AI RECOMMENDATION ARCHITECTURE FOR MEDICAL PRESCRIPTIONS

Application No. 18/222,392

• Filed: Jul 14, 2023
• Examiner: AKOGYERAM II, NICHOLAS A
• Art Unit: 3686
• Tech Center: 3600 — Transportation & Electronic Commerce
• Assignee: ORACLE INTERNATIONAL Corporation
• OA Round: 3 (Non-Final)

Office Action

DETAILED ACTION

Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .

Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on September 10, 2026 has been entered.

Status of Claims
Claims 1, 4, 6-9, 11, 14, 16-19, and 21-28, as recited in an amendment filed on February 12, 2026, were previously pending and subject to a final office action filed on June 16, 2026 (the “June 16, 2026 Final Office Action”). On July 17, 2026, Applicant filed a response to the June 16, 2026 Final Office Action, which: (i) amended claims 1 and 11; (ii) canceled claims 26 and 27; and (iii) added new claims 29 and 30 (the “July 17, 2026 After Final Response). On July 30, 2026, an Advisory Action was mailed to Applicant advising that the amendments in the “July 17, 2026 After Final Response would not be entered, because: (1) they raised new issues that would require further consideration and/or search; and (2) they were not deemed to place the application in better form for appeal by materially reducing or simplifying the issues for appeal.
On September 10, 2026, Applicant filed a Request for Continued Examination in accordance with 37 CFR 1.114, where Applicant resubmitted: (i) the amendments to claims 1 and 11; (ii) the cancellations of claims 26 and 27; (iii) the addition of new claims 29 and 30; and (iv) requested entry and consideration of claims 1, 4, 6-9, 11, 14, 16-19, 21-25, and 28-30 (the “September 10, 2026 RCE”). As such, claims 1, 4, 6-9, 11, 14, 16-19, 21-25, and 28-30, as recited in the September 10, 2026 RCE, are entered, currently pending, and subject to the non-final office action below.

Response to Applicant’s Remarks
Response to Applicant’s Remarks Concerning Rejections under 35 U.S.C. § 101
Applicant’s arguments, see Applicant’s Remarks, pp. 13-15, Claim Rejections - 35 U.S.C. §101 Section, filed September 10, 2026, with respect to rejections of claim 1, 4, 6-9, 11, 14, 16-19, and 21-28 under 35 U.S.C. § 101 have been fully considered, but they are not persuasive. Further, in light of the 2019 Revised Patent Subject Matter Eligibility Guidance (available at MPEP § 2106) (the “2019 Revised PEG”), the § 101 rejections of claims 1, 4, 6-9, 11, 14, 16-19, and 21-25 are maintained and the § 101 rejections of new claims 29 and 30 are added in this final office action.
Specifically, Applicant argues that the claims provide an improvement in the functioning of a computer by explaining that “traditional item selection models are not capable of incorporating a large text corpus into modeling primarily due to significant latency at training and inference, and also due to scalability and that the embodiments leverage an LLM/ML language model to convert unstructured summary note data into structured data that a ML recommendation model can consume”. See Applicant’s Remarks, at p. 13-14. Examiner respectfully disagrees. When evaluating whether claims recite an improvement to the functioning of a computer or a technical field, the disclosure must provide sufficient details such that one of ordinary skill in the art would recognize the claimed invention as providing an improvement. MPEP § 2106.05(a). The specification need not explicitly set forth the improvement, but it must describe the invention such that the improvement would be apparent to one of ordinary skill in the art. Id.
Conversely, if the specification explicitly sets forth an improvement but in a conclusory manner (i.e., a bare assertion of an improvement without the detail necessary to be apparent to a person of ordinary skill in the art), the examiner should not determine the claim improves technology. An indication that the claimed invention provides an improvement can include a discussion in the specification that identifies a technical problem and explains the details of an unconventional technical solution expressed in the claim, or identifies technical improvements realized by the claim over the prior art. After the examiner has consulted the specification and determined that the disclosed invention improves technology, the claim must be evaluated to ensure the claim itself reflects the disclosed improvement in technology. MPEP § 2106.05(a). That is, the claim must include the components or steps of the invention that provide the improvement described in the specification. Id.
In the present case, Applicant’s claims do not include components or steps of the invention that provide the improvement that Applicant argues for in the remarks. That is, the claims do not include a step directed to converting data from an unstructured format into a structured format. For example, while the claims recite a step directed to receiving the set of feature values as structured data from the ML language model, the claims do not recite receiving any unstructured data or converting any unstructured data into a structured format. Specifically, independent claims 1 and 11 merely recite receiving summary not data and generating a set of values from the summary note data using a machine-learned (ML) language model. This does not amount to an improvement to technology or a technical field, because these steps merely amount to receiving data and outputting data using a generic language model (i.e., Applicant has not disclosed the algorithm or steps that the model goes through in order to generate the set of feature values). Even if the claims did recite converting unstructured data into a structured format, a person could practically perform such conversion in his/her mind with pen and paper. For instance, in the case where a doctor’s unstructured notes included the text Condition: Diabetes - Type 2; Complications: Hyperglycemia, a person could easily convert such text into the JSON format {"condition": "Diabetes - Type 2", "complications": "hyperglycemia"}. These arguments were addressed in a previous office action as well.
Further, as discussed in an interview between the Examiner and Applicant’s attorneys on July 16, 2026, the Examiner suggested that in regard to potentially addressing the § 101 rejections, it would be helpful to define the ML models with some specificity. As the claims are currently drafted, the ML models are merely added to analyze or generate data. Then the generated data is merely displayed on a screen. This does not amount to a technical improvement, but rather is similar to applying an abstract idea on a generic computer device/insignificant extra solution activity. It would be helpful when claiming a technical improvement to amend the claims to specify or demonstrate improvements to the models themselves, such as, the ML architecture, the training processes, or inference mechanisms (e.g., something that enhances the performance of the models); or include specific training steps and applying the trained model to perform some relevant function that demonstrates an improvement to the function. The claims are lacking that currently, because the claims merely recite that the various ML models are “trained on first, second, or third training” (however, it is not specified what this training data is) and that the ML models use “first, second, or third machine learning techniques” (although, it is not specific what machine learning techniques are used). Therefore, the ML models are still claimed at a high level of generality and this argument is not persuasive for these reasons.
For these reasons, this argument is not persuasive. Please see the amended rejections under the Claim Rejections – 35 U.S.C. § 101 Section below, for further clarification and complete analysis.

Response to Applicant’s Remarks Concerning Rejections under 35 U.S.C. § 103
Applicant’s arguments, see Applicant’s Remarks, pp. 9-13, Claim Rejections Section, filed September 10, 2026, with respect to rejections of claim 1, 4, 6-9, 11, 14, 16-19, and 21-28 under 35 U.S.C. § 103 have been fully considered, but they are not persuasive. Applicant argues that Amarasingham et al. (Pub. No. US 2023/0104655) does not teach the “ML reinforcement model that is trained on a machine learning technique”. See Applicant’s Remarks, at p. 10. The Examiner disagrees.
In the present case, since the ML reinforcement learning is not well-defined in the claims, Amarasingham is deemed to teach a ML reinforcement learning model that ranks candidate medical items. Specifically, paragraph [0054] in Amarasingham generally teaches that the knowledge graph (KG) 214 may also rank which medical procedures and/or treatments are more important than others given a particular medical condition and/or symptoms, where paragraph [0065] teaches that the treatments, pending diagnostic orders, and active treatment orders are prioritized in a ranked list with a prioritization module (i.e., generating, by an ML reinforcement model, a ranking of the plurality of candidate medical items based on the plurality of candidate medical items received from the ML recommendation model, where the prioritization module is interpreted as the equivalent of an ML reinforcement learning model because one of ordinary skill in the arts would recognize that prioritization and optimization models are types of reinforcement learning models). Further, paragraph [0066] in Amarasingham teaches that the prioritization module is trained to exclude or remove the identified treatment orders and diagnostic orders which are no longer required or have been fulfilled from the ranked list (i.e., the ML reinforcement model is trained on third training data), where paragraph [0038] generally teaches that each component of the AI engine 112 [which includes the prioritization module] may be trained to perform a particular task or combination of tasks, and the training may occur using historical patient data or other types of datasets, including datasets that may be accessible over network 106 of FIG. 1 (i.e., the ML reinforcement model is trained on third training data). Paragraph [0039] in Amarasingham also teaches that the AI engine may include various models, such as, recurrent neural network (RNN) models, neural net models including convolutional neural network (CNN) models, Bidirectional Encoder Representations from Transformers (BERT) models or their variants, and/or Generative Pre-trained Transformer (GPT) models, GPT-2 models, GPT-3 models and the like (i.e., the ML reinforcement model may use a third machine learning technique).
Based on these paragraphs, one of ordinary skill in the art would recognize that Amarasingham teaches the amended limitations. As described in the response to the § 101 rejections, in addressing the § 103 rejections, it may be helpful to described the specific structure and the specific training steps of the ML reinforcement learning model in order to distinguish Applicant’s ML reinforcement learning model over the cited prior art. Therefore, the § 103 rejections of claims 1 and 11 are updated in this office action. Please see the rejections under the Claim Rejections – 35 U.S.C. § 103 Section below, for further clarification and complete analysis.

Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.

Claims 1, 4, 6-9, 11, 14, 16-19, 21-25, and 28-30 are rejected under 35 U.S.C. §101 because the claimed invention is directed to an abstract idea without significantly more. See MPEP § 2106 (hereinafter referred to as the “2019 Revised PEG”).

Step 1 of the 2019 Revised PEG
Following Step 1 of the 2019 Revised PEG, claims 1, 4, 6-9, 21-25, 29, and 30 are directed to a method, which is within one of the four statutory categories (i.e., a process). See MPEP § 2106.03. Claims 11, 14, 16-19, and 28 are directed to one or more non-transitory storage media, which is also within one of the four statutory categories (i.e., a manufacture). See id.

Step 2A of the 2019 Revised PEG - Prong One
Following Prong One of Step 2A of the 2019 PEG, the claim limitations are to be analyzed to determine whether they “recite” a judicial exception or in other words whether a judicial exception is “set forth” or “described” in the claims. See MPEP §2106.04. An “abstract idea” judicial exception is subject matter that falls within at least one of the following groupings: (1) Mathematical Concepts; (2) Certain Methods of Organizing Human Activity, and (3) Mental Processes. See MPEP § 2106.04(a).
Claims 1, 4, 6-9, 11, 14, 16-19, 21-25, and 28-30 are rejected under 35 U.S.C. § 101, because the claimed invention is directed to an abstract idea without significantly more. Representative independent claims 1 and 11 include limitations that recite an abstract idea. Note that independent claim 1 is a method, while claim 11 covers the matching one or more non-transitory storage media. Specifically, independent claim 11 recites the following limitations:
One or more non-transitory storage media storing instructions which, when executed by one or more computing devices, cause:

receiving summary note data that is composed by a physician for a patient;

generating, by a machine-learned (ML) language model, based on the summary note data, a set of feature values, where the set of feature values comprise a plurality of values, each of the plurality of values corresponding to a respective feature of a plurality of features extracted from the summary note, where the plurality of features includes one or more of a potential diagnosis, a risk factor, or potential treatment, and where the set of feature values are stored as structured data, wherein the ML language model is trained on first training data using a first machine learning technique;

identifying a profile of the patient and a profile of the physician;

determining, by an ML recommendation model, based on the profile of the patient, the profile of the physician, and the set of feature values, a plurality of candidate medical items, where the set of feature values are received by the ML recommendation model as the structured data from the ML language model, wherein the ML recommendation model is trained on second training data using a second machine learning technique;

generating, by an ML reinforcement learning model, a ranking of the plurality of candidate medical items based on the plurality of candidate medial items received from the ML recommendation model, wherein the ML reinforcement learning model is trained on third training data using a third machine learning technique;

causing a subset of the plurality of candidate medical items to be presented on a screen of a computing device based on the ranking.

However, the Examiner submits that the foregoing underlined limitations constitute a process that, under its broadest reasonable interpretation, falls within the “Mental Processes” grouping of abstract ideas. See 2019 Revised PEG. The Mental Processes category covers concepts which are capable of being performed in the human mind or encompasses a human performing the step(s) mentally with the aid of a pen and paper (including an observation, evaluation, judgment, or opinion) (i.e., a method, comprising: generating a set of feature values based on summary note data that is composed by a physician for a patient; identifying a profile of the patient and a profile of the physician; determining a plurality of candidate medical items based on the profile of the patient, the profile of the physician, and the set of feature values; and generating a ranking of the plurality of candidate medical items). See MPEP § 2106.04(a)(2)(III). That is, other than reciting some computer components and functions (the foregoing limitations in claim 11 which are not underlined), the context of claims 1 and 11 encompasses concepts that are capable of being performed in the human mind or encompasses a human performing the step(s) mentally with the aid of a pen and paper (including an observation, evaluation, judgment, and/or opinion) (i.e., a method, comprising: generating a set of feature values based on summary note data that is composed by a physician for a patient; identifying a profile of the patient and a profile of the physician; determining a plurality of candidate medical items based on the profile of the patient, the profile of the physician, and the set of feature values; and generating a ranking of the plurality of candidate medical items).
The aforementioned claim limitations described in claims 1 and 11 are analogous to claim limitations directed toward concepts which are capable of being performed in the human mind or encompasses a human performing the step(s) mentally with the aid of a pen and paper, because they merely recite limitations which encompass a person mentally and/or manually: (1) generating a set of feature values based on summary note data that is composed by a physician for a patient (i.e., a type of observation, evaluation, judgment, and/or opinion where a person could mentally and/or manually generate a set of feature values based on summary note data that is composed by a physician for a patient); (2) ; identifying a profile of the patient and a profile of the physician (i.e., a type of observation, evaluation, judgment, and/or opinion where a person could mentally and/or identify a patient profile and a physician profile); (3) determining a plurality of candidate medical items based on the profile of the patient, the profile of the physician, and the set of feature values (i.e., a type of observation, evaluation, judgment, and/or opinion where a person could manually and/or manually determine a plurality of candidate medical items based on the patient profile, physician profile, and generated set of feature values); and (4) generating a ranking of the plurality of candidate medical items (i.e., a type of observation, evaluation, judgment, and/or opinion where a person could mentally and/or manually arrange/order/rank the generated plurality of candidate medical items).
Further, Applicant’s claims are similar to claims which have been held to recite an abstract mental process. For example, the Federal Circuit held that a claim directed to “collecting information, analyzing it, and displaying certain results of the collection and analysis”, where the data analysis steps are recited at a high level of generality amounted to steps that could practically be performed in the human mind. See MPEP § 2106.04(a)(2)(III)(A) (citing Electric Power Group v. Alstom, S.A.). Similarly, Applicant’s claims recite steps for collecting information (i.e., the steps directed to generating/determining the set of feature values, candidate medical items, and ranking of the plurality of candidate medical items); analyzing the data (i.e., identifying a patient profile and a physician profile); and displaying certain results about the collection and analysis (i.e., causing a subset of the plurality of candidate medical items to be presented on a screen of a computing device based on the ranking), at a high level of generality. Therefore, the aforementioned underlined claim limitations may reasonably be interpreted as mental/manual observations, evaluations, judgments, and/or opinions made by a person, such as healthcare provider. If a claim limitation, under its broadest reasonable interpretation, covers concepts which are capable of being performed in the human mind or encompasses a human performing the step(s) mentally with the aid of a pen and paper, then it falls within the “Mental Processes” grouping of abstract ideas. See 2019 Revised PEG. Accordingly, claims 1 and 11 recite an abstract idea that falls within the Mental Processes category.
Furthermore, Examiner notes that dependent claims 4, 6-9, 14, 16-19, 21-25, and 28-30 further define the at least one abstract idea (and thus fail to make the abstract idea any less abstract) as set forth below. Examiner notes that dependent claims 4, 14, 21, 22, and 25 do not provide any limitations that are deemed to be additional elements which require further analysis under Prong Two of Step 2A; and dependent claims 6-9, 16-19, 23, 24, and 28-30 include limitations that are deemed to be additional elements, and require further analysis under Prong Two of Step 2A. For example, claims 4 and 14 merely recite the additional mental steps for identifying one or more medical items that were previously prescribed to the patient and determining the plurality of candidate medical items based on the one or more medical items (i.e., these steps are deemed to be reasonably performed mentally, because they add additional steps that could be observations, evaluations, judgments, and/or opinions made by a person, such as a healthcare professional). Further, claims 21, 22, and 25 merely recite additional mental steps for increasing or decreasing a score of a candidate medical item that became available within a freshness window, or basing the ranking of the candidate medical items on relevance scores (i.e., similarly, these steps are deemed to be reasonably performed mentally, because they add additional steps that could be observations, evaluations, judgments, and/or opinions made by a person, such as a healthcare professional).

Step 2A of the 2019 Revised PEG – Prong Two
Regarding Prong Two of Step 2A of the 2019 Revised PEG, it must be determined whether the claim as a whole integrates the abstract idea into a practical application. As noted in the 2019 Revised PEG, it must be determined whether any additional elements in the claims are indicative of integrating the abstract idea into a practical application in a manner that imposes a meaningful limit on the judicial exception. The courts have indicated that additional elements merely using a computer to implement an abstract idea, adding insignificant extra solution activity, or generally linking use of a judicial exception to a particular technological environment or field of use do not integrate a judicial exception into a “practical application.” See MPEP §§ 2106.05 (f)-(h).
In the present case, for independent claim 11, the additional limitations beyond the above-noted at least one abstract idea are as follows (where the bolded portions are the “additional limitations” while the underlined portions continue to represent the at least one “abstract idea”):
One or more non-transitory storage media storing instructions (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f)) which, when executed by one or more computing devices (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f)), cause:

receiving summary note data that is composed by a physician for a patient (adding insignificant extra-solution activity as noted below, see MPEP § 2106.05(g); the Examiner further submits that such steps are not unconventional as they merely consist of receiving or transmitting data over a network, as evidenced by the Intellectual Ventures v. Symantec case, as noted below in the Step 2B Analysis Section, see MPEP § 2106.05(d));

generating, by a machine-learned (ML) language model (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)), based on the summary note data, a set of feature values, where the set of feature values comprise a plurality of values, each of the plurality of values corresponding to a respective feature of a plurality of features extracted from the summary note, where the plurality of features includes one or more of a potential diagnosis, a risk factor, or potential treatment, and where the set of feature values are stored as structured data (adding insignificant extra-solution activity as noted below, see MPEP § 2106.05(g); the Examiner further submits that such steps are not unconventional as they merely consist of storing and retrieving information in memory, as evidenced by the Versata Dev. Group, Inc. v. SAP Am., Inc. case, as noted below in the Step 2B Analysis Section, see MPEP § 2106.05(d)), wherein the ML language model is trained on first training data using a first machine learning technique (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h));

identifying a profile of the patient and a profile of the physician;

determining, by an ML recommendation model (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)), based on the profile of the patient, the profile of the physician, and the set of feature values, a plurality of candidate medical items, where the set of feature values are received by the ML recommendation model as the structured data from the ML language model, wherein the ML recommendation model is trained on second training data using a second machine learning technique (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h));

generating, by an ML reinforcement learning model (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)), a ranking of the plurality of candidate medical items based on the plurality of candidate medial items received from the ML recommendation model, wherein the ML reinforcement learning model is trained on third training data using a third machine learning technique (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h));

causing a subset of the plurality of candidate medical items to be presented on a screen of a computing device based on the ranking (adding insignificant extra-solution activity as noted below, see MPEP § 2106.05(g); the Examiner further submits that such steps are not unconventional as they merely consist of receiving or transmitting data over a network, as evidenced by the Intellectual Ventures v. Symantec case, as noted below in the Step 2B Analysis Section, see MPEP § 2106.05(d)).

However, the recitation of these generic computer components and functions in claims 1 and 11 are recited at a high-level of generality (i.e., using a generic computer device to perform the abstract idea of: a method, comprising: generating a set of feature values based on summary note data that is composed by a physician for a patient; identifying a profile of the patient and a profile of the physician; determining a plurality of candidate medical items based on the profile of the patient, the profile of the physician, and the set of feature values; and generating a ranking of the plurality of candidate medical items), such that it amounts to no more than: (1) adding the words “apply it” (or is the equivalent of) with the judicial exception; mere instructions to implement an abstract idea on a computer; or merely uses a computer as a tool to perform an abstract idea; (2) adding insignificant extra-solution activity to the judicial exception; and (3) generally linking the use of a judicial exception to a particular technological environment or field of use. See MPEP §§ 2106.05(f)-(h). For the following reasons, the Examiner submits that the above identified additional limitations do not integrate the above-noted at least one abstract idea into a practical application.
- The following is an example of court decisions that demonstrate merely applying instructions by reciting the computer structure as a tool to implement the claimed limitations (e.g., see MPEP § 2106.05(f)):
- Invoking computers or other machinery merely as a tool to perform an existing process, e.g. see, Affinity Labs v. DirecTV – similarly, the current invention invokes computers (i.e., the one or more non-transitory storage media storing instructions; one or more computing devices; machine-learned (ML) language model; ML recommendation model; ML reinforcement learning model; and screen of a computing device) to perform the existing processes of: a person generating a set of feature values based on summary note data that is composed by a physician for a patient; identifying a profile of the patient and a profile of the physician; determining a plurality of candidate medical items based on the profile of the patient, the profile of the physician, and the set of feature values; and generating a ranking of the plurality of candidate medical items); and
- Using a computer or other machinery in its ordinary capacity for economic or other tasks (e.g., to receive, store, or transmit data), e.g. see, TLI Communications LLC v. AV Auto, LLC – similarly, the current invention invokes the one or more non-transitory storage media storing instructions; one or more computing devices; machine-learned (ML) language model; ML recommendation model; ML reinforcement learning model; and screen of a computing device, in their ordinary capacity to process and transmit data generated by the abstract mental process.
- Requiring the use of software to tailor information and provide it to the user on a generic computer, e.g., see Intellectual Ventures I LLC v. Capital One Bank (USA) – similarly, the current invention requires configuring the computing system (i.e., where the one or more non-transitory storage media storing instructions; one or more computing devices; machine-learned (ML) language model; ML recommendation model; ML reinforcement learning model; and screen of a computing device are deemed to be the equivalent of implementing generic software) to tailor information (e.g., generate the ranking of the plurality of candidate medical items) and provide it to the user on a generic computer (e.g., cause a subset of ranked candidate medical items to be presented on a screen of the one or more computing devices).
- The following is an example of an insignificant extra-solution activity (e.g., see MPEP § 2106.05(g)):
- Examples of Mere Data Gathering/Mere Data Outputting:
- Obtaining information about transactions using the Internet to verify credit card transactions, e.g., see CyberSource v. Retail Decisions, Inc. – similarly, the step directed to: “receiving summary note data that is composed by a physician for a patient”, described in claims 1 and 11, is a necessary data gathering step (i.e., “receiving summary note data that is composed by a physician for a patient” is a necessary data gathering step in order to collect the data that is later used to perform the abstract idea, namely, generating the set of feature values and candidate medical items.).
- Consulting and updating an activity log, e.g., see Ultramercial, Inc. v. Hulu, LLC – similarly, the step directed to: “where the set of feature values are stored as structured data”, described in claims 1 and 11, is deemed to be necessary data gathering, because this step does not add a meaningful limit to the process of generating the ranking of the plurality of candidate medical items (i.e., storing the set of feature values as structured data represents necessary data gathering, because it is the equivalent of updating data in a database/activity log.).
- Printing or downloading generated menus, e.g., see Apple, Inc. v. Ameranth, Inc. – similarly, the step directed to: “causing a subset of the plurality of candidate medical items to be presented on a screen of a computing device based on the ranking”, described in claims 1 and 11, is deemed to be necessary data outputting, because this step does not add a meaningful limit to the process of generating the ranking of the plurality of candidate medical items (i.e., presenting a subset of the plurality of candidate medical items on a screen of a computing device based on the ranking represents necessary data outputting, because it presenting/displaying data is necessary in order to convey this information to a user.).
- The following is an example of limitations that courts have described as merely indicating a field of use or technological environment in which to apply a judicial exception (e.g., see MPEP § 2106.05(h)):
- Specifying that the abstract idea of monitoring audit log data relates to transactions or activities that are executed in a computer environment, because this requirement merely limits the claims to the computer field, i.e., to execution on a generic computer, e.g., see FairWarning v. Iatric Sys., - similarly, using the machine-learned (ML) language model; ML recommendation model; ML reinforcement learning model to perform the abstract mental steps; and the steps directed to “where the set of feature values are received by the ML recommendation model as the structured data from the ML language model”; “wherein the ML language model is trained on first training data using a first machine learning technique”; “wherein the ML recommendation model is trained on second training data using a second machine learning technique”; and “wherein the ML reinforcement learning model is trained on third training data using a third machine learning technique”, described in claims 1 and 11, merely specifies that the aforementioned abstract concepts of: (i) generating a set of feature values based on summary note data that is composed by a physician for a patient; (ii) determining a plurality of candidate medical items based on the profile of the patient, the profile of the physician, and the set of feature values; and (iii) generating a ranking of the plurality of candidate medical items, are executed in a computer environment, because they merely limit the claims to the field of machine learning technology.
Thus, the additional elements in independent claims 1 and 11 are not indicative of integrating the judicial exception into a practical application. Similarly, dependent claims 4, 14, 21, 22, and 25 do not recite any additional elements outside of those identified as being directed to the abstract idea described above. Examiner notes that dependent claims 6-9, 16-19, 23, 24, and 28-30 recite the following additional elements identified in bold font below (with limitations deemed to be part of the above identified abstract idea identified in underlined font):
receiving, from a user, input that indicates that one or more candidate medical items in the subset are incorrect recommendations (adding insignificant extra-solution activity as noted below, see MPEP § 2106.05(g); the Examiner further submits that such steps are not unconventional as they merely consist of receiving or transmitting data over a network, as evidenced by the Intellectual Ventures v. Symantec case, as noted below in the Step 2B Analysis Section, see MPEP § 2106.05(d)); updating the ML reinforcement learning model based on the input (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)) (as described in claims 6 and 16);

receiving, from a user, input that indicates one or more reasons why one or more candidate medical items in the subset are incorrect recommendations (adding insignificant extra-solution activity as noted below, see MPEP § 2106.05(g); the Examiner further submits that such steps are not unconventional as they merely consist of receiving or transmitting data over a network, as evidenced by the Intellectual Ventures v. Symantec case, as noted below in the Step 2B Analysis Section, see MPEP § 2106.05(d)); updating the ML language model based on the input (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)) (as described in claims 7 and 17);

receiving, from a user, input that indicates one or more reasons why one or more candidate medical items in the subset are incorrect recommendations (adding insignificant extra-solution activity as noted below, see MPEP § 2106.05(g); the Examiner further submits that such steps are not unconventional as they merely consist of receiving or transmitting data over a network, as evidenced by the Intellectual Ventures v. Symantec case, as noted below in the Step 2B Analysis Section, see MPEP § 2106.05(d)); updating the summary note data based on the input to generate an updated summary note data; generating, by the ML language model (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)), based on the updated summary note data, a second set of feature values; generating, by the ML recommendation model (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)), based on the profile of the patient, the profile of the physician, and the second set of feature values, a second plurality of candidate medical items; generating, by the ML reinforcement learning model (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)), a particular ranking of the second plurality of candidate medical items; causing a subset of the second plurality of candidate medical items to be presented on the screen of the computing device based on the particular ranking (adding insignificant extra-solution activity as noted below, see MPEP § 2106.05(g); the Examiner further submits that such steps are not unconventional as they merely consist of receiving or transmitting data over a network, as evidenced by the Intellectual Ventures v. Symantec case, as noted below in the Step 2B Analysis Section, see MPEP § 2106.05(d)) (as described in claim 8 and 18);

receiving the summary note data is performed by a prompt engine (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)); the method further comprising: generating a set of instructions, and causing the ML language model to access the set of instructions and the summary note data (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)); generating the set of feature values comprises generating the set of feature values also based on the set of instructions (as described in claims 9 and 19);

wherein the set of feature values includes potential treatments; wherein the method further comprises in a cold-start scenario, training the ML recommendation model using the potential treatments as label data (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)) (as described in claim 23);

storing prior medical items in an order database, each stored prior medical item including a timestamp (adding insignificant extra-solution activity as noted below, see MPEP § 2106.05(g); the Examiner further submits that such steps are not unconventional as they merely consist of storing and retrieving information in memory, as evidenced by the Versata Dev. Group, Inc. v. SAP Am., Inc. case, as noted below in the Step 2B Analysis Section, see MPEP § 2106.05(d)); identifying a temporal order of the medical items; providing the temporal order as an input to the ML recommendation model (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)); wherein the plurality of candidate medical items is determined based on the temporal order (as described in claims 24 and 28);

receiving an intervention corresponding to a ranking of the subset of the plurality of candidate medical items (adding insignificant extra-solution activity as noted below, see MPEP § 2106.05(g); the Examiner further submits that such steps are not unconventional as they merely consist of receiving or transmitting data over a network, as evidenced by the Intellectual Ventures v. Symantec case, as noted below in the Step 2B Analysis Section, see MPEP § 2106.05(d)); penalizing, based on the intervention, a reward corresponding to the reward function for the ranking generated by the ML reinforcement learning model (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)); and updating the ML reinforcement learning model based on the penalized reward (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)) (as described in claim 29); and

receiving, from the physician, a selection of a candidate medical item from the subset of the plurality of candidate medical items (adding insignificant extra-solution activity as noted below, see MPEP § 2106.05(g); the Examiner further submits that such steps are not unconventional as they merely consist of receiving or transmitting data over a network, as evidenced by the Intellectual Ventures v. Symantec case, as noted below in the Step 2B Analysis Section, see MPEP § 2106.05(d)); storing, in an order database, data identifying the physician, the patient, and the selected candidate medical item (adding insignificant extra-solution activity as noted below, see MPEP § 2106.05(g); the Examiner further submits that such steps are not unconventional as they merely consist of storing and retrieving information in memory, as evidenced by the Versata Dev. Group, Inc. v. SAP Am., Inc. case, as noted below in the Step 2B Analysis Section, see MPEP § 2106.05(d)); generating the third training data for the ML reinforcement learning model from data stored in the order database, the third training data comprising a plurality of training instances, each training instance comprising (i) input feature data comprising a set of scores for a list of candidate medical items and (ii) label data comprising the selected candidate medical item (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)); and training the ML reinforcement learning model based on the third training data (the Examiner submits that this additional element amounts to adding the words “apply it” (or an equivalent), or mere instructions to implement the abstract idea on a computer, see MPEP § 2106.05(f); and generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP § 2106.05(h)) (as described in claim 30).

As such, the additional elements in claims 1, 6-9, 11, 16-19, 23, 24, and 28-30 are not indicative of integrating the judicial exception into a practical application. Looking at the additional limitations as an ordered combination adds nothing that is not already present when looking at the elements taken individually. For instance, unlike the claims that have been held as a whole to be directed to an improvement or otherwise directed to something more than the abstract idea, claims 1, 4, 6-9, 11, 14, 16-19, 21-25, and 28-30: (1) are not directed to improvements to the functioning of a computer, or to any other technology or technical field similar to the Enfish, LLC v. Microsoft Corp. case (see MPEP § 2106.05(a)); (2) do not apply or use a judicial exception to effect a particular treatment or prophylaxis for a disease or medical condition (see MPEP § 2106.04(d)(2)); (3) do not apply the judicial exception with, or by use of, a particular machine (see MPEP § 2106.05(b)); (4) do not effect a transformation or reduction of a particular article to a different state or thing (see MPEP § 2106.05(c)); nor do they (5) apply or use the judicial exception in some other meaningful way beyond generally linking the use of the judicial exception to a particular technological environment, such that the claim as whole is more than a drafting effort designed to monopolize the exception (see MPEP § 2106.05(e) and MPEP § 2106.04(d)(2)). For these reasons, claims 1, 4, 6-9, 11, 14, 16-19, 21-25, and 28-30 do not recite additional elements that integrate the judicial exception into a practical application.

Step 2B of the 2019 Revised PEG
Regarding Step 2B of the 2019 Revised PEG, claims 1, 4, 6-9, 11, 14, 16-19, 21-25, and 28-30 do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, with respect to integration of abstract idea into a practical application, the additional elements of claims 1, 6-9, 11, 16-19, 23, 24, and 28-30 amount to no more than: (1) adding the words “apply it” (or is the equivalent of) with the judicial exception; mere instructions to implement an abstract idea on a computer; or merely uses a computer as a tool to perform an abstract idea; (2) adding insignificant extra-solution activity to the judicial exception; and (3) generally linking the use of a judicial exception to a particular technological environment or field of use. See MPEP §§ 2106.05(f)-(h). Further the additional elements, other than the abstract idea per se, when considered both individually and as an ordered combination, amount to no more than limitations consistent with what the courts recognize, or those having ordinary skill in the art would recognize, to be well-understood, routine, and conventional computer components. See MPEP § 2106.05 (d).
Specifically, the Examiner submits that the additional elements of claims 1, 6-9, 11, 16-19, 23, 24, and 28-30, as recited, the one or more non-transitory storage media storing instructions; one or more computing devices; machine-learned (ML) language model; ML recommendation model; ML reinforcement learning model; screen of a computing device; order database; prompt engine; and the steps directed to: “receiving summary note data that is composed by a physician for a patient”; “where the set of feature values are stored as structured data”; “wherein the ML language model is trained on first training data using a first machine learning technique”; “where the set of feature values are received by the ML recommendation model as the structured data from the ML language model”; “wherein the ML recommendation model is trained on second training data using a second machine learning technique”; “wherein the ML reinforcement learning model is trained on third training data using a third machine learning technique”; “causing a subset of the plurality of candidate medical items to be presented on a screen of a computing device based on the ranking”; “receiving, from a user, input that indicates that one or more candidate medical items in the subset are incorrect recommendations”; “updating the ML reinforcement learning model based on the input”; “receiving, from a user, input that indicates one or more reasons why one or more candidate medical items in the subset are incorrect recommendations”; “updating the ML language model based on the input”; “receiving, from a user, input that indicates one or more reasons why one or more candidate medical items in the subset are incorrect recommendations”; “causing a subset of the second plurality of candidate medical items to be presented on the screen of the computing device based on the particular ranking”; “receiving the summary note data is performed by a prompt engine”; “causing the ML language model to access the set of instructions and the summary note data”; “wherein the method further comprises in a cold-start scenario, training the ML recommendation model using the potential treatments as label data”; “storing prior medical items in an order database, each stored prior medical item including a timestamp”; “providing the temporal order as an input to the ML recommendation model”; “receiving an intervention corresponding to a ranking of the subset of the plurality of candidate medical items”; “penalizing, based on the intervention, a reward corresponding to the reward function for the ranking generated by the ML reinforcement learning model”; “updating the ML reinforcement learning model based on the penalized reward”; “receiving, from the physician, a selection of a candidate medical item from the subset of the plurality of candidate medical items”; “storing, in an order database, data identifying the physician, the patient, and the selected candidate medical item”; “generating the third training data for the ML reinforcement learning model from data stored in the order database, the third training data comprising a plurality of training instances, each training instance comprising (i) input feature data comprising a set of scores for a list of candidate medical items and (ii) label data comprising the selected candidate medical item”; and “training the ML reinforcement learning model based on the third training data”, are well-understood, routine, and conventional functions. See MPEP § 2106.05(d)(II).
- In regard to the one or more non-transitory storage media storing instructions; one or more computing devices; machine-learned (ML) language model; ML recommendation model; ML reinforcement learning model; screen of a computing device; order database; prompt engine; and the steps directed to: “wherein the ML language model is trained on first training data using a first machine learning technique”; “where the set of feature values are received by the ML recommendation model as the structured data from the ML language model”; “wherein the ML recommendation model is trained on second training data using a second machine learning technique”; “wherein the ML reinforcement learning model is trained on third training data using a third machine learning technique”; “updating the ML reinforcement learning model based on the input”; “updating the ML language model based on the input”; “receiving the summary note data is performed by a prompt engine”; “causing the ML language model to access the set of instructions and the summary note data”; “wherein the method further comprises in a cold-start scenario, training the ML recommendation model using the potential treatments as label data”; “providing the temporal order as an input to the ML recommendation model”; “penalizing, based on the intervention, a reward corresponding to the reward function for the ranking generated by the ML reinforcement learning model”; “updating the ML reinforcement learning model based on the penalized reward”; “generating the third training data for the ML reinforcement learning model from data stored in the order database, the third training data comprising a plurality of training instances, each training instance comprising (i) input feature data comprising a set of scores for a list of candidate medical items and (ii) label data comprising the selected candidate medical item”; and “training the ML reinforcement learning model based on the third training data”, these additional elements or combination of elements in the claims, other than the abstract idea per se, amount to no more than well-understood, routine, and conventional activities previously known to the industry, because:
- Applicant’s disclosure supports this assertion. For example, Applicant describes the one or more non-transitory storage media as “optical disks, magnetic disks, or solid-state drives, such as storage device […] a floppy disk, a flexible disk, hard disk, solid-state drive, magnetic tape, or any other magnetic data storage medium, a CD-ROM, any other optical data storage medium, any physical medium with patterns of holes, […] any other memory chip or cartridge”. Applicant’s specification as filed on July 14, 2023, paragraph [0074]. Further, Applicant describes the computing device as a client device, such as a “desktop computer, a laptop computer, a tablet computer, a watch or other wearable device, or any other portable computing device.” Applicant’s specification as filed on July 14, 2023, paragraphs [0019] and [0057]. Still further, Applicant describes the various machine learning models as not limited to any particular machine learning technique, which “include linear regression, logistic regression, random forests, naive Bayes, and Support Vector Machines (SVMs )” and “supervised while others may be unsupervised”. Applicant’s specification as filed on July 14, 2023, paragraph [0015]. Next, Applicant describes the prompt engine as being “implemented in a software, hardware, or any combination of software and hardware”. Applicant’s specification as filed on July 14, 2023, paragraph [0022]. Lastly, Applicant describes the order database as “stor[ing] data above previous orders for medical items, such as, medicine, tests, or other treatments”. Applicant’s specification as filed on July 14, 2023, paragraph [0036]. These passages in Applicant’s specification indicate that the various computer components and machine learning models are conventional in nature (i.e., well-understood, routine, and conventional computer devices and software), because the descriptions in the specification are recited at high-level of generality/represent generic computer components and functions. Therefore, the Examiner submits that these additional elements represent well-understood, routine, and conventional computer devices and functions which are known in the medical industry.
- The Examiner submits that these limitations amount to merely using a computer or other machinery as tools for performing their typical functionality in conjunction with performing the above-noted at least one abstract idea (see MPEP § 2106.05(f) and analysis of these limitations under Step 2A, Prong Two above).
- The Examiner submits that these limitations generally link the use of the judicial exception to a particular technological environment or field of use – for example, the limitations directed to: machine-learned (ML) language model; ML recommendation model; ML reinforcement learning model; and the steps directed to: “wherein the ML language model is trained on first training data using a first machine learning technique”; “where the set of feature values are received by the ML recommendation model as the structured data from the ML language model”; “wherein the ML recommendation model is trained on second training data using a second machine learning technique”; “wherein the ML reinforcement learning model is trained on third training data using a third machine learning technique”; “updating the ML reinforcement learning model based on the input”; “updating the ML language model based on the input”; “receiving the summary note data is performed by a prompt engine”; “causing the ML language model to access the set of instructions and the summary note data”; “wherein the method further comprises in a cold-start scenario, training the ML recommendation model using the potential treatments as label data”; “providing the temporal order as an input to the ML recommendation model”; “penalizing, based on the intervention, a reward corresponding to the reward function for the ranking generated by the ML reinforcement learning model”; “updating the ML reinforcement learning model based on the penalized reward”; “generating the third training data for the ML reinforcement learning model from data stored in the order database, the third training data comprising a plurality of training instances, each training instance comprising (i) input feature data comprising a set of scores for a list of candidate medical items and (ii) label data comprising the selected candidate medical item”; and “training the ML reinforcement learning model based on the third training data”, amount to limiting the abstract idea to the field of machine learning models/algorithms (see MPEP § 2106.05(h) and analysis of these limitations under Step 2A, Prong Two above).
Therefore, these limitations are also deemed to be well-understood, routine, and conventional under Step 2B for similar reasons since they are claimed in a generic manner.
- Regarding the steps and features directed to: the steps directed to: “receiving summary note data that is composed by a physician for a patient”; “causing a subset of the plurality of candidate medical items to be presented on a screen of a computing device based on the ranking”; “receiving, from a user, input that indicates that one or more candidate medical items in the subset are incorrect recommendations”; “receiving, from a user, input that indicates one or more reasons why one or more candidate medical items in the subset are incorrect recommendations”; “receiving, from a user, input that indicates one or more reasons why one or more candidate medical items in the subset are incorrect recommendations”; “causing a subset of the second plurality of candidate medical items to be presented on the screen of the computing device based on the particular ranking”; “receiving an intervention corresponding to a ranking of the subset of the plurality of candidate medical items”; “receiving, from the physician, a selection of a candidate medical item from the subset of the plurality of candidate medical items”; “storing, in an order database, data identifying the physician, the patient, and the selected candidate medical item” - The following represents examples that courts have identified to be well-understood, routine, and conventional activities (e.g., see MPEP § 2106.05(d)):
- Receiving or transmitting data over a network, e.g., see Intellectual Ventures v. Symantec – similarly the limitations directed to: “receiving summary note data that is composed by a physician for a patient”; “where the set of feature values are stored as structured data”; “causing a subset of the plurality of candidate medical items to be presented on a screen of a computing device based on the ranking”; “receiving, from a user, input that indicates that one or more candidate medical items in the subset are incorrect recommendations”; “receiving, from a user, input that indicates one or more reasons why one or more candidate medical items in the subset are incorrect recommendations”; “receiving, from a user, input that indicates one or more reasons why one or more candidate medical items in the subset are incorrect recommendations”; “causing a subset of the second plurality of candidate medical items to be presented on the screen of the computing device based on the particular ranking”; “receiving an intervention corresponding to a ranking of the subset of the plurality of candidate medical items”; and “receiving, from the physician, a selection of a candidate medical item from the subset of the plurality of candidate medical items”, are similarly deemed to be well-understood, routine, and conventional activity in the medical field, because they also represent the mere collection and transmission of data over a network (i.e., merely receiving data from a user and presenting/displaying data on a screen are the equivalent of transmitting data over a network). See MPEP § 2106.05(d).
- Storing and retrieving information in memory, e.g., see Versata Dev. Group, Inc. v. SAP Am., Inc. – similarly the limitations directed to: “where the set of feature values are stored as structured data”; “storing prior medical items in an order database, each stored prior medical item including a timestamp”; and “storing, in an order database, data identifying the physician, the patient, and the selected candidate medical item”, are similarly deemed to be well-understood, routine, and conventional activity in the medical field, because they also represent the mere storage of data in memory/storage device (i.e., merely storing medical data in a database is the equivalent of storing data in a memory). See MPEP § 2106.05(d).
Therefore, the additional elements described in claims 1, 6-9, 11, 16-19, 23, 24, and 28-30 are deemed to be additional elements which do not amount to significantly more than the abstract idea identified above.
Thus, taken alone, the additional elements of claims 1, 6-9, 11, 16-19, 23, 24, 27, and 28-30 do not amount to significantly more than the above-identified judicial exception (the abstract idea). Furthermore, looking at the limitations as an ordered combination adds nothing that is not already present when looking at the elements taken individually. There is no indication that the combination of elements improves the functionality of a computer or improves any other technology, and their collective functions merely provide conventional computer implementation. Therefore, whether taken individually or as an ordered combination, claims 1, 6-9, 11, 16-19, 23, 24, and 28-30 are nonetheless rejected under 35 U.S.C. § 101 as being directed to non-statutory subject matter.
Additionally, dependent claims 4, 14, 21, 22, and 25 (which individually depend on claims 1 and 11 due to their respective chains of dependency), do not include additional elements that are sufficient to amount to significantly more than the judicial exception. Examiner notes that dependent claims 4, 14, 21, 22, and 25 do not include any additional elements beyond those identified as well-understood, routine, and conventional components as described above in the subject matter eligibility rejections of independent claims 1 and 11. Dependent claims 4, 14, 21, 22, and 25 merely add limitations that further narrow the abstract idea described in independent claims 1 and 11. Therefore, claims 1, 4, 6-9, 11, 14, 16-19, 21-25, and 28-30 are nonetheless rejected under 35 U.S.C. § 101 as being directed to non-statutory subject matter.

Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.

The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.

Claims 1, 4, 9, 11, 14, 19, 23, and 25 are rejected under 35 U.S.C. 103 as being unpatentable over:
- Amarasingham et al. (Pub. No. US 2023/0104655), in view of:
- Friedmann (Pub. No. US 2022/0068492).

Regarding claims 1 and 11,
- Amarasingham et al. (Pub. No. US 2023/0104655) teaches:
- a method comprising (as described in claim 1) (Amarasingham, paragraph [0074]; Paragraph [0074] teaches a method 300 for generating a summary.):
- one or more non-transitory storage media storing instructions which, when executed by one or more computing devices, cause (as described in claim 11) (Amarasingham, paragraph [0074]; Paragraph [0074] teaches that the method 300 may be implemented, at least in part, in the form of executable code stored on non-transitory, tangible, machine-readable media that, when run by one or more processors may cause the one or more processors to perform one or more of the processes 302-328.):
- receiving summary note data that is composed by a physician for a patient (as described in claims 1 and 11) (Amarasingham, paragraphs [0020], [0029], and [0044]; Paragraph [0029] teaches that the Application server 108 may receive and process various patient data (i.e., receiving summary note data that is for a patient), including structured and unstructured patient data, such as patient data stored in memory storage 104. For example, paragraph [0020] teaches that a raw summary may be generated from doctors’ notes (i.e., receiving summary note data that is composed by a physician for a patient). Further, paragraph [0044] teaches that an NLP model 202A may receive data input 201 that includes patient data, such as doctor’s notes (i.e., receiving summary note data that is composed by a physician for a patient) and identify the patient’s reason for admission to the hospital or another medical care facility.);
- generating, by a machine-learned (ML) language model, based on the summary note data, a set of feature values (Amarasingham, paragraphs [0054] and [0056]; Paragraph [0054] teaches that the AI engine 112 includes a knowledge graph (KG) module that includes a knowledge graph (KG) 214 that links clinical issues and their relationships, including links to other clinical issues, and/or treatments. For example, the KG 214 may categorize relationships among medical problems, symptoms, procedures, and/or treatments. Paragraph [0056] teaches that the KG 214 may be generated using a relational language model 218 (i.e., using a relational language model to generate a knowledge graph that includes links clinical issues and their relationships, including links to other clinical issues, medical problems, symptoms, procedures, and/or treatments, based on the received input patient data, is interpreted as generating a set of feature values, by a machine-learning (ML) language model, based on the summary note data).), where the set of features values comprise a plurality of values, each of the plurality of values corresponding to a respective feature of a plurality of features extracted from the summary note data (Amarasingham, paragraph [0088]; Paragraph [0088] teaches that the NLP models 202 may extract some or all concepts, including clinical issues, diagnoses, symptoms, treatments, etc., along with their attributes (e.g., section, negation, past, etc.) from the unstructured data (i.e., the set of values comprise a plurality of value, and each of the plurality of values correspond to a plurality of features extracted from the summary note data).), where the plurality of features includes one or more of a potential diagnosis, a risk factor, or potential treatment (Amarasingham, paragraphs [0020] and [0107]; Paragraph [0020] teaches that the summary generated for the patient may include current conditions and diagnoses that the patient may be experiencing (i.e., the plurality of features include one or more potential diagnoses) and current recommendations for medications and treatments (i.e., the plurality of features include one or more potential treatments). Paragraph [0107] teaches that the prioritization module 222 may prioritize risks (i.e., the plurality of features include one or more risk factors) into a ranked list.), and where the set of feature values are stored as structured data (Amarasingham, paragraph [0029]; Paragraph [0029] teaches that the application server 108 may receive and process the various patient data, including structured patient data stored in memory storage (i.e., the set of feature values are stored as structured data).), wherein the ML language model is trained on first training data using a first machine learning technique (Amarasingham, paragraph [0058]; Paragraph [0058] teaches the relational language model 218 may be trained by a deep learning language modeling approach based on a large set of clinical notes (i.e., this shows that all of the components of the AI engine, including the ML language model, are trained on first training data using a first machine learning technique) Paragraph [0038] teaches that each of the components and modules in AI engine 112 may be trained to perform a particular task or a combination of tasks (i.e., this shows that all of the components of the AI engine, including the ML language model, are trained). The training may occur using historical patient data or other types of datasets (i.e., the ML language model is trained on first training data), including datasets that may be accessible over network 106 of FIG. 1.) (as described in claims 1 and 11);
- identifying a profile of the patient (as described in claims 1 and 11) (Amarasingham, paragraph [0020]; Paragraph [0020] teaches that patient profile information may be identified and automatically processed to identify content of various inputs, such as a person, age, sex, reason for visiting the medical facility, current conditions, previous conditions, medications administered, etc.)…;
- determining, by an ML recommendation model, based on the profile of the patient, … , and the set of feature values, a plurality of candidate medical items (Amarasingham, paragraph [0030]; Paragraph [0030] teaches that the AI engine 112 executing within application server 108 receives data input 102 and generates data output 113 such as patient summary 114, where the summary 114 may include patient conditions 116, diagnostics and diagnoses 118 performed or recommended to determine additional conditions, and treatments 120 for known conditions (i.e., determining based on the profile of the patient, a plurality of candidate medical items) including both chronic and short-term conditions, where paragraph [0067] teaches that the AI engine includes NLG models that may be trained to make recommendations (i.e., the determinations of the plurality of candidate medical items is made with an ML recommendation model, where the NLG models are interpreted as the equivalent of the “ML recommendation model”).), where the set of feature values are received by the ML recommendation model as the structure data from the ML language model (as described in claims 1 and 11) (Amarasingham, paragraph [0067]; Paragraph [0067] teaches that the AI engine may include predictive models that receive data input 201 that includes the structured data, where the structured data includes the patient data and data generated using NLP models (i.e., the set of feature values are received by the ML recommendation model as structured data from the ML language model).), wherein the ML recommendation model is trained on second training data using a second machine learning technique (Amarasingham, paragraph [0067]; Paragraph [0067] teaches that an NLG model 206A may be trained on a summary task to summarize key health issues or conditions of the patient (i.e., the ML recommendation model is trained on second training data), where paragraph [0067] also teaches that the NLG models may be BERT models, GPT models, GPT-2 models, GPT-3 models, and the like (i.e., the ML recommendation model is trained using a second machine learning technique).) (as described in claims 1 and 11);
- generating, by an ML reinforcement learning model, a ranking of the plurality of candidate medical items based on the plurality of candidate medical items received from the ML recommendation model (as described in claims 1 and 11) (Amarasingham, paragraph [0054]; Paragraph [0054] generally teaches that the knowledge graph (KG) 214 may also rank which medical procedures and/or treatments are more important than others given a particular medical condition and/or symptoms, where paragraph [0065] teaches that the treatments, pending diagnostic orders, and active treatment orders are prioritized in a ranked list with a prioritization module (i.e., generating, by an ML reinforcement model, a ranking of the plurality of candidate medical items based on the plurality of candidate medical items received from the ML recommendation model, where the prioritization module is interpreted as the equivalent of an ML reinforcement learning model because one of ordinary skill in the arts would recognize that prioritization and optimization models are types of reinforcement learning models).), wherein the ML reinforcement learning model is trained on third training data using a third machine learning technique (Amarasingham, paragraphs [0038], [0039], and [0066]; Paragraph [0066] teaches that the prioritization module is trained to exclude or remove the identified treatment orders and diagnostic orders which are no longer required or have been fulfilled from the ranked list (i.e., the ML reinforcement model is trained on third training data), where paragraph [0038] generally teaches that each component of the AI engine 112 [which includes the prioritization module] may be trained to perform a particular task or combination of tasks, and the training may occur using historical patient data or other types of datasets, including datasets that may be accessible over network 106 of FIG. 1 (i.e., the ML reinforcement model is trained on third training data) and paragraph [0039] generally teaches that the AI engine may include various models, such as, recurrent neural network (RNN) models, neural net models including convolutional neural network (CNN) models, Bidirectional Encoder Representations from Transformers (BERT) models or their variants, and/or Generative Pre-trained Transformer (GPT) models, GPT-2 models, GPT-3 models and the like (i.e., the ML reinforcement model may use a third machine learning technique).) (as described in claims 1 and 11);
- causing a subset of the plurality of candidate medical items to be presented on a screen of a computing device based on the ranking (as described in claims 1 and 11) (Amarasingham, paragraphs [0029] and [0033]-[0035]; Paragraph [0029] teaches that the patient summaries 114 that include data that is ranked and prioritized (i.e., generating a subset of the plurality of candidate medical items that are ranked/prioritized) may be sent to computing devices of various interested entities, including, but not limited to, primary care physicians, surgeons, specialist physicians, pharmacists, close family members, extended family members, nurses, out-patient facilities, and other members of the medical facility, where paragraphs [0033]-[0035] explicitly teach that these summaries may be displayed on the devices of the physicians, surgeons, specialist physicians, pharmacists, close family members, extended family members, nurses, out-patient facilities, and other members of the medical facility (i.e., displaying the subset of plurality of candidate medical items on a screen of a computing device based on the ranking).); and
- wherein the method is performed by one or more computing devices (as described in claim 1) (Amarasingham, paragraph [0031]; Paragraph [0031] generally teaches that computing devices are used to receive the data, receive updates, provide access to the patient data, including the summaries; and display the summaries generated by the AI engine (i.e., the method is performed by one or more computing devices).).
- Amarasingham does not explicitly teach, however, in analogous art of medical systems and methods for predicting and detecting medical conditions, Friedmann (Pub. No. US 2022/0068492) teaches a method and non-transitory computer-readable medium comprising instructions that cause a computing device to perform the steps of:
- identifying a profile of a physician (as described in claims 1 and 11) (Friedmann, paragraph [0234]; Paragraph [0234]; Paragraph [0234] teaches that in step 1840, the processor may determine the physician profile based on the association.); and
- determining, based on the profile of the physician, a plurality of candidate medical items (as described in claims 1 and 11) (Friedmann, paragraphs [0079] and [0234]; Paragraph [0234] teaches that determining the physician profile may include determining, for a given medical condition and a patient’s profile a predicted decision (i.e., detecting a plurality of candidate medical items based on the profile of the physician). Paragraph [0079] teaches that this feature is beneficial for assisting physicians to accurately diagnose the patient and to provide a correct required treatment to the patient.).
Therefore, it would have been obvious to one of ordinary skill in the art of medical systems and methods for predicting and detecting medical conditions at the time of the effective filing date of the claimed invention to modify the method and non-transitory computer-readable medium for generating a clinical summary, taught by Amarasingham, to incorporate steps and features directed to: (i) identifying a physician profile, and (ii) determining candidate medical items based on the physician profile, as taught by Friedmann, in order to assist physicians to accurately diagnose the patient and to provide a correct required treatment to the patient. See Friedmann, paragraph [0079]; see also MPEP § 2143 G.

Regarding claims 4 and 14,
- The combination of: Amarasingham, as modified in view of Friedmann, teaches the limitations of: claim 1 (which claim 4 depends on) and claim 11 (which claim 14 depends on), as described above.
- Amarasingham teaches a method and one or more non-transitory storage media, further comprising:
- identifying, based on a patient identifier of the patient, one or more medical items, each of which was previously prescribed to the patient (as described in claims 4 and 14) (Amarasingham, paragraph [0024]; Paragraph [0024] teaches that the data input 102 may include medical input associated with the patient, including prescribed treatments (i.e., identifying one or more medical items, each of which was previously prescribed to the patient).); and
- wherein determining the plurality of candidate medical items by the ML recommendation model is also based on the one or more medical items (as described in claims 4 and 14) (Amarasingham, paragraph [0029]; Paragraph [0029] teaches that the application server 108 may receive and process various patient data [the input patient data] (i.e., this includes the identified one or more medical items and the previously prescribed treatments described in paragraph [0024] above), and the application server 108 includes an AI engine that correlates the patient data gathered from multiple sources with a knowledge base of medical treatment options, diagnostics, and other related information to generate a patient summary that includes the patient conditions 116, diagnoses, treatments, test/procedures, and additional conditions (i.e., the determined candidate medical items are based on the one or more medical items).).

Regarding claims 9 and 19,
- The combination of: Amarasingham, as modified in view of Friedmann, teaches the limitations of: claim 1 (which claim 9 depends on) and claim 11 (which claim 19 depends on), as described above.
- Amarasingham teaches a method and one or more non-transitory storage media, wherein:
- receiving the summary note data is performed by a prompt engine (as described in claims 9 and 19) (Amarasingham, paragraphs [0025] and [0029]; Paragraph [0029] teaches that an AI engine (i.e., a prompt engine) processes the patient data that includes structured and unstructured patient data (i.e., receiving patient data by a prompt engine), where paragraph [0025] teaches that the unstructured data may be written data, such as words written on doctor’s notes (i.e., the received patient data is patient summary note data).); and
- the method further comprising:
- generating a set of instructions (as described in claims 9 and 19) (Amarasingham, paragraph [0031]; Paragraph [0031] teaches that computing devices 122 may receive input from various parties, including a patient, family of the patient, doctors, nurses, medical students, residents, out-patient care facility, pharmacy, home-based sensors, etc., where in some instances, computing devices 122 may send a request message to AI engine to generate a summary 114 (i.e., generating a set of instructions).);
- causing the ML language model to access the set of instructions and the summary note data (as described in claims 9 and 19) (Amarasingham, paragraph [0074]; Paragraph [0074] teaches that the AI engine 112 may then query memory storage 104 at predefined intervals, upon request (i.e., accessing the set of instructions), or when memory storage 104 is updated with the patient data (i.e., accessing the summary note date), and generate a summary of the patient conditions.); and
- generating the set of feature values comprises generating the set of feature values also based on the set of instructions (as described in claims 9 and 19) (Amarasingham, paragraph [0032]; Paragraph [0032] teaches that the AI engine 112 may generate summary 114A that includes patient summary in layman’s terms and that is directed to one or more patients or their family members, or the medical summary 114B directed to medical professionals, such as during a handover from a physician-to-physician, doctor-to-nurse, etc., may include patient information in medical terms, such as key conditions 116B (e.g., tachycardia, pneumonia, sepsis), various key diagnostics 118B (e.g., sputum sample), and key treatments 120B (e.g., antibiotics Azithromycin, Amoxicillin) (i.e., generating the set of feature values based on the set of instructions to request the generation of the summary).).

Regarding claim 23,
- The combination of: Amarasingham, as modified in view of Friedmann, teaches the limitations of claim 1 (which claim 23 depends on), as described above.
- Amarasingham further teaches a method and one or more non-transitory storage media, wherein:
- the set of feature values includes potential treatments (Amarasingham, paragraph [0020]; Paragraph [0020] teaches that the summary generated for the patient may include current recommendations for medications and treatments (i.e., the plurality of features include one or more potential treatments).; and
- the instructions, when executed by the one or more computing devices, further cause in a cold-start scenario, training the ML recommendation model using the
potential treatments as label data (Amarasingham, paragraph [0040]; Paragraph [0040] teaches that the NLP models 202 may be pre-trained (i.e., a cold-start scenario) on training data that includes historical data associated with numerous patients, their conditions, treatments (i.e., training the ML recommendation model using the potential treatments in a cold-start scenario) and corresponding known outcomes.).

Regarding claim 25,
- The combination of: Amarasingham, as modified in view of Friedmann, teaches the limitations of claim 1 (which claim 25 depends on), as described above.
- Amarasingham teaches a method, wherein:
- the plurality of candidate items have relevance scores (Amarasingham, paragraph [0105]; Paragraph [0105] teaches that the order processing model 220 may identify key active treatment orders from the orders for medication and may prioritize the treatment orders according to relevance (i.e., the plurality of candidate items have relevance scores) using KG module 212.); and
- wherein the ranking is generated based on using the relevance scores in a reward function of the ML reinforcement learning model (Amarasingham, paragraph [0105]; Paragraph [0105] teaches that the order processing model 220 may determine which treatment orders would treat key active clinical issues identified in operation 312 and rank the treatment orders according to the prioritized active clinical issues (i.e., the ranking is based on using the relevance scores). The output of the order processing model 220 may include a ranked list of key active treatment orders (i.e., the ranking is based is based on using the relevance scores in a reward function of the ML reinforcement learning model).).

Claims 6, 7, 16, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over:
- The combination of: Amarasingham et al. (Pub. No. US 2023/0104655), as modified in view of Friedmann (Pub. No. US 2022/0068492), as applied to claims 1 and 11 above, and further in view of:
- Atkins et al. (Pub. No. US 2024/0071579).

Regarding claims 6 and 16,
- The combination of: Amarasingham, as modified in view of Friedmann, teaches the limitations of: claim 1 (which claim 6 depends on) and claim 11 (which claim 16 depends on), as described above.
- The combination of: Amarasingham, as modified in view of Friedmann, does not explicitly teach, however, in analogous art of methods and systems for predicting medical treatments and procedures, Atkins et al. (Pub. No. US 2024/0071579) teaches a method and one or more non-transitory storage media, further comprising:
- receiving, from a user, input that indicates that one or more candidate medical items in the subset are incorrect recommendations (as described in claims 6 and 16) (Atkins, paragraph [0043]; Paragraph [0043] teaches that if a user indicates that a predicted procedure was not performed (i.e., receiving input indicating that the one or more candidate medical items in the subset are incorrect recommendations), then the machine learning algorithm might infer an incorrect prediction (i.e., the recommended one or more recommended candidate medical items in the subset are deemed to be incorrect).); and
- updating the ML reinforcement learning model based on the input (as described in claims 6 and 16) (Atkins, paragraph [0043]; Paragraph [0043] teaches that the user feedback data may be sent to the machine learning algorithm to train the machine learning algorithm (i.e., training the ML algorithm based on the user feedback is the equivalent of updating the ML reinforcement learning model based on the user’s input). Paragraph [0043] teaches that this feature is beneficial for further refining and improving the machine learning algorithm. For example, paragraph [0043] teaches that the machine learning algorithm may positively reinforce the algorithm upon correct predictions, and may analyze data and adjust analysis where predictions are inaccurate.).
Therefore, it would have been obvious to one of ordinary skill in the art of methods and systems for predicting medical treatments and procedures at the time of the effective filing date of the claimed invention to further modify the method and non-transitory computer-readable medium for generating a clinical summary, taught by Amarasingham, as modified in view of Friedmann, to incorporate steps and features directed to: (i) receiving input from a user that indicates that the one or more candidate medical items are incorrect, and (ii) updating the machine learning model based on the user’s input, as taught by Atkins, in order to further refine and improve the machine learning algorithm. For example, paragraph [0043] teaches that the machine learning algorithm may positively reinforce the algorithm upon correct predictions, and may analyze data and adjust analysis where predictions are inaccurate. See Atkins, paragraph [0043]; see also MPEP § 2143 G.

Regarding claims 7 and 17,
- The combination of: Amarasingham, as modified in view of Friedmann, teaches the limitations of: claim 1 (which claim 7 depends on) and claim 11 (which claim 17 depends on), as described above.
- The combination of: Amarasingham, as modified in view of Friedmann, does not explicitly teach, however, in analogous art of methods and systems for predicting medical treatments and procedures, Atkins et al. (Pub. No. US 2024/0071579) teaches a method and one or more non-transitory storage media, further comprising:
- receiving, from a user, input that indicates one or more reasons why one or more candidate medical items in the subset are incorrect recommendations (as described in claims 7 and 17) (Atkins, paragraph [0043]; Paragraph [0043] teaches that if a user indicates that a predicted procedure was not performed, then the machine learning algorithm might infer an incorrect prediction and may look into contextual data provided as to why the predicted medical procedure was not performed (i.e., input indicating one or more reasons why one or more candidate medical items in the subset are incorrect recommendations). If the user elected to skip the procedure even though it was recommended, the system may still infer a correct prediction. If no procedure was recommended by the doctor (i.e., receiving input indicating one or more reasons why the one or more candidate medical items in the subset are incorrect recommendations), then the system may take the feedback as an incorrect prediction. If a tangentially related medical procedure was performed instead, the system may infer a neutral result and seek to update weighting and rankings for potential subsequent medical procedure predictions.); and
- updating the ML language model based on the input (as described in claims 7 and 17) (Atkins, paragraph [0043]; Paragraph [0043] teaches that the user feedback data may be sent to the machine learning algorithm to train the machine learning algorithm (i.e., training the ML algorithm based on the user feedback is the equivalent of updating the ML language model based on the user’s input). Paragraph [0043] teaches that this feature is beneficial for further refining and improving the machine learning algorithm. For example, paragraph [0043] teaches that the machine learning algorithm may positively reinforce the algorithm upon correct predictions, and may analyze data and adjust analysis where predictions are inaccurate.).
Therefore, it would have been obvious to one of ordinary skill in the art of methods and systems for predicting medical treatments and procedures at the time of the effective filing date of the claimed invention to further modify the method and non-transitory computer-readable medium for generating a clinical summary, taught by Amarasingham, as modified in view of Friedmann, to incorporate steps and features directed to: (i) receiving input from a user that indicates one or more reasons why the one or more candidate medical items are incorrect, and (ii) updating the machine learning model based on the user’s input, as taught by Atkins, in order to further refine and improve the machine learning algorithm. For example, paragraph [0043] teaches that the machine learning algorithm may positively reinforce the algorithm upon correct predictions, and may analyze data and adjust analysis where predictions are inaccurate. See Atkins, paragraph [0043]; see also MPEP § 2143 G.

Claims 8 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over:
- The combination of: Amarasingham et al. (Pub. No. US 2023/0104655), as modified in view of Friedmann (Pub. No. US 2022/0068492), as applied to claims 1 and 11 above, and further in view of:
- Van Assel et al. (Pub. No. US 2021/0233658); and
- Heyman (Pub. No. US 2013/0030832).

Regarding claims 8 and 18,
- The combination of: Amarasingham, as modified in view of Friedmann, teaches the limitations of: claim 1 (which claim 8 depends on) and claim 11 (which claim 18 depends on), as described above.
- The combination of: Amarasingham, as modified in view of Friedmann, does not explicitly teach, however, in analogous art of methods and systems for facilitating accurate medical diagnosis, Van Assel et al. (Pub. No. US 2021/0233658) teaches a method and one or more non-transitory storage media, further comprising:
- receiving, from a user, input that indicates one or more reasons why one or more candidate medical items in the subset are incorrect recommendations (as described in claims 8 and 18) (Van Assel, paragraphs [0222] and [0224]; Paragraph [0222] teaches that in step S119, the system of FIG. 2 makes a diagnosis (for example using the method explained above). The system then determines which further questions should be asked of the patient 101. Paragraph [0224] teaches that once the user supplies further information, where for example, paragraph [0224] generally teaches that the user information could be that a headache is present where instead the previous information was that a headache was absent (i.e., receiving, from a user, input indicating one or more reasons why the one or more candidate medical items are incorrect recommendations), then this is passed back to the diagnosis engine 111 to update evidence to produce updated probabilities.);
- updating the summary note data based on the input to generate an updated summary note data (as described in claims 8 and 18) (Van Assel, paragraph [0224]; Paragraph [0224] teaches that the further information supplied by the user is used to produce updated probabilities of the medical diagnosis (i.e., generating updated summary note data based on the user input).);
- generating, by the ML language model, based on the updated summary note data, a second set of feature values (as described in claims 8 and 18) (Van Assel, paragraph [0224]; Paragraph [0224] teaches that the evidence vector is updated with the new information from the updated probabilities (i.e., updating the evidence with the new information is deemed to be the equivalent of generating second set of feature values).); and
- generating, by the ML recommendation model, based on the profile of the patient, the profile of the physician, and the second set of feature values, a second plurality of candidate medical items (as described in claims 8 and 18) (Van Assel, paragraphs [0167] and [0224]; Paragraph [0224] teaches that the after all of the iterations of user questions are performed, a final diagnosis is output (i.e., generating a second plurality of candidate medical items). Paragraph [0167] teaches that this features are beneficial for giving more context to a diagnosis system to allow for improvement of a diagnosis.).
…
…
Therefore, it would have been obvious to one of ordinary skill in the art of methods and systems for facilitating accurate medical diagnosis at the time of the effective filing date of the claimed invention to further modify the method and non-transitory computer-readable medium for generating a clinical summary, taught by Amarasingham, as modified in view of Friedmann, to incorporate steps and features directed to: (i) receiving input from a user that indicates one or more reasons why the candidate medical items are incorrect; (ii) updating the summary note data based on the input from the user to generate an updated summary note data; (iii) generating a second set of feature values based on the updated summary note data; and (iv) generating a second plurality of candidate medical items based on the set of feature values, patient profile, and physician profile, as taught by Van Assel, in order to give more context to a diagnosis system to allow for improvement of a diagnosis. See Van Assel, paragraph [0167]; see also MPEP § 2143 G.
- Further, the combination of: Amarasingham, as modified in view of: Friedmann and Van Assel, does not explicitly teach, however, in analogous art of methods and systems for facilitating accurate medical diagnosis, Heyman (Pub. No. US 2013/0030832) teaches a method and one or more non-transitory storage media, further comprising:
- generating, by the ML reinforcement learning model, a particular ranking of the second plurality of candidate medical items (as described in claims 8 and 18) (Heyman, paragraph [0014]; Paragraph [0014] teaches that input from each of the plurality of participants regarding the likelihood of each of the potential diagnoses being correct is averaged at the server, thereby generating a ranking of the potential diagnoses from highest likelihood of being correct to lowest likelihood of being correct (i.e., generating a ranking of the second plurality of candidate medical items).); and
- causing a subset of the second plurality of candidate medical items to be presented on the screen of the computing device based on the particular ranking (as described in claims 8 and 18) (Heyman, paragraphs [0010] and [0014]; Paragraph [0014] teaches that the ranking is transmitted from the server to the patient computer (i.e., transmitting the ranking to a patient computer is deemed to be the equivalent of causing a subset of the second plurality of candidate medical items to be presented on the screen of a computing device based on the particular ranking). Further, paragraph [0014] also teaches that an indication of which of the potential diagnoses was found to be a correct diagnosis by a physician who examined the patient is received at the server from the patient computer (i.e., this is also deemed to be the equivalent of causing a subset of the second plurality of candidate medical items to be presented on the screen of a computing device based on the particular ranking, because the server receives and displays only the potential diagnoses that were found to be correct by the physician who examined the patient, i.e., the correct diagnoses are the highest ranked diagnoses.) Paragraph [0010] teaches that these features are beneficial for enabling accurate predictions regarding the real world success of concepts being tested.).
Therefore, it would have been obvious to one of ordinary skill in the art of methods and systems for facilitating accurate medical diagnosis at the time of the effective filing date of the claimed invention to further modify the method and non-transitory computer-readable medium for generating a clinical summary, taught by Amarasingham, as modified in view of: Friedmann and Van Assel, to incorporate steps and features directed to: (i) generating a particular ranking of the second plurality of candidate medical items; and (ii) transmitting and presenting the particular ranking of the second plurality of candidate medical items on the screen of the computing device, as taught by Heyman, in order to enable accurate predictions regarding the real world success of concepts being tested. See Heyman, paragraph [0010]; see also MPEP § 2143 G.

Claims 21 and 22 are rejected under 35 U.S.C. 103 as being unpatentable over:
- The combination of: Amarasingham et al. (Pub. No. US 2023/0104655), as modified in view of Friedmann (Pub. No. US 2022/0068492), as applied to claim 1 above, and further in view of:
- Devarakonda et al. (Pub. No. US 2015/0356270).

Regarding claim 21,
- The combination of: Amarasingham, as modified in view of Friedmann, teaches the limitations of claim 1 (which claim 21 depends on), as described above.
- The combination of: Amarasingham, as modified in view of Friedmann, does not explicitly teach, however, in analogous art of methods and systems for automatically generating predictive outputs for medical events, Devarakonda et al. (Pub. No. US 2015/0356270) teaches a method and one or more non-transitory storage media, wherein:
- generating the ranking comprises increasing a score of a candidate medical item that became available within a freshness window by a predetermined percentage (Devarakonda, paragraphs [0029] and [0030]; Paragraph [0029] teaches that machine learning process may increase the weighting of feature values based on the locations of the medical problem terms within the structure of the electronic medical records (i.e., increasing a score of a candidate medical item that became available within a freshness window), where paragraph [0030] teaches that if the final combined feature score for candidate medical problem classified by a standardized medical concept is below a threshold, that candidate medical problem will be removed from the group of probable standardized medical concepts in item 410 (i.e., a predetermined percentage for the candidate medical item determines if it becomes available within the freshness window, where the threshold is interpreted as being the equivalent of “a predetermined percentage”). Paragraph [0030] teaches that this feature is beneficial for increasing accuracy and making the final results output to the user more acceptable.).
Therefore, it would have been obvious to one of ordinary skill in the art of methods and systems for automatically generating predictive outputs for medical events at the time of the effective filing date of the claimed invention to further modify the method and non-transitory computer-readable medium for generating a clinical summary, taught by Amarasingham, as modified in view of Friedmann, to incorporate a step and feature directed to increasing a score of a candidate medical item that became available based on a predetermined percentage, as taught by Devarakonda, in order to increase accuracy and make the final results output to the user more acceptable. See Devarakonda, paragraph [0030]; see also MPEP § 2143 G.

Regarding claim 22,
- The combination of: Amarasingham, as modified in view of: Friedmann and Devarakonda, teaches the limitations of claim 21 (which claim 22 depends on), as described above.
- Devarakonda further teaches a method, wherein:
- the predetermined percentage decreases over time until reaching zero after expiration of the freshness window (Devarakonda, paragraph [0030]; Paragraph [0030] teaches that the machine learning process may decrease the weighting of features scores (i.e., the predetermined percentage decreases over time) based on how recently the medical problem terms appear within the electronic medical records in response to feedback, where paragraph [0018] teaches that if a score is less than a threshold, the final feature score for the feature is zero (i.e., the predetermined percentage decrease over time until reaching zero).).
The motivations and rationales for modifying the method and non-transitory computer readable storage medium for generating a clinical summary taught by Amarasingham, in view of: Friedmann and Devarakonda, described in the obviousness rejections of claims 1, 11, 21, and 26 above similarly apply to this obviousness rejection, and are incorporated herein by reference.

Claims 24 and 28 are rejected under 35 U.S.C. 103 as being unpatentable over:
- The combination of: Amarasingham et al. (Pub. No. US 2023/0104655), as modified in view of Friedmann (Pub. No. US 2022/0068492), as applied to claims 1 and 11 above, and further in view of:
- Grantcharov et al. (Pub. No. US 2021/0076966).

Regarding claims 24 and 28,
- The combination of: Amarasingham, as modified in view of Friedmann, teaches the limitations of: claim 1 (which claim 24 depends on) and claim 11 (which claim 28 depends on), as described above.
- The combination of: Amarasingham, as modified in view of Friedmann, does not explicitly teach, however, in analogous art of methods and systems for automatically generating predictive outputs for medical events, Grantcharov et al. (Pub. No. US 2021/0076966) teaches a method and one or more non-transitory storage media, further comprising:
- storing prior medical items in an order database, each stored prior medical item including a timestamp (as described in claims 24 and 28) (Grantcharov, paragraph [0518]; Paragraph [0518] the time-stamped clinical events within the session container file is stored with associated metadata for duration and frequency of each time-stamped clinical event (i.e., storing prior medical items in an order database, each stored prior medical item including a timestamp).);
- identifying a temporal order of the medical items (as described in claims 24 and 28) (Grantcharov, paragraphs [0409] and [0505]; Paragraph [0505] teaches that the interface indicator identifies each of the time-stamped clinical events and paragraph [0409] teaches that the method provides a temporal order among a set of events (i.e., identifying a temporal order of the medical items) and the associated confidence levels.);
- providing the temporal order as an input to the ML recommendation model (as described in claims 24 and 28); and wherein the plurality of candidate medical items is determined based on the temporal order (as described in claims 24 and 28) (Grantcharov, paragraphs [0409] and [0539]; Paragraph [0539] teaches that a central content server generates an interface indicator for a visual sequence of the time-stamped clinical events within the session container file and correspondence assessments; and generates a predictive data model for refining protocol generation using support vector machines or artificial intelligence network data structures (i.e., providing the temporal order as input to the ML recommendation model, where the plurality of candidate medical items is determined based on the temporal order). Paragraph [0409] teaches that this feature is beneficial for extracting data in a meaningful and structured way and facilitating data analysis and the development of predictive algorithms.).
Therefore, it would have been obvious to one of ordinary skill in the art of methods and systems for automatically generating predictive outputs for medical events at the time of the effective filing date of the claimed invention to further modify the method and non-transitory computer-readable medium for generating a clinical summary, taught by Amarasingham, as modified in view of Friedmann, to incorporate steps and features directed to: (i) storing medical items in an order database with associated time-stamp data for each medical item; (ii) identifying a temporal order of the medical items; (iii) providing the temporal order as an input to the ML recommendation model; and (iv) wherein the plurality of candidate medical items is determined based on the temporal order, as taught by Grantcharov, in order to extract data in a meaningful and structured way, which helps to facilitate data analysis and the development of predictive algorithms. See Grantcharov, paragraph [0409]; see also MPEP § 2143 G.

Claim 29 is rejected under 35 U.S.C. 103 as being unpatentable over:
- The combination of: Amarasingham et al. (Pub. No. US 2023/0104655), as modified in view of Friedmann (Pub. No. US 2022/0068492), as applied to claim 25 above, and further in view of:
- Hussain et al. (Pub. No. US 2023/0170093).

Regarding claim 29,
- The combination of: Amarasingham, as modified in view of Friedmann, teaches the limitations of claim 25 (which claim 29 depends on), as described above.
- The combination of: Amarasingham, as modified in view of Friedmann, does not explicitly teach, however, in analogous art of medical predictive methods and systems for recommending clinical interventions, Hussain et al. (Pub. No. US 2023/0170093) teaches a method, further comprising:
- receiving an intervention corresponding to a ranking of the subset of the plurality of candidate medical items (Hussain, paragraph [0086]; Paragraph [0086] teaches that clinicians may be able to choose from the list of recommended on-guideline care actions and intermediate to long term care pathway plans (i.e., receiving an intervention corresponding to a ranking of the subset of the plurality of candidate medical items).);
- penalizing, based on the intervention, a reward corresponding to the reward function for the ranking generated by the ML reinforcement learning model (Hussain, paragraphs [0066] and [0075]; Paragraph [0075] teaches that the unexplored state action handler 407 is able to reward of penalize performance of a clinical action (i.e., penalizing a reward corresponding to the reward function for the ranking generated by the ML reinforcement learning model) at a defined clinical state st based at least in part on occurrence rate of historical experience records, where paragraph [0066] teaches that the various historical experience records each describe that a recorded performance of a pruned clinical action at a first defined clinical state has led to a second defined clinical state which may or may not be different than the first defined clinical state (i.e., the penalization of the reward is based on the intervention, because the historical experience records describe the clinical actions including the clinical actions which were chosen by the clinical actions).); and
- updating the ML reinforcement learning model based on the penalized reward (Hussain, paragraph [0070]; Paragraph [0070] teaches that the trainable parameters of the dynamics of the machine learning model 405 are updated (i.e., updating the ML reinforcement learning model) based on the historical experience record . Paragraph [0070] teaches that these features are beneficial for minimizing the error function of the machine learning model.).
Therefore, it would have been obvious to one of ordinary skill in the art of medical predictive methods and systems for recommending clinical interventions at the time of the effective filing date of the claimed invention to further modify the method and non-transitory computer-readable medium for generating a clinical summary, taught by Amarasingham, as modified in view of Friedmann, to incorporate steps and features directed to: (i) receiving an intervention corresponding to a ranking of the subset of the plurality of candidate medical items; (ii) penalizing a reward that corresponds to the reward function for the ranking of candidate medical items; and (iii) updating the machine learning reinforcement learning model based on the penalized reward, as taught by Hussain, in order to minimize the error function of the machine learning model. See Hussain, paragraph [0070]; see also MPEP § 2143 G.

Claim 30 is rejected under 35 U.S.C. 103 as being unpatentable over:
- The combination of: Amarasingham et al. (Pub. No. US 2023/0104655), as modified in view of Friedmann (Pub. No. US 2022/0068492), as applied to claim 1 above, and further in view of:
- Getz et al. (Pub. No. US 2023/0099880).

Regarding claim 30,
- The combination of: Amarasingham, as modified in view of Friedmann, teaches the limitations of claim 1 (which claim 30 depends on), as described above.
- Friedmann teaches a method, further comprising:
- receiving, from the physician, a selection of a candidate medical item from the subset of the plurality of candidate medical items (Friedmann, paragraph [0080]; Paragraph [0080] teaches that the system 100 may receive one or more data elements from one or more EMRs, where paragraph [0081] teaches that the EMR data may include previous data and/or analysis (e.g., made by a physician, a healthcare professional, and the like) of at least one of a physical, medical, intellectual, social and/or mental condition of the patient, and a medical history of diagnosed physical, cognitive and/or mental diseases that the patient may have been diagnosed with (i.e., receiving, from the physician, a selection of a diagnosed disease, where the diagnosed disease is interpreted as the equivalent of a “candidate medical item”, from the subset of the plurality of candidate medical items).);
…
- generating the third training data for the ML reinforcement learning model from data stored in the order database (Friedmann, paragraph [0200]; Paragraph [0200] teaches that the method may further include training a machine learning (ML) module to predict the medical condition based on the completed data and the level of importance.), the third training data comprising a plurality of training instances, each training instance comprising (i) input feature data comprising a set of scores for a list of candidate medical items (Friedmann, paragraphs [0235]; Paragraph [0235] teaches that, in relation to the training of the ML module, the processor may (i) assess whether the diagnosis was correct by, for example, comparing the diagnosis of a specific physician to diagnosis given to similar women (e.g., as set forth by use similarity module 411) by one or more other physicians; and (ii) give a reliability score to each doctor and may determine how much weight to give the decisions and diagnoses of each specific physician (i.e., each training instance comprises a set of scores for a list of candidate medical items).) and (ii) label data comprising the selected candidate medical item (Friedmann, paragraphs [0097] and [0098]; Paragraph [0097] teaches that the system may collected labeled data from the hospital where paragraph [0098] teaches that the parameterizing module 340 may include a library of signals received from hospitals and labeled accordingly, and then label the data that was collected from wearable devices, such as analyzing the data to detect events or patterns of a heart attack (i.e., labeling the data comprises the selected candidate medical item).); and
- train the ML reinforcement learning model based on the third training data (Friedmann, paragraphs [0070], [0146], and [0207]; Paragraph [0146] teaches that the ML module 50 may include any machine learning model known in the art, where paragraph [0070] teaches that the machine learning model may include reinforcement learning (i.e., the machine learning module is an ML reinforcement learning model), and the required parameters and/or completed data may be provided to the ML module 50, so as to train ML module 50 to predict a medical condition or precondition (i.e., training the ML reinforcement learning model based on the third training data). Paragraph [0207] teaches that these features are beneficial for reinforcement between prediction and detection, in order to reduce misdiagnosis.).
Therefore, it would have been obvious to one of ordinary skill in the art of medical systems and methods for predicting and detecting medical conditions at the time of the effective filing date of the claimed invention to further modify the method for generating a clinical summary, taught by Amarasingham, as modified in view of Friedmann, to incorporate steps and features directed to: (i) receiving selections of candidate medical items from a physician, (ii) generating training data for the ML reinforcement learning model from data identifying the physician, patient, and selected candidate medical items, and (iii) training the ML reinforcement model based on the training data, also taught by Friedmann, in order to assist physicians to improve the reinforcement between prediction and detection, in order to reduce misdiagnosis. See Friedmann, paragraph [0207]; see also MPEP § 2143 G.
- The combination of: Amarasingham, as modified in view of Friedmann, does not explicitly teach, however, in analogous art of medical predictive methods and systems for diagnosing diseases, Getz et al. (Pub. No. US 2023/0099880) teaches a method, further comprising:
- storing, in an order database, data identifying the physician, the patient, and the selected candidate medical item (Getz, paragraphs [0070], [0071], [0115], and [0145]; Paragraph [0071] teaches that in step 110, the doctor or other health care provider, generates retrospective feedback on the diagnostic accuracy of the output generated by the system. The physician's analysis of the system's performance is input to the expert medical logic database of step 102 (i.e., storing, in an order database, data identifying the physician). Paragraph [0115] generally teaches that the at least one database comprises output of the system, such as a diagnosis or list of possible diagnoses, ranked in order of likelihood (i.e., storing, in an order database, data identifying the selected candidate medical item). Paragraph [0145] generally teaches that the large population database contains health related data of the large population (i.e., storing, in an order database, data identifying the patient). Paragraphs [0070] and [0071] teaches that this feature is beneficial for updating and improving the accuracy of the output indicating the likelihood that the given individual may develop an autoimmune disease in the future.).
Therefore, it would have been obvious to one of ordinary skill in the art of medical predictive methods and systems for diagnosing diseases at the time of the effective filing date of the claimed invention to further modify the method for generating a clinical summary, taught by Amarasingham, as modified in view of Friedmann, to incorporate a step and feature directed to storing data that identifies a physician, a patient, and selected candidate medical items, as taught by Getz, in order to update and improve the accuracy of the output indicating the likelihood that the given individual may develop an autoimmune disease in the future. See Getz, paragraphs [0070] and [0071]; see also MPEP § 2143 G.

Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Nicholas Akogyeram II whose telephone number is [phone]. The examiner can normally be reached Monday - Friday, between 8:00am - 5:00pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Jason Dunham can be reached at [phone]. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
Official replies to this Office action may now be submitted electronically by registered
users of the EFS-Web system. Information on EFS-Web tools is available on the Internet at:
http://www.uspto.gov/patents/processlfi!elefslguidance/index.isp. An EFS-Web Quick-Start
Guide is available at: http://www.uspto.gov/ebc/portallefslquick-start.pdf.
Alternatively, official replies to this Office Action may still be submitted by any one of fax, mail, or hand delivery.
Faxed replies should be directed to the central fax at [phone].
Mailed replies should be addressed to:
United States Patent and Trademark Office:
Commissioner of Patents and Trademarks
P.O. Box 1450
Alexandria, VA 22313-1450
Hand delivered responses should be brought to the United States Patent and Trademark Office Customer Service Window:
Randolph Building
401 Dulany Street
Alexandria, VA 22314-1450

/N.A.A./Examiner, Art Unit 3686

/JONATHON A. SZUMNY/Primary Examiner, Art Unit 3686

Prosecution Timeline

• Jul 14, 2023: Application Filed
• Nov 21, 2025: Non-Final Rejection mailed — §101, §103
• Feb 10, 2026: Examiner Interview Summary
• Feb 10, 2026: Applicant Interview (Telephonic)
• Feb 12, 2026: Response Filed
• Jun 16, 2026: Final Rejection mailed — §101, §103
• Jul 16, 2026: Examiner Interview Summary
• Jul 16, 2026: Applicant Interview (Telephonic)
• Jul 17, 2026: Response after Non-Final Action
• Sep 10, 2026: Request for Continued Examination
• Sep 16, 2026: Response after Non-Final Action
• Oct 05, 2026: Non-Final Rejection mailed — §101, §103 (current)