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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 6/9/2026 has been entered.
Status of the Claims
The status of the claims as of the response filed 6/9/2026 is as follows: Claims 8-9 and 18 are cancelled, and all previously given rejections for these claims are considered moot. Claims 1, 15, and 20 are currently amended. Claims 2-7, 10-14, 16-17, and 19 are original. Claims 1-7, 10-17, and 19-20 are currently pending in the application and have been considered below.
Response to Amendment
Rejection Under 35 USC 101
The claims have been amended but the 35 USC 101 rejections are upheld.
Rejection Under 35 USC 103
The amendments made to the claims introduce limitations that are not fully addressed in the previous office action, and thus the corresponding 35 USC 103 rejections are withdrawn. However, Examiner will consider the amended claims in light of an updated prior art search and address their patentability with respect to prior art below.
Response to Arguments
Rejection Under 35 USC 101
On page 7 of the response filed 6/9/2026 Applicant argues that the “combination of 1) a sensor in conformal contact with a user, 2) a hardware processor that trains and predicts the onset of physiological events with a time to event loss function, and 3) a hardware processing that modifies a monitored window of physiological signals based on a detected occurrence of a physiological event is more than an abstract idea.” Applicant analogizes such features to the set of claims found eligible in SRI, which were directed to detecting suspicious activity using network monitoring of network packets. Applicant’s arguments are fully considered, but are not persuasive. Examiner notes that the fact patterns of the instant case and SRI are different, and that each case is evaluated for eligibility on its own merits. In the instant case, clinicians are frequently tasked with analyzing data from physiological sensors to identify or detect significant physiological events, whereas it would not be practical for a human actor to monitor network traffic data such as network packet data transfer commands, network packet data transfer errors, network packet data volume, network connection requests, network connection denials, error codes included in a network packet, network connection acknowledgements, and network packets indicative of well-known network-service protocols to detect suspicious network activity as in SRI. Examiner further notes that training a machine learning model with a time to event loss function is a mathematical concept, such that this limitation recites an abstract idea in a manner similar to the neural network training step of Example 47. Examiner submits that modifying a monitored window of physiological signals based on a detected occurrence of a physiological event is also abstract, because it describes how a person might widen or narrow the window of data used in future event detection operations (e.g. making a note to look at a full month of data rather than just a day’s worth of data if the user does have worrying signs of a physiological event). The use of a sensor and hardware processor to provide data for and perform these operations does not preclude the claims from reciting an abstract idea, and these features are instead evaluated as additional elements in Steps 2A – Prong 2 and 2B.
On pages 7-8 Applicant outlines a list of claim features believed to provide integration into a practical application by “improv[ing] the technological process.” Applicant further analogizes the instant claims to those found eligible in Example 47 (claim 3), Ex parte Carmody, and Ex parte Desjardins and reiterates that “the features of claim 1 constitute an improvement to a wearable monitoring device that predicts the onset of physiological events.” Applicant’s arguments are fully considered, but are not persuasive. Examiner first notes that many of the features that Applicant points to as providing the practical application are part of the abstract idea itself rather than additional elements; for example:
accessing training data for a plurality of users is an abstract data obtaining operation that a human actor could achieve by collecting or retrieving a dataset for training a model;
accessing physiological status indicators for each user wherein the status indicator is determined based on the training data is an abstract data obtaining and/or determination operation that a human actor could achieve by looking at the training dataset to identify associated status indicators for each user and/or perform calculations on the data to determine the indicators;
training a machine learning model using the training data and the physiological status indicator is an abstract mathematical concept because it comprises applying a failure time and a censor variable to a time to event loss function that predicts probabilities of survival/failure within a specified timeframe, which describes applying variables to a mathematical function to train the model;
modifying a window of monitored physiological signals to detect an occurrence of the physiological event based on the prediction of the onset of the physiological event is an abstract data manipulation operation that a human actor could achieve by adjusting the timespan of data that they look at and evaluate to perform event detection for a user, e.g. by looking at an entire month’s worth of data rather than examining only one day’s worth of data when a patient has experienced a cardiac arrhythmia in the past;
modifying a selection of physiological signals to transmit to an external monitoring device is an abstract data selection operation that a human actor could achieve by thinking about and picking out a specific dataset (e.g. of particular physiological signal types, designated timeframes, etc.) that they would like to be sent to an external system.
Because these features are part of the abstract idea itself, they do not provide “significantly more” than the abstract idea and thus do not confer eligibility (see MPEP 2106.05(a): “It is important to note, the judicial exception alone cannot provide the improvement. The improvement can be provided by one or more additional elements.” See also 2106.05(a)(II): “it is important to keep in mind that an improvement in the abstract idea itself… is not an improvement in technology”). As indicated above, the additional elements of a sensor to provide the data and a hardware processor to perform the event detection, training, adjustment, and other steps of the invention are evaluated under Steps 2A – Prong 2 and 2B. In the instant case, the adhesive assembly and sensor additional elements amount to insignificant extra-solution activity in the form of data gathering because they merely serve as means of obtaining the physiological signals needed for the main data analysis steps of the invention (see MPEP 2106.05(g)). The use of a hardware processor to perform the various steps of the invention amounts to instructions to “apply” the abstract idea using generic computer components because this element merely serves as a tool with which the otherwise-abstract functions of applying signals to a model, accessing various types of data, training a model, and modifying windows and data selections are digitized and/or automated (see MPEP 2106.05(f)).
Regarding Applicant’s arguments with respect to claim 3 of Example 47, Examiner notes that the instant claims are not analogous to that eligible claim because they do not provide an analogous non-abstract, technical remedial action that provides integration into a practical application. The claim in question included steps (e) and (f), which automatically drop the detected malicious network packet(s) and block future traffic from the associated source address, which provided specific computer solutions that utilized the output from the ANN to provide security solutions to the detected anomalies. In contrast, the resulting actions from the detected physiological events in the instant claims include modifying a window of physiological signals monitored to detect occurrences of the physiological event and modifying a selection of physiological signals transmitted to an external monitoring device, each of which describe abstract determinations and data manipulations (as explained above) rather than specific computer/ technical solutions to an identified technical problem.
Regarding Applicant’s arguments with respect to Carmody and Desjardins, Examiner notes that these cases both described specific technical improvements to methods of training machine learning models that resulted in technical advantages described in their respective specifications. For example, in Desjardins, the claims reflected an improvement to how a machine learning model itself is trained and operates to address the technical problem of ‘catastrophic forgetting’ encountered in continual learning systems, which was identified and explained as a technical problem in the specification. In contrast, the instant specification does not outline a specific technical problem in machine learning technology whose solution is reflected in the claims. Examiner respectfully disagrees that the instant claims provide improvements to the technical field of machine learning or methods of training machine learning models themselves, and instead appear to apply known machine learning techniques to the clinical field of cardiac arrhythmia prediction. This high-level use of computerized machine learning techniques to digitize and/or automate the prediction of clinical events from physiological data does not provide a technical improvement to a technical problem (as in Desjardins and Carmody), and instead amount to instructions to “apply” the exception with a computer.
For the reasons outlined above, the 35 USC 101 rejections are upheld for claims 1-7, 10-17, and 19-20.
Rejection Under 35 USC 103
On page 9 Applicant argues that “Gensheimer does not disclose both training a machine learning model using a physiological status indicator and applying a failure time and a censor variable to a time to event loss function. At most, only one of the claimed features is disclosed.” Applicant’s arguments are fully considered, but are not persuasive. In Gensheimer, a clinical prediction machine learning model such as a neural network is trained as explained in the “Implementation” section on Pg 5. The training includes a custom loss function to teach the model to output a vector representing predicted probability of surviving a given time interval, i.e. a time to event loss function. The training data that is used includes a plurality of individuals who each have a known failure/censoring time t and a censoring indicator (also considered equivalent to physiological status indicators determined based on the training data because they represent outcomes of the training data), such that training with the loss function as explained above is considered equivalent to training the model using a physiological status indicator and applying a failure time and a censor variable to a time to event loss function as claimed. Accordingly, Examiner maintains that Gensheimer sufficiently suggests this feature of the independent claims.
On pages 9-10 Applicant argues that neither Thakur nor Gensheimer discloses the newly-introduced claim language directed to modifying a window of physiological signals and modifying a selection of physiological signals transmitted to an external monitoring device. Applicant’s arguments are fully considered, but are not persuasive. Examiner submits that Thakur does teach the limitations at issue, as follows:
wherein the hardware processor is further configured to modify a window of physiological signals monitored to detect an occurrence of the physiological event based on the prediction of the onset of the physiological event (Thakur [0086], noting the system can tune or adjust aspects of the sensors such as sensor data acquisition time, schedule, frequency, duration (i.e. window), sampling rate, etc. based on a current estimate of arrhythmia risk for a patient), and
wherein modifying the window of physiological signals modifies a selection of physiological signals transmitted to an external monitoring device that implements a second machine learning model configured to classify the physiological event (Thakur [0053]-[0055], noting an external system 125 may receive device data from the AMD sensor including real-time or stored physiologic data from the patient and perform further classification processing on the data, e.g. to verify that a detected medical event is a true positive. Taken together with the tunable/adjustable data acquisition timing, scheduling, frequency, duration, sampling rate, etc. of [0086], these disclosures show that the system may modify the selection of physiological signals that are actively acquired by the system (e.g. by selecting the amount of data acquired over a modified time window) and thus transmitted to the external system for further processing as in [0053]-[0055], which meets the broadest reasonable interpretation of this limitation as outlined in the claim interpretation section below).
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 6/9/2026 is in accordance with the provisions of 37 CFR 1.97 and is considered by the Examiner. Examiner notes that foreign reference CN 102917638 has not been considered because no English translation or explanation of relevance for this reference was provided.
Claim Interpretation
The implementation of the second machine learning model by an external monitoring device is not positively recited in independent claims 1, 15, or 20 and instead reflects an intended/desired result or use of the transmitted signals; the only positively recited function of the final limitation is the hardware processor / method / non-transitory computer storage medium modifying a selection of physiological signals transmitted to an external monitoring device. The external monitoring device is not a positively recited part of the claimed electronic device / method / non-transitory computer storage medium, and there is no indication that the external monitoring device is required to implement the second machine learning model to classify the physiological event as part of the structure and functioning of the claimed electronic device / method non-transitory computer storage medium. That is, the positively recited structure and functioning of each respective claimed invention would remain the same if the external device received the selected signals and implemented a second machine learning model as desired, received the selected signals and performed some other analysis or processing, received the selected signals and did no further processing, failed to receive the selected signals at all, etc. Accordingly, the broadest reasonable interpretation of this limitation includes modifying a selection of physiological signals transmitted to an external monitoring device for any purpose. See MPEP 2111.
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 1-7, 10-17, and 19-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1
In the instant case, claims 1-7 and 10-14 are directed to a device (i.e. a machine), claims 15-17 and 19 are directed to a method (i.e. a process), and claim 20 is directed to a non-transitory computer storage medium (i.e. a manufacture). Thus, each of the claims falls within one of the four statutory categories. Nevertheless, the claims fall within the judicial exception of an abstract idea.
Step 2A – Prong 1
Independent claims 1, 15, and 20 recite steps that, under their broadest reasonable interpretations, cover certain methods of organizing human activity, e.g. managing personal behavior, relationships, or interactions between people, as well as mathematical concepts. Specifically, claim 1 (as representative) recites:
An electronic device for monitoring physiological signals of a user, the electronic device comprising:
an adhesive assembly comprising a housing that encloses a circuit board;
a sensor in electrical communication with the circuit board and configured to be positioned in conformal contact with the surface of the user to detect the physiological signals of the user; and
a hardware processor configured to apply the physiological signals to a machine learning model, wherein the machine learning model is configured to predict an onset of a physiological event based on the physiological signals of the user, and wherein the machine learning model is trained by:
accessing training data for a plurality of users separate from the user, wherein the training data comprises physiological data of the plurality of users;
for each user of the plurality of users, accessing a physiological status indicator, wherein the physiological status indicator is determined based on the training data; and
training the machine learning model based on the training data and the physiological status indicator to predict the onset of the physiological event, wherein training the machine learning model comprises applying a failure time and a censor variable to a time to event loss function that predicts probabilities of survival/failure within a specified timeframe, wherein the hardware processor is further configured to modify a window of physiological signals monitored to detect an occurrence of the physiological event based on the prediction of the onset of the physiological event, and wherein modifying the window of physiological signals modifies a selection of physiological signals transmitted to an external monitoring device that implements a second machine learning model configured to classify the physiological event.
But for the recitation of generic computer components like a hardware processor and high-level machine learning, the italicized functions, when considered as a whole, describe a clinical event detection and model fitting operation that could be achieved via mathematical operations and by a human actor such as a clinician or other medical professional managing their personal behavior and/or interactions with others. For example, a clinician could look at sensor readouts of physiological signals for a patient and apply the signals to a predictive model fitted with labelled data from other patients to predict an onset of a physiological event. The application of a failure time and a censor variable to a time to event loss function that predicts probabilities of survival/failure within a specified timeframe describes mathematical processes for training the model. A human actor would also be capable of modifying a monitoring window based on occurrence of a detected event (e.g. by looking at an entire month’s worth of data rather than examining only one day’s worth of data when a patient has experienced a cardiac arrhythmia in the past) and modifying selection of data desired to be transmitted to an external monitoring device (e.g. by thinking about and picking out a specific dataset (e.g. including particular physiological signal types, designated timeframes, etc.) that they would like to be sent to an external system). Accordingly, claim 1 recites an abstract idea in the form of mathematical concepts and a certain method of organizing human activity. Claims 15 and 20 recite substantially similar subject matter as claim 1 and are found to recite an abstract idea under the same analysis.
Dependent claims 2-7, 10-14, 16-17, and 19 inherit the limitations that recite an abstract idea from their dependence on claims 1 and 15, respectively, and thus these claims also recite an abstract idea under the Step 2A – Prong 1 analysis. In addition, claims2-7, 10-14, 16-17, and 19 recite additional limitations that further describe the abstract idea identified in the independent claims.
Specifically, claims 2 and 16 specify that the physiological signals do not indicate an ongoing occurrence of the physiological event, which a clinician would be capable of ascertaining by looking at the physiological signals of a patient not currently undergoing an acute physiological event.
Claims 3 and 16 specify that predicting the onset of the physiological event comprises predicant a future occurrence of the physiological event, which a clinician could achieve by looking at the physiological signals of a patient not currently undergoing an acute physiological event but indicative of the potential for a future acute event.
Claims 4 and 17 specify that predicting the onset of the physiological event comprises determining a risk group within a risk stratification, which a clinician could achieve by using their medical expertise to classify a patient into a high, medium, or low risk group based on their physiological signals.
Claim 17 further recites predicting a future occurrence of the physiological event based at least in part on the risk group, which a clinician could achieve by considering the risk group when predicting a future event (e.g. finding a higher probability of a future event if the patient is in a high risk group).
Claim 5 specifies that determining the risk group comprises determining treatment guidance associated with the risk group, which a clinician could achieve by using their medical expertise to recommend a treatment type for each group, e.g. continued monitoring for a low risk group, preventative medication for a medium risk group, and surgical intervention for a high risk group.
Claim 6 recites generating a risk score corresponding to a risk of an occurrence of the physiological event at a future time, which is a mathematical operation; such an operation could also be achieved by a clinician using their medical expertise to calculate a risk score.
Claim 7 recites transmitting an alert to the user or a healthcare provider in response to the risk score satisfying a threshold score, which a clinician could achieve by comparing the calculated risk score to a threshold and then communicating an alert (e.g. verbally, in writing, etc.) to the patient or a colleague if the risk score exceeds the threshold.
Claim 10 recites outputting a recommendation of a treatment or intervention based on the predicted onset of the event, which a clinician could achieve by using their medical expertise to recommend a treatment for a patient.
Claims 11 and 19 recite identifying a patient cluster of a plurality of patient clusters based on similarity of a physiological characteristic and selecting the treatment recommendation based on the patient cluster, which a clinician could achieve by evaluating several potential patient clusters each associated with certain characteristics or traits and corresponding treatment recommendations, and selecting the cluster most similar to the current patient’s physiological characteristics so that the corresponding treatment recommendation of that cluster may be selected.
Claim 12 recites that each patient cluster corresponds to a centroid of a plurality of centroids and that the recommendation is selected based in part on a distance in latent space between a representation of the user and corresponding centroid of the identified patient cluster, which a clinician could achieve by utilizing mathematical graphing and distance concepts to identify the similar patient cluster and relevant corresponding treatment recommendation.
Claim 13 specifies that the physiological characteristic comprises one or more of a status of cardiac arrhythmia, an intervention type, or an outcome corresponding to an instance of an intervention of the intervention type. A clinician would be capable of evaluating these types of patient characteristics to identify an appropriate cluster of patients with similar characteristics as in claim 11.
Claim 14 specifies that the event is cardiac arrhythmia, which is a type of event that a clinician would be capable of assessing or predicting from physiological signal data.
However, recitation of an abstract idea is not the end of the analysis. Each of the claims must be analyzed for additional elements that indicate the abstract idea is integrated into a practical application to determine whether the claim is considered to be “directed to” an abstract idea.
Step 2A – Prong 2
The judicial exception is not integrated into a practical application. In particular, independent claims 1, 15, and 20 do not include additional elements that integrate the abstract idea into a practical application. The additional elements of claim 1 include an electronic device comprising an adhesive assembly comprising a housing that encloses a circuit board, a sensor in electrical communication with the circuit board and configured to be positioned in conformal contact with the surface of the user to detect the physiological signals of the user, and a hardware processor configured to perform the various steps of the invention, as well as specifying that the model is a machine learning model. Claims 15 and 20 recite substantially similar additional elements. The adhesive assembly and sensor additional elements amount to insignificant extra-solution activity in the form of data gathering because they merely serve as means of obtaining the physiological signals needed for the main data analysis steps of the invention (see MPEP 2106.05(g)). The use of a hardware processor (e.g. executing computer-executable instructions stored in a non-transitory computer storage medium as in claim 20) to perform the various steps of the invention amounts to instructions to “apply” the abstract idea using generic computer components because this element merely serves as a tool with which the otherwise-abstract functions of applying signals to a model, accessing various types of data, and training a model are digitized and/or automated (see MPEP 2106.05(f)). Similarly, specifying that the model is a machine learning model merely invokes this high-level type of computerized element as a tool with which to automate and/or digitize the otherwise-abstract functions of mathematical model fitting and execution to analyze data. Accordingly, claims 1, 15, and 20 as a whole are each directed to an abstract idea without integration into a practical application.
The judicial exception recited in dependent claims 2-7, 10-14, 16-17, and 19 is also not integrated into a practical application under a similar analysis as above. Claims 2-6, 10-14, 16-17, and 19 merely further describe the abstract idea of the independent claims without introducing any new additional elements of their own, and thus do not provide integration into a practical application. Claim 7 recites that the hardware processor transmits an alert to a device of the user or a healthcare provider, which merely digitizes/automates the otherwise-abstract function of sharing data between human actors such that this element also amounts to instructions to “apply” the exception in a computer environment.
Accordingly, the additional elements of claims 1-7, 10-17, and 19-20 do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. Claims 1-7, 10-17, and 19-20 are directed to an abstract idea.
Step 2B
The claims 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 the abstract idea into a practical application, the additional elements of an electronic device with a processor executing instructions to perform the applying, predicting, accessing, training, modifying, and machine learning model aspects of the invention amount to mere instructions to apply the exception using generic computer components. As evidence of the generic nature of the above recited additional elements, Examiner notes paras. [0155]-[0156] & [0161] of Applicant’s specification, where the computer elements (e.g. microprocessor) are described in terms of known, conventional components and device types (e.g. server, laptop computer, mobile device, etc.). See also paras. [0157]-[0158], noting various generic types of I/O devices for interaction with a user (e.g. encompassing a device of a user or healthcare provider as in claim 7) such as a touchpad or touchscreen, display device, GUI, etc.
Regarding the machine learning aspect of the event detection model, Examiner notes that it is well-understood, routine, and conventional to utilize machine learning models like neural networks for the purpose of physiological event detection, as evidenced by at least Thakur et al. (US 20200297230 A1) paras. [0078]-[0079]; Boleyn et al. (US 20190090769 A1) paras. [0025]-[0026] & [0106]; and Fornwalt et al. (US 20210076960 A1) paras. [0003] & [0180].
Regarding the adhesive assembly and sensor elements of the electronic device, as noted above, these additional elements amount to insignificant extra-solution activity in the form of a means of data gathering. Examiner notes that use of an adhesive assembly comprising a housing that encloses a circuit board and a sensor in electrical communication with the circuit board and configured to be positioned in conformal contact with the surface of the user to detect physiological signals is well-understood, routine, and conventional in the art, as evidenced by at least Hughes et al. (US 20160120433 A1) paras. [0007]-[0008] & [0084]-[0085]; Thakur paras. [0048]-[0050]; and Boleyn paras. [0007]-[0008] & [0043].
Further, the combination of these additional elements is not expanded upon in the specification as a unique arrangement and as such relies on the knowledge of one of ordinary skill in the art to understand the combination of components as a well-known and generic combination for automating an abstract idea that could otherwise be performed as a certain method of organizing human activity and thus do not provide an inventive concept. Additionally, the combination of adhesive sensor devices, computer processing hardware, and machine learning models to achieve physiological event monitoring and prediction is well-understood, routine, and conventional, as evidenced by at least Thakur Fig. 2 and paras. [0048]-[0050] & [0078]-[0079]; Boleyn Fig. 3 and paras. [0025]-[0026], [0043], & [0106]; and Fornwalt abstract and paras. [0153] & [0180]. Thus, when considered as a whole and in combination, claims 1-7, 10-17, and 19-20 are not patent eligible.
Claim Rejections - 35 USC § 103
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.
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 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.
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 1-6, 10, 14-17, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Thakur et al. (US 20200297230 A1) in view of Gensheimer et al. (Reference V on the PTO-892 mailed 3/11/2026).
Claim 1
Thakur teaches an electronic device for monitoring physiological signals of a user (Thakur Fig. 2, [0061]-[0062], noting arrhythmic risk stratification system monitors a user’s physiological signals to predict onset of a cardiac arrhythmia), the electronic device comprising:
an adhesive assembly comprising a housing that encloses a circuit board (Thakur [0050], noting ambulatory medical device includes a hermetically sealed housing that contains various circuits; the AMD may be a patch-based device as noted in [0048], indicating that the it may be embodied with an adhesive (i.e. patch-based) assembly);
a sensor in electrical communication with the circuit board and configured to be positioned in conformal contact with the surface of the user to detect the physiological signals of the user (Thakur [0048]-[0050], [0066], noting the ambulatory medical device includes sensor devices in communication with the various circuits, e.g. a cardiac sensor that senses cardiac information from electrodes positioned on a patient’s body surface); and
a hardware processor configured to apply the physiological signals to a machine learning model, wherein the machine learning model is configured to predict an onset of a physiological event based on the physiological signals of the user (Thakur [0069], [0078]-[0079], noting arrhythmic risk stratifier is embodied as part of a microprocessor circuit or other processor hardware and uses a machine learning model to determine an arrhythmia risk indication (e.g. onset timing) based on input physiologic information), and wherein the machine learning model is trained by:
accessing training data for a plurality of users separate from the user, wherein the training data comprises physiological data of the plurality of users (Thakur [0078], noting the machine learning model is trained using sensor data from a patient population (i.e. training data comprising physiological data of a plurality of users separate from the user));
training the machine learning model based on the training data (Thakur [0078], noting the machine learning model is trained using sensor data from a patient population),
wherein the hardware processor is further configured to modify a window of physiological signals monitored to detect an occurrence of the physiological event based on the prediction of the onset of the physiological event (Thakur [0086], noting the system can tune or adjust aspects of the sensors such as sensor data acquisition time, schedule, frequency, duration (i.e. window), sampling rate, etc. based on a current estimate of arrhythmia risk for a patient), and wherein modifying the window of physiological signals modifies a selection of physiological signals transmitted to an external monitoring device that implements a second machine learning model configured to classify the physiological event (Thakur [0053]-[0055], noting an external system 125 may receive device data from the AMD sensor including real-time or stored physiologic data from the patient and perform further classification processing on the data, e.g. to verify that a detected medical event is a true positive. Taken together with the tunable/adjustable data acquisition timing, scheduling, frequency, duration, sampling rate, etc. of [0086], these disclosures show that the system may modify the selection of physiological signals that are actively acquired by the system (e.g. by selecting the amount of data acquired over a modified time window) and thus transmitted to the external system for further processing as in [0053]-[0055], which meets the broadest reasonable interpretation of this limitation as outlined above).
In summary, Thakur teaches a system for applying non-invasively collected physiological data to a machine learning model that has been trained from patient population data to predict future onset of a physiological event such as a cardiac arrhythmia. The machine learning model can take many forms, including linear regression, decision tree, Naïve Bayes, support vector machine, neural network, etc. (see [0078]). However, this reference fails to explicitly disclose that the machine learning model training process includes for each user of the plurality of users, accessing a physiological status indicator, wherein the physiological status indicator is determined based on the training data; training the machine learning model based on the training data and the physiological status indicator, and wherein training the machine learning model comprises applying a failure time and a censor variable to a time to event loss function that predicts probabilities of survival/failure within a specified timeframe as required by the instant claim. However, Gensheimer teaches a specific method of training a clinical prediction machine learning model such as a neural network that includes accessing training data of a plurality of patients that each have known outcome indicators including failure time and censor variables (also considered equivalent to physiological status indicators determined based on the training data) and applying the known outcome indicators in the training data to a loss function to train the model to predict probabilities of survival or failure within a specified timeframe (Gensheimer abstract & “Implementation” section on Pg 5). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the machine learning model training method of Thakur to include use of known physiological status indicators like failure time and censor variables in a loss function for training the model to predict probabilities of survival/failure in a time period as in Gensheimer in order to utilize a modeling approach that avoids information loss when training the model and enables the generation of predicted survival curves (as suggested by Gensheimer abstract & “Implementation” section on Pg 5).
Claim 2
Thakur in view of Gensheimer teaches the electronic device of claim 1, and the combination further teaches wherein the physiological signals do not indicate an ongoing occurrence of the physiological event (Thakur [0043], [0079], noting the arrhythmia is predicted for a future onset rather than being currently indicated in the physiological signals).
Claim 3
Thakur in view of Gensheimer teaches the electronic device of claim 2, and the combination further teaches wherein predicting the onset of the physiological event comprises predicting a future occurrence of the physiological event (Thakur abstract, [0079], noting prediction of future onset of an arrhythmia).
Claim 4
Thakur in view of Gensheimer teaches the electronic device of claim 1, and the combination further teaches wherein predicting the onset of the physiological event comprises determining a risk group within a risk stratification (Thakur [0079], [0097], noting the arrhythmia risk indication can stratify a patient into categorical risk groups such as high, medium, or low).
Claim 5
Thakur in view of Gensheimer teaches the electronic device of claim 4, and the combination further teaches wherein determining the risk group comprises determining treatment guidance associated with the risk group (Thakur [0083], [0099], noting the system can deliver a recommendation or therapy (i.e. determine treatment guidance) based on the user’s risk indication being categorized as high risk).
Claim 6
Thakur in view of Gensheimer teaches the electronic device of claim 1, and the combination further teaches wherein predicting the onset of the physiological event comprises generating a risk score corresponding to a risk of an occurrence of the physiological event at a time subsequent to detection by the sensor of the physiological signals of the user (Thakur [0076], [0079], noting the arrhythmia risk indication includes a risk score indicating the risk of the patient developing an arrhythmia in the future).
Claim 10
Thakur in view of Gensheimer teaches the electronic device of claim 1, and the combination further teaches wherein the hardware processor is further configured to output a recommendation of a treatment or an intervention based on a prediction of the onset of the physiological event (Thakur [0082], [0099], noting the system can generate recommendations for treatments or interventions such as more aggressive arrhythmia monitoring, further testing to be performed, initiating or adjusting patient medication or other types of treatment, etc. based on the patient risk).
Claim 14
Thakur in view of Gensheimer teaches the electronic device of claim 1, and the combination further teaches wherein the physiological event comprises cardiac arrhythmia (Thakur abstract).
Claim 15
Thakur teaches a method comprising:
detecting physiological signals of a user using a sensor of a physiological signal monitor, wherein the sensor is configured to be placed in conformal contact with a surface of the user (Thakur [0048]-[0050], [0066], noting the ambulatory medical device includes patch-based or wearable sensor devices that collect physiological signals of a patient, e.g. a cardiac sensor that senses cardiac information from electrodes positioned on a patient’s body surface); and
applying the physiological signals detected by the sensor to a machine learning model, wherein the machine learning model is configured to predict an onset of a physiological event based on the physiological signals of the user (Thakur [0069], [0078]-[0079], noting arrhythmic risk stratifier uses a machine learning model to determine an arrhythmia risk indication (e.g. onset timing) based on input physiologic information), wherein the machine learning model is trained by:
accessing training data for a plurality of users, wherein the training data comprises historical physiological data of the plurality of users (Thakur [0078], noting the machine learning model is trained using sensor data from a patient population (i.e. training data comprising historical physiological data of a plurality of users));
training the machine learning model based on the training data (Thakur [0078], noting the machine learning model is trained using sensor data from a patient population),
wherein a window of physiological signals monitored to detect an occurrence of the physiological event is modified based on the prediction of the onset of the physiological event (Thakur [0086], noting the system can tune or adjust aspects of the sensors such as sensor data acquisition time, schedule, frequency, duration (i.e. window), sampling rate, etc. based on a current estimate of arrhythmia risk for a patient), and wherein modifying the window of physiological signals modifies a selection of physiological signals transmitted to an external monitoring device that implements a second machine learning model configured to classify the physiological event (Thakur [0053]-[0055], noting an external system 125 may receive device data from the AMD sensor including real-time or stored physiologic data from the patient and perform further classification processing on the data, e.g. to verify that a detected medical event is a true positive. Taken together with the tunable/adjustable data acquisition timing, scheduling, frequency, duration, sampling rate, etc. of [0086], these disclosures show that the system may modify the selection of physiological signals that are actively acquired by the system (e.g. by selecting the amount of data acquired over a modified time window) and thus transmitted to the external system for further processing as in [0053]-[0055] , which meets the broadest reasonable interpretation of this limitation as outlined above).
In summary, Thakur teaches a method for applying non-invasively collected physiological data to a machine learning model that has been trained from patient population data to predict future onset of a physiological event such as a cardiac arrhythmia. However, this reference fails to explicitly disclose that the machine learning model training process includes for each user of the plurality of users, accessing a physiological status indicator, wherein the physiological status indicator is determined based on the training data; training the machine learning model based on the training data and the physiological status indicator, and wherein training the machine learning model comprises applying a failure time and a censor variable to a time to event loss function that predicts probabilities of survival/failure within a specified timeframe as required by the instant claim. However, Gensheimer teaches a specific method of training a clinical prediction machine learning model such as a neural network that includes accessing training data of a plurality of patients that each have known outcome indicators including failure time and censor variables (also considered equivalent to physiological status indicators determined based on the training data) and applying the known outcome indicators in the training data to a loss function to train the model to predict probabilities of survival or failure within a specified timeframe (Gensheimer abstract & “Implementation” section on Pg 5). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the machine learning model training method of Thakur to include use of known physiological status indicators like failure time and censor variables in a loss function for training the model to predict probabilities of survival/failure in a time period as in Gensheimer in order to utilize a modeling approach that avoids information loss when training the model and enables the generation of predicted survival curves (as suggested by Gensheimer abstract & “Implementation” section on Pg 5).
Claim 16
Thakur in view of Gensheimer teaches the method of claim 15, and the combination further teaches wherein the physiological signals do not indicate an ongoing occurrence of the physiological event, and wherein the method further comprises predicting a future occurrence of the physiological event based on the physiological signals of the user (Thakur abstract, [0043], [0079], noting the arrhythmia is predicted for a future onset rather than being currently indicated in the physiological signals).
Claim 17
Thakur in view of Gensheimer teaches the method of claim 15, and the combination further teaches determining a risk group of the user within a risk stratification; and predicting a future occurrence of the physiological event based at least in part on the risk group (Thakur [0079], noting the arrhythmia risk indication can stratify a patient into categorical risk groups such as high, medium, or low and predict corresponding onset timing or timeframe of a future cardiac event).
Claim 20
Thakur teaches a non-transitory computer storage medium storing computer-executable instructions that, when executed by a processor, cause the processor to perform operations (Thakur [0069]-[0070], noting arrhythmic risk stratifier is embodied with processing hardware that can carry out operations encoded in a computer readable medium) comprising:
detecting physiological signals of a user using a sensor of a physiological signal monitor, wherein the sensor is configured to be placed in conformal contact with a surface of the user (Thakur [0048]-[0050], [0066], noting the ambulatory medical device includes patch-based or wearable sensor devices that collect physiological signals of a patient, e.g. a cardiac sensor that senses cardiac information from electrodes positioned on a patient’s body surface); and
applying the physiological signals detected by the sensor to a machine learning model, wherein the machine learning model is configured to predict an onset of a physiological event based on the physiological signals of the user (Thakur [0069], [0078]-[0079], noting arrhythmic risk stratifier uses a machine learning model to determine an arrhythmia risk indication (e.g. onset timing) based on input physiologic information), wherein the machine learning model is trained by:
accessing training data for a plurality of users, wherein the training data comprises historical physiological data of the plurality of users (Thakur [0078], noting the machine learning model is trained using sensor data from a patient population (i.e. training data comprising historical physiological data of a plurality of users));
training the machine learning model based on the training data (Thakur [0078], noting the machine learning model is trained using sensor data from a patient population),
wherein a window of physiological signals monitored to detect an occurrence of the physiological event is modified based on the prediction of the onset of the physiological event (Thakur [0086], noting the system can tune or adjust aspects of the sensors such as sensor data acquisition time, schedule, frequency, duration (i.e. window), sampling rate, etc. based on a current estimate of arrhythmia risk for a patient), and wherein modifying the window of physiological signals modifies a selection of physiological signals transmitted to an external monitoring device that implements a second machine learning model configured to classify the physiological event (Thakur [0053]-[0055], noting an external system 125 may receive device data from the AMD sensor including real-time or stored physiologic data from the patient and perform further classification processing on the data, e.g. to verify that a detected medical event is a true positive. Taken together with the tunable/adjustable data acquisition timing, scheduling, frequency, duration, sampling rate, etc. of [0086], these disclosures show that the system may modify the selection of physiological signals that are actively acquired by the system (e.g. by selecting the amount of data acquired over a modified time window) and thus transmitted to the external system for further processing as in [0053]-[0055] , which meets the broadest reasonable interpretation of this limitation as outlined above).
In summary, Thakur teaches a system for applying non-invasively collected physiological data to a machine learning model that has been trained from patient population data to predict future onset of a physiological event such as a cardiac arrhythmia. However, this reference fails to explicitly disclose that the machine learning model training process includes for each user of the plurality of users, accessing a physiological status indicator, wherein the physiological status indicator is determined based on the training data; training the machine learning model based on the training data and the physiological status indicator, and wherein training the machine learning model comprises applying a failure time and a censor variable to a time to event loss function that predicts probabilities of survival/failure within a specified timeframe as required by the instant claim. However, Gensheimer teaches a specific method of training a clinical prediction machine learning model such as a neural network that includes accessing training data of a plurality of patients that each have known outcome indicators including failure time and censor variables (also considered equivalent to physiological status indicators determined based on the training data) and applying the known outcome indicators in the training data to a loss function to train the model to predict probabilities of survival or failure within a specified timeframe (Gensheimer abstract & “Implementation” section on Pg 5). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the machine learning model training method of Thakur to include use of known physiological status indicators like failure time and censor variables in a loss function for training the model to predict probabilities of survival/failure in a time period as in Gensheimer in order to utilize a modeling approach that avoids information loss when training the model and enables the generation of predicted survival curves (as suggested by Gensheimer abstract & “Implementation” section on Pg 5).
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Thakur and Gensheimer as applied to claims 1 and 6 above, and further in view of Fornwalt et al. (US 20210076960 A1).
Claim 7
Thakur in view of Gensheimer teaches the electronic device of claim 6, and the combination further teaches wherein (Thankur [0045], [0056], [0082], claim 10, noting the system may generate an alert to notify a system user (e.g. a clinician or other healthcare personnel) at an output device about the risk or predicted future event).
In summary, the present combination teaches transmitting an alert to a device of a healthcare provider based on the generated risk or predicted event (which may be determined by comparison to thresholds as noted in [0076]-[0080]). However, the present combination fails to explicitly disclose that the alert is transmitted specifically in response to the risk score satisfying a threshold score. However, Fornwalt teaches generating a report (equivalent to an alert) for output at medical personnel user devices specifically responsive to the risk score exceeding a predetermined threshold (Fornwalt [0014], [0165]). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the alerting function of the combination such that it occurs specifically responsive to the risk score satisfying a threshold as in Fornwalt in order to alert a clinician only in cases where the risk score indicates that the patient will suffer from an event within a predetermined time period (as suggested by Fornwalt [0165]), thereby facilitating immediate intervention and reducing unnecessary alerts.
Claims 11-13 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Thakur and Gensheimer as applied to claims 1 and 10 or 15 above, and further in view of Van Berkel et al. (US 20220084669 A1).
Claim 11
Thakur in view of Gensheimer teaches the electronic device of claim 10, but the combination fails to explicitly disclose wherein the hardware processor is further configured to identify a patient cluster based at least in part on a similarity of a physiological characteristic between the user and patients of the patient cluster, wherein the patient cluster is one of a plurality of patient clusters, and wherein the recommendation of the treatment or the intervention is selected based in part on the patient cluster.
However, Van Berkel teaches selecting an appropriate treatment for a user by: identifying a patient cluster based at least in part on a similarity of a physiological characteristic between the user and patients of the patient cluster, wherein the patient cluster is one of a plurality of patient clusters (Van Berkel [0060], noting the system selects the most similar virtual patient for a new patient (i.e. user), where a virtual patient represents a near-homogeneous cluster of patients grouped based on clinical, demographic, socioeconomic, utilization, etc. features/characteristics as noted in [0052]-[0054] & [0062]; thus, selecting a most similar virtual patient out of a plurality of virtual patients is considered equivalent to identifying a patient cluster of a plurality of patient clusters based at least in part on a similarity of a physiological characteristic or feature of the new patient (user) and the virtual patient (cluster)); and wherein the recommendation of the treatment or the intervention is selected based in part on the patient cluster (Van Berkel [0060], noting once the system has selected the most similar virtual patient (i.e. cluster) for a given new patient (i.e. user), it selects a care plan for the new patient based on the care plan that is associated with the virtual patient). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the treatment recommendation function of the combination to include the specific patient cluster identification and selection of associated treatment plans as in Van Berkel in order to match new patients to optimized care plans that are likely to benefit those patients the most based on known features and outcomes of similar patients (as suggested by Van Berkel [0049]), thereby improving patient care.
Claim 12
Thakur in view of Gensheimer and Van Berkel teaches the electronic device of claim 11, and the combination further teaches wherein each patient cluster of the plurality of patient clusters corresponds to a medoid of a plurality of medoids (Van Berkel [0054], noting each virtual patient represents a medoid of its associated patient cluster), and wherein the hardware processor is further configured to select the recommendation based in part on a distance in latent space between a representation of the user and a corresponding medoid of the patient cluster identified based on the similarity of the physiological characteristic between the user and the patients of the patient cluster (Van Berkel Fig. 8, [0060]-[0061], noting the most similar virtual patient (i.e. medoid representation of a patient cluster) is identified for a new patient based on distance between a representation of the new patient and the virtual patient (i.e. medoid of a patient cluster) as mapped in a patient space such that the associated treatment plan is selected based on the shortest distance from a patient to a virtual patient medoid).
In summary, the present combination teaches a system that may select an appropriate treatment plan for a user by identifying a most similar patient cluster (represented by a virtual patient representing the medoid of the cluster) based on shortest distance in a mapped patient space between representations of the new patient and the virtual patient. Though the virtual patient represents a composite metric of an entire patient cluster such as a medoid, the present combination fails to explicitly disclose that the composite metric of a patient cluster is a centroid. However, Van Berkel also contemplates that a centroid may be a cluster parameter used for distance and similarity comparisons (Van Berkel [0059]). It therefore would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the virtual patient to be a representation of a given patient cluster’s centroid rather than its medoid because centroid is also shown to be a known cluster parameter that can be used for distance and similarity comparisons, and simple substitution of one known element for another producing a predictable result (i.e. selection of the cluster’s centroid instead of medoid as its virtual patient representation) renders the claim obvious.
Claim 13
Thakur in view of Gensheimer and Van Berkel teaches the electronic device of claim 12, and the combination further teaches wherein the physiological characteristic comprises one or more of a status of cardiac arrhythmia, an intervention type, or an outcome corresponding to an instance of an intervention of the intervention type (Van Berkel [0050]-[0052], noting patients are grouped as “similar” based on clinical, demographic, socioeconomic, utilization (i.e. intervention), etc. features/characteristics, considered to include clinical features like cardiac arrhythmia risk or status when considered in the context of the combination with Thakur and Gensheimer).
Claim 19
Thakur in view of Gensheimer teaches the method of claim 15, and the combination further teaches (Thakur [0082], [0099], noting the system can generate recommendations for treatments or interventions such as more aggressive arrhythmia monitoring, further testing to be performed, initiating or adjusting patient medication or other types of treatment, etc. based on the patient risk).
Though the present combination teaches generating treatment recommendations responsive to the predicted patient onset of a cardiac event, it fails to explicitly disclose identifying a patient cluster from a plurality of patient clusters based at least in part on a similarity of a physiological characteristic between the user and patients of the patient cluster and determining the treatment recommendation based at least in part on the patient cluster. However, Van Berkel teaches selecting an appropriate treatment for a user by: identifying a patient cluster from a plurality of patient clusters based at least in part on a similarity of a physiological characteristic between the user and patients of the patient cluster (Van Berkel [0060], noting the system selects the most similar virtual patient for a new patient (i.e. user), where a virtual patient represents a near-homogeneous cluster of patients grouped based on clinical, demographic, socioeconomic, utilization, etc. features/characteristics as noted in [0052]-[0054] & [0062]; thus, selecting a most similar virtual patient out of a plurality of virtual patients is considered equivalent to identifying a patient cluster of a plurality of patient clusters based at least in part on a similarity of a physiological characteristic or feature of the new patient (user) and the virtual patient (cluster)); and determining a recommendation of a treatment based at least in part on the patient cluster (Van Berkel [0060], noting once the system has selected the most similar virtual patient (i.e. cluster) for a given new patient (i.e. user), it selects a care plan for the new patient based on the care plan that is associated with the virtual patient). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the treatment recommendation function of the combination to include the specific patient cluster identification and selection of associated treatment plans as in Van Berkel in order to match new patients to optimized care plans that are likely to benefit those patients the most based on known features and outcomes of similar patients (as suggested by Van Berkel [0049]), thereby improving patient care.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Zadeh et al. (Reference U on the accompanying PTO-892) describes methods for training neural network machine learning time-to-event models using censored events and cost functions.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KAREN A HRANEK whose telephone number is (571)272-1679. The examiner can normally be reached M-F 8:00-4:00 ET.
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, Shahid Merchant can be reached at 571-270-1360. 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.
/KAREN A HRANEK/ Primary Examiner, Art Unit 3684
Prosecution Insights