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 .
Status of Claims
Claims 1-6, 8-13, 15-19, and 21-23 are currently pending and have been examined.
Claims 1-6, 8-13, and 15-19 have been amended.
Claims 7, 14, and 20 have been canceled.
Claims 21-23 are new.
Claims 1-6, 8-13, 15-19, and 21-23 have been rejected.
Information Disclosure Statement
The information disclosure statement (IDS) was submitted on 03/06/2024. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
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 1-6, 8-13, 15-19, and 21-23 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claimed invention is directed to an abstract idea without significantly more. Claims 1-6, 8-13, 15-19, and 21-23 are directed to a system, method, or product which are one of the statutory categories of invention. (Step 1: YES).
Independent Claim 1 discloses a method comprising: receiving medical data of a patient; determining whether the patient has a medical condition using the medical data and a diagnostic model including a first artificial intelligence (AI) model through an n-th AI model; and transmitting or displaying information identifying the determination of whether the patient has the medical condition, wherein the diagnostic model including the first AI model through the n-th AI model is trained by: receiving training data including a majority class of samples corresponding to medical data of patients that do not have the medical condition and a minority class of samples corresponding to medical data of patients that do have the medical condition; grouping a first group of samples of the majority class into a first sub-group based on determining that each sample of the first group of samples shares a first feature; grouping an n-th group of samples of the majority class into an n-th sub-group based on determining that each sample of the n-th group of samples shares an n-th feature; generating a first sub-group training dataset that includes the first group of samples and samples of the minority class of samples; generating an n-th sub-group training dataset that includes the n-th group of samples and samples of the minority class of samples; training the first AI model of the diagnostic model using the first sub-group training dataset; and training the n-th AI model of the diagnostic model using the nth sub-group training dataset.
Independent Claim 8 discloses a device comprising: a memory configured to store instructions; and one or more processors configured to execute the instructions to perform operations comprising: receiving medical data of a patient; determining whether the patient has a medical condition using the medical data and a diagnostic model including a first artificial intelligence (AI) model through an n-th AI model; and transmitting or displaying information identifying the determination of whether the patient has the medical condition, wherein the diagnostic model including a first AI model through the n-th AI model is trained by: receiving training data including a majority class of samples corresponding to medical data of patients that do not have the medical condition and a minority class of samples corresponding to medical data of patients that do have the medical condition; grouping a first group of samples of the majority class into a first sub-group based on determining that each sample of the first group of samples shares a first feature; grouping an n-th group of samples of the majority class into an n-th sub-group based on determining that each sample of the n-th group of samples shares an n-th feature; generating a first sub-group training dataset that includes the first group of samples and samples of the minority class of samples; generating an n-th sub-group training dataset that includes the n-th group of samples and samples of the minority class of samples; training the first AI model of the diagnostic model using the first sub-group training dataset; and training the n-th AI model of the diagnostic model using the nth sub-group training dataset.
Independent Claim 15 discloses a non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising: receiving medical data of a patient; determining whether the patient has a medical condition using the medical data and a diagnostic model including a first artificial intelligence (AI) model through an n-th AI model; and transmitting or displaying information identifying the determination of whether the patient has the medical condition, wherein the diagnostic model including the first AI model through the n-th AI model is trained by: receiving training data including a majority class of samples corresponding to medical data of patients that do not have the medical condition and a minority class of samples corresponding to medical data of patients that do have the medical condition; grouping a first group of samples of the majority class into a first sub-group based on determining that each sample of the first group of samples shares a first feature; grouping an n-th group of samples of the majority class into an n-th sub-group based on determining that each sample of the n-th group of samples shares an n-th feature; generating a first sub-group training dataset that includes the first group of samples and samples of the minority class of samples; generating an n-th sub-group training dataset that includes the n-th group of samples and samples of the minority class of samples; training theof the diagnostic model using the first sub-group training the n-th AI model of the diagnostic model using the nth sub-group training dataset.
The examiner is interpreting the above bolded limitations as additional elements as further discussed below. The following un-bolded limitations, given the broadest reasonable interpretation, cover the abstract idea of a mental process because they recite a process that could be practically performed in the human mind (i.e. observations, evaluations, judgements, and / or opinions). In this case, the steps of receiving medical data of a patient, determining whether the patient has a medical condition using the medical data and a diagnostic model and transmitting or displaying information identifying the determination of whether the patient has the medical condition, and receiving majority class of samples corresponding to medical data of patients that do not have the medical condition and a minority class of samples corresponding to medical data of patients that do have the medical condition; determining sub-groups of the majority class of samples based on features of the majority class of samples; generating sub-group training datasets that each include respective samples of the sub-groups of the majority class of samples and samples of the minority class of samples is reasonably interpreted as at least evaluations that could be performed by a human mentally or using a pen and paper.
Further, the remaining un-bolded limitations, are also merely directed to rules or instructions to determine whether a patient has a medical condition. The series of steps recited above describe managing personal behavior or relationships or interactions between people and thus are grouped as certain methods of organizing human activity which is an abstract idea.
The abstract ideas are being considered together as a single abstract idea for further analysis. (Step 2A- Prong 1: YES. The claims are abstract).
This judicial exception is not integrated into a practical application. Limitations that are not indicative of integration into a practical application include: (1) Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05.f), (2) Adding insignificant extra- solution activity to the judicial exception (MPEP 2106.05.g), (3) Generally linking the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05.h).
Independent Claim 1 discloses the following additional elements:
A first artificial intelligence (AI) model through an n-th model
Independent Claim 8 discloses the following additional elements:
A device comprising: a memory configured to store instructions; and one or more processors configured to execute the instructions to perform operations
A first artificial intelligence (AI) model through an n-th model
Independent Claim 15 discloses the following additional elements:
A non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations
A first artificial intelligence (AI) model through an n-th model
In particular, the first artificial intelligence (AI) model through the n-th model (of claims 1, 8, and 15), the device comprising: a memory configured to store instructions; and one or more processors configured to execute the instructions to perform operations (of claim 8), and the non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations (of claim 15) are recited at a high-level of generality such that it amounts to no more than mere instructions to implement an abstract idea by adding the words ‘apply it’ (or an equivalent) with the judicial exception.
Applicant’s specification para 31 states - The user device 180 may be configured to display information received from the diagnostic platform 120. For example, the user device 180 may be a smartphone, a laptop computer, a desktop computer, a wearable device, a medical device, a radiology device, or the like.
Wherein Fig. 1 of the instant application discloses the diagnostic platform implements the diagnostic model which comprises of AI models. Thus, the user device is applying the AI model to display information received from the diagnostic platform to implement the abstract idea. The user device is behaving as expected and is not improved in any way.
Further, the Applicant’s specification para 26 discloses, “the medical device 110 may be an electrocardiogram (ECG) device, an electroencephalogram (EEG) device, an ultrasound device, a magnetic resonance imaging (MRI) device, an X-ray device, a computed tomography (CT) device, or the like.” Therefore, the medical device is collecting data as is expected and is not being improved. The data is not collected in a more efficient way nor is the data collected that which could not be collected before.
Accordingly, these additional elements, when considered separately and as an ordered combination, do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
Accordingly, claim(s) 1, 8, and 15 are directed to an abstract idea(s) without a practical application. (Step 2A-Prong 2: NO: the additional claimed elements are not integrated into a practical application).
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim does 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 the first artificial intelligence (AI) model through the n-th AI model (of claims 1, 8, and 15), the device comprising: a memory configured to store instructions; and one or more processors configured to execute the instructions to perform operations (of claim 8), and the non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations (of claim 15) amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept ("significantly more’). MPEP2106.05(I)(A) indicates that merely saying "apply it” or equivalent to the abstract idea cannot provide an inventive concept ("significantly more").
Accordingly, even in combination, this additional element does not provide significantly more. As such the independent claims 1, 8, and 15 are not patent eligible. (Step 2B: NO. The claims do not provide significantly more).
Dependent claim(s) 2-6, 9-13, 16-19, and 21-23 are similarly rejected because they either further define/narrow the abstract idea and/or do not further limit the claim to a practical application or provide an inventive concept such that the claims are subject matter eligible even when considered individually or as an ordered combination. Dependent claims 4, 11, and 18 do further disclose the additional element(s) of a deep learning ensemble model which is further narrowing the additional element of the AI models as described above.
As such, the deep learning ensemble model (of claims 4, 11, and 18) further narrows the AI models of claims 1, 8, and 15 and further is recited at a high-level of generality such that it amounts to no more than mere instructions to implement an abstract idea by adding the words ‘apply it’ (or an equivalent) with the judicial exception. Accordingly, these additional elements, when considered separately and as an ordered combination, do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application.
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim does 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 element of the deep learning ensemble model amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept ("significantly more’). MPEP2106.05(I)(A) indicates that merely saying "apply it” or equivalent to the abstract idea cannot provide an inventive concept ("significantly more"). Accordingly, this additional element does not provide significantly more.
Therefore, the dependent claims are also directed to an abstract idea.
Thus, Claims 1-6, 8-13, 15-19, and 21-23 are 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.
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.
Claim(s) 1-3, 5, 8-10, 12, 15-17, 19 and 21-23 are rejected under 35 U.S.C. 103 as being unpatentable over Kashiwagi (US PG Pub 2023/0284983 A1) in view of Al-Shamaa (The Use of Hellinger Distance Undersampling Model to Improve the Classification of Disease Class in Imbalanced Medical Datasets), further in view of Mufti (Exploiting Machine Learning Algorithms and Methods for the Prediction of Agitated Delirium After Cardiac Surgery: Models Development and Validation Study).
Regarding Claim 1, Kashiwagi discloses:
A method comprising:
receiving medical data of a patient; determining whether the patient has a medical condition using the medical data and (Para 55 discloses a therapy selection support program that generate a discriminator (identifier) as a diagnostic marker or a classifier as a stratification marker through machine learning on the basis of measurement data on brain activities [received medical data] and that provide information related to selection of a therapy for a subject with depression symptoms on the basis of the results of measurement of brain activities of the subject using the discriminator (identifier) or the classifier as a biomarker.)
a diagnostic model including a first artificial intelligence (AI) model through an n-th AI model; and (Para 62 discloses a process of generating the identifier through the machine learning involves ensemble learning in which a plurality of identifier sub-models [a first through n-th AI model] are generated for the plurality of training sub-samples [a first through n-th sub-group training dataset] and the plurality of identifier sub-models are integrated with each other to generate the identifier model [and thus discloses a first through n-th model and a first through n-th sub-group training dataset based on the first through n-th sub groups])
transmitting or displaying information identifying the determination of whether the patient has the medical condition, wherein (Para 59 discloses the support information providing device includes a clustering processor and an interface device; the clustering processor calculates a probability at which the first subject is classified as belonging to each of the clusters by the clustering classifier, and reads at least two pieces of the therapy information selected in accordance with the probability from the database device; and the interface device outputs data for displaying the selected clusters and the corresponding pieces of the therapy information in association with each other.)
training the first AI model of the diagnostic model using the first sub-group training dataset; and training the n-th AI model of the diagnostic model using the n-th sub-group training dataset. (Para 58 discloses the processor is configured to, in the machine learning to generate the identifier model, generate a plurality of training sub-samples by executing under-sampling and sub-sampling from the first cohort and the second cohort, select features for clustering from a sum-set of features that are used to generate the identifier through the machine learning in accordance with a degree of importance of features that belong to the sum-set for each of the training sub-samples, and generate the clustering classifier by the multiple co-clustering method on the basis of the selected features for the clustering. Para 62 discloses a process of generating the identifier through the machine learning involves ensemble learning in which a plurality of identifier sub-models [a first through n-th AI model] are generated for the plurality of training sub-samples [a first through n-th sub-group training dataset] and the plurality of identifier sub-models are integrated with each other to generate the identifier model. Para 76 discloses the plurality of second subjects include a first cohort having a diagnosis label of a depression and a second cohort not having the diagnosis label of the depression. The clusters obtained as a result of the stratification are obtained by a clustering classifier obtained through a clustering process for results of measurement of brain functional connectivity correlation values. The clustering classifier is generated through a processing step of executing the clustering process for the plurality of second subjects.)
While Kashiwagi discloses the above limitations, and specifically Para 62 discloses, “a process of generating the identifier through the machine learning involves ensemble learning in which a plurality of identifier sub-models [a first through n-th AI model] are generated for the plurality of training sub-samples [a first through n-th sub-group training dataset] and the plurality of identifier sub-models are integrated with each other to generate the identifier model,” and thus discloses a first through n-th model and a first through n-th sub-group training dataset based on the first through n-th sub groups, it does not specifically disclose that the sub-groups are grouped “based on determining that each sample of the [first through n-th] group of samples shares a [first through n-th] feature.” Al-Shamaa’s “Introduction” section discloses, “a novel undersampling technique, named the Hellinger Distance Undersampling (HDUS) model, aimed at solving the imbalanced classification problem in medical datasets. The proposed model reduces healthy class samples to improve the classifying performance of the rare disease class. It adopts the Hellinger distance to measure the similarity between majority class instance and its neighbouring minority class instances, then chooses a number of the highest Hellinger distance values, and sums them up to be a similarity value for each majority instance. Finally, the model selects a subset from the majority instances, having top similarity values [wherein the shared feature is the top similarity values], and combined with the original minority class instances.” As such, the combination of the feature calculated by Al-Shamaa to determine a sub group of a majority class with the various sub-models generated for the plurality of training sub-samples as taught by Kashiwagi reads on the following limitations, “grouping a first group of samples of the majority class into a first sub-group based on determining that each sample of the first group of samples shares a first feature; grouping an n-th group of samples of the majority class into an n-th sub-group based on determining that each sample of the n-th group of samples shares an n-th feature; generating a first sub-group training dataset that includes the first group of samples and samples of the minority class of samples; [[and]] generating an n-th sub-group training dataset that includes the n-th group of samples and samples of the minority class of samples…”
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 therapy selection support device (brain functional connectivity correlation value clustering device, method, system) as taught by Kashiwagi with the use of Hellinger distance undersampling model as taught by Al-Shamaa to improve the classification of disease class in imbalanced medical datasets and effectively separate major class instances and minor class instances and boost the discrimination power for each class, thereby improving the classification accuracy for rare class (Al-Shamaa Introduction).
Kashiwagi discloses the above limitations, and Para 62 discloses, “a process of generating the identifier through the machine learning involves ensemble learning in which a plurality of identifier sub-models [a first through n-th AI model] are generated for the plurality of training sub-samples [a first through n-th sub-group training dataset] and the plurality of identifier sub-models are integrated with each other to generate the identifier model,” and thus discloses a first through n-th model and a first through n-th sub-group training dataset based on the first through n-th sub groups and thus reads on the diagnostic model including the first artificial intelligence (AI) model through an n-th AI model] is trained by… Al-Shamaa’s “Introduction” section further discloses, “a novel undersampling technique, named the Hellinger Distance Undersampling (HDUS) model, aimed at solving the imbalanced classification problem in medical datasets. The proposed model reduces healthy class samples to improve the classifying performance of the rare disease class… Finally, the model selects a subset from the majority instances, having top similarity values, and combined with the original minority class instances.” Wherein Algorithm 1: HDUS pseudocode of Al-Shamaa discloses the input is an imbalanced training dataset and the output is a balanced training dataset and thus reads one the following limitation, Mufti further discloses the following limitation:
receiving training data including a majority class of samples corresponding to medical data of patients that do not have the medical condition and a minority class of samples corresponding to medical data of patients that do have the medical condition; (The issue of outcome class imbalance section discloses in our dataset, the outcome class distribution is notably imbalanced (only 11.4% of patients developed delirium [minority class = patients that do have the medical condition, thus reads on majority class = patients that do not have the medical condition]). Typically, classification algorithms tend to predict the majority class very well but perform poorly on the minority class due to 3 main reasons [47-49]: (1) the goal of minimizing the overall error (maximize accuracy), to which the minority class contributes very little; (2) algorithm’s assumption that classes are balanced; and (3) the assumption that impact of making an error is equal. See further: Prediction model’s performance evaluation section)
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 combination of the therapy selection support device (brain functional connectivity correlation value clustering device, method, system) as taught by Kashiwagi and the use of Hellinger distance undersampling model as taught by Al-Shamaa with the majority and minority classes as taught by Mufti in order to generate a new balanced dataset by decreasing the number of the majority class instances, in order to reduce the difference between the minority and the majority classes and make the training more efficient (Mufti: The Issue of Outcome Class Imbalance section, para 3).
Regarding Claim 2, this claim recites the limitations of Claim 1 and as to those limitations is rejected for the same basis and reasons as disclosed above. The combination of Kashiwagi, Al-Shamaa and Mufti discloses the following limitation that Al-Shamaa further discloses:
The method of claim 1, wherein the first feature and the n-th feature are determined using clinical metadata associated with the majority class of samples. (The “Introduction” section discloses a novel undersampling technique, named the Hellinger Distance Undersampling (HDUS) model, aimed at solving the imbalanced classification problem in medical datasets. The proposed model reduces healthy class samples to improve the classifying performance of the rare disease class. It adopts the Hellinger distance to measure the similarity between majority class instance [the broadest reasonable interpretation of clinical metadata associated with the majority class of samples] and its neighbouring minority class instances , then chooses a number of the highest Hellinger distance values, and sums them up to be a similarity value for each majority instance. Finally, the model selects a subset from the majority instances, having top similarity values, and combined with the original minority class instances.)
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 combination of the therapy selection support device (brain functional connectivity correlation value clustering device, method, system) as taught by Kashiwagi and the majority and minority classes as taught by Mufti with the use of Hellinger distance undersampling model as taught by Al-Shamaa to improve the classification of disease class in imbalanced medical datasets and effectively separate major class instances and minor class instances and boost the discrimination power for each class, thereby improving the classification accuracy for rare class (Al-Shamaa Introduction).
Regarding Claim 3, this claim recites the limitations of Claim 1 and as to those limitations is rejected for the same basis and reasons as disclosed above. The combination of Kashiwagi, Al-Shamaa and Mufti discloses the following limitation that Al-Shamaa further discloses:
The method of claim 1, wherein the first feature and the n-th feature are determined by extracting the first feature and the n-th feature from the training data using a feature extraction technique. (The “Introduction” section discloses a novel undersampling technique, named the Hellinger Distance Undersampling (HDUS) model, aimed at solving the imbalanced classification problem in medical datasets. The proposed model reduces healthy class samples to improve the classifying performance of the rare disease class. It adopts the Hellinger distance to measure the similarity [feature extraction] between majority class instance and its neighbouring minority class instances , then chooses a number of the highest Hellinger distance values, and sums them up to be a similarity value for each majority instance. Finally, the model selects a subset from the majority instances, having top similarity values, and combined with the original minority class instances.)
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 combination of the therapy selection support device (brain functional connectivity correlation value clustering device, method, system) as taught by Kashiwagi and the majority and minority classes as taught by Mufti with the use of Hellinger distance undersampling model as taught by Al-Shamaa to improve the classification of disease class in imbalanced medical datasets and effectively separate major class instances and minor class instances and boost the discrimination power for each class, thereby improving the classification accuracy for rare class (Al-Shamaa Introduction).
Regarding Claim 5, this claim recites the limitations of Claim 1 and as to those limitations is rejected for the same basis and reasons as disclosed above. Further, while Kashiwagi discloses the above limitations, and specifically Para 62 discloses, “a process of generating the identifier through the machine learning involves ensemble learning in which a plurality of identifier sub-models [a first through n-th AI model] are generated for the plurality of training sub-samples [a first through n-th sub-group training dataset] and the plurality of identifier sub-models are integrated with each other to generate the identifier model,” and thus discloses a first through n-th model and a first through n-th sub-group training dataset based on the first through n-th sub groups, it does not specifically disclose that the sub-groups are grouped “based on determining that each sample of the [first through n-th] group of samples shares a [first through n-th] feature.” Al-Shamaa’s “Introduction” section discloses, “a novel undersampling technique, named the Hellinger Distance Undersampling (HDUS) model, aimed at solving the imbalanced classification problem in medical datasets. The proposed model reduces healthy class samples to improve the classifying performance of the rare disease class. It adopts the Hellinger distance to measure the similarity between majority class instance and its neighbouring minority class instances , then chooses a number of the highest Hellinger distance values, and sums them up to be a similarity value for each majority instance. Finally, the model selects a subset from the majority instances, having top similarity values [wherein the shared feature is the top similarity values], and combined with the original minority class instances.” As such, the combination of the feature calculated by Al-Shamaa to determine a sub group of a majority class with the various sub-models generated for the plurality of training sub-samples as taught by Kashiwagi reads on, “The method of claim 1, wherein the first feature and the n-th feature a same type of feature,” wherein the type of feature is the similarity value utilized to determine the balanced dataset of Al-Shamaa and it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to apply the process taught by Al-Shamaa to each sub-model taught by Kashiwagi in order to improve the classification of disease class in imbalanced medical datasets and effectively separate major class instances and minor class instances and boost the discrimination power for each class, thereby improving the classification accuracy for rare class (Al-Shamaa Introduction) and to improve the accuracy of the output by integrating a plurality of sub-models through ensemble learning.
Regarding claim 21, this claim recites the limitations of Claim 1 and as to those limitations is rejected for the same basis and reasons as disclosed above. As presented above, Kashiwagi discloses the first through n-th sub-group training datasets wherein a plurality of sub-models are trained based on various sub-samples, specifically Para 62 discloses, “a process of generating the identifier through the machine learning involves ensemble learning in which a plurality of identifier sub-models [a first through n-th AI model] are generated for the plurality of training sub-samples [a first through n-th sub-group training dataset] and the plurality of identifier sub-models are integrated with each other to generate the identifier model,” and thus discloses a first through n-th model and a first through n-th sub-group training dataset based on the first through n-th sub groups. The combination of Kashiwagi, Al-Shamaa and Mufti discloses the following limitation that Al-Shamaa further discloses:
(New) The method of claim 1, wherein the [first] sub-group training dataset includes all of the samples of the minority class of samples, and wherein the [n-th] sub-group training dataset includes all of the samples of the minority class of samples. (The “Introduction” section discloses a novel undersampling technique, named the Hellinger Distance Undersampling (HDUS) model, aimed at solving the imbalanced classification problem in medical datasets. The proposed model reduces healthy class samples to improve the classifying performance of the rare disease class. It adopts the Hellinger distance to measure the similarity between majority class instance and its neighbouring minority class instances , then chooses a number of the highest Hellinger distance values, and sums them up to be a similarity value for each majority instance. Finally, the model selects a subset from the majority instances, having top similarity values, and combined with the original minority class instances.)
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 combination of the therapy selection support device (brain functional connectivity correlation value clustering device, method, system) as taught by Kashiwagi and the majority and minority classes as taught by Mufti with the use of Hellinger distance undersampling model as taught by Al-Shamaa to improve the classification of disease class in imbalanced medical datasets and effectively separate major class instances and minor class instances and boost the discrimination power for each class, thereby improving the classification accuracy for rare class (Al-Shamaa Introduction).
As to claims 8-10, 12, and 22, the claims are directed to the device implementing the method of claims 1-3 and 21 and further recite a memory configured to store instructions and one or more processors configured to execute the instructions to perform operations (e.g., see Kashiwagi Para 283 teaching a computer main body including: a processor, a RAM for temporarily storing instructions of an application program…a non-volatile storage device for storing application programs, system programs, and data and para 284 teaching various functions… are realized by operation processes performed by CPU in accordance with a program) and are similarly rejected
As to claims 15-17, 19 and 23, the claims are directed to the non-transitory computer-readable medium implementing the method of claims 1-3 and 21 and are similarly rejected, and further recite instructions that, when executed by one or more processors, cause the one or more processors to perform operations (e.g., see Kashiwagi Para 283 teaching a computer main body including: a processor, a RAM for temporarily storing instructions of an application program…a non-volatile storage device for storing application programs, system programs, and data and para 284 teaching various functions… are realized by operation processes performed by CPU in accordance with a program).
Claim(s) 4, 11 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Kashiwagi (US PG Pub 2023/0284983 A1) in view of Al-Shamaa (The Use of Hellinger Distance Undersampling Model to Improve the Classification of Disease Class in Imbalanced Medical Datasets), further in view of Mufti (Exploiting Machine Learning Algorithms and Methods for the Prediction of Agitated Delirium After Cardiac Surgery: Models Development and Validation Study) and Nguyen (Deep Ensemble Learning Approaches in Healthcare to Enhance the Prediction and Diagnosing Performance: The Workflows, Deployments, and Surveys on the Statistical, Image-Based, and Sequential Datasets).
Regarding Claim 4, this claim recites the limitations of Claim 1 and as to those limitations is rejected for the same basis and reasons as disclosed above. While Kashiwagi para 62 discloses an ensemble model, “a process of generating the identifier through the machine learning involves ensemble learning in which a plurality of identifier sub-models are generated for the plurality of training sub-samples and the plurality of identifier sub-models are integrated with each other to generate the identifier model,” the combination of Kashiwagi, Al-Shamaa, and Mufti does not fully disclose the following limitation that Nguyen discloses:
The method of claim 1, wherein the first AI and the n-th AI model are deep learning ensemble models. (Section 1: Introduction, para 3 discloses in the healthcare support system, reaching the optimal performance is always the top priority for prediction and classification, where the decision of the healthcare workers needs to be accurate and tailored to each patient. Under these circumstances, ensemble techniques are one of the best choices [ 3]. Essentially, the ensemble learning technique combines many similar or different weak prediction models into a robust model. In other words, the technique is able to reduce variance and prevent overfitting phenomena in the training process [ 4,5]. As a result, it improves the accuracy and stability of the prediction model in classification and regression tasks. By incorporating DL models with ensemble learning techniques in this study, we propose three approaches collectively known as deep ensemble learning: deep-stacked generalization ensemble learning, gradient deep learning boosting, and Deep aggregation learning. In other words, by replacing the core learning units of the corresponding ensemble technique with suitable DL models, our proposed method can perform well with higher efficiency on all three data types… Section 2: Literature Review section, para 6 discloses the ensemble learning technique gains much reliability owing to its performance in combining multiple predictive and classification models into one strong model. Ensemble learning with the core learning unit as DL is an innovative and prospective method. Several studies implemented this combination and showed its high performance and reliability as a result. Suk et al. presented a deep ensemble sparse regression network model to diagnose brain diseases [ 20]… A study by An et al. proposed the concept of deep ensemble learning, named deep belief network, to classify Alzheimer’s disease [ 21]. See further Section 3.3 proposed deep ensemble learning approaches section.)
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 combination of the therapy selection support device (brain functional connectivity correlation value clustering device, method, system) as taught by Kashiwagi, the use of Hellinger distance undersampling model as taught by Al-Shamaa and the majority and minority classes as taught by Mufti with the deep ensemble learning approaches in healthcare to enhance the prediction and diagnosing performance as taught by Nguyen in order to reduce variance and prevent overfitting phenomena in the training process (Section 1: Introduction, para 3) and implement a reliable prediction model with high performance (Section 2: Literature Review, para 6).
As to claim 11, this claim is directed to the device implementing the method of claim 4 and further recites a memory configured to store instructions and one or more processors configured to execute the instructions to perform operations (e.g., see Kashiwagi Para 283 teaching a computer main body including: a processor, a RAM for temporarily storing instructions of an application program…a non-volatile storage device for storing application programs, system programs, and data and para 284 teaching various functions… are realized by operation processes performed by CPU in accordance with a program) and are similarly rejected.
As to claim 18, this claim is directed to the non-transitory computer-readable medium implementing the method of claim 4 and further recites instructions that, when executed by one or more processors, cause the one or more processors to perform operations (e.g., see Kashiwagi Para 283 teaching a computer main body including: a processor, a RAM for temporarily storing instructions of an application program…a non-volatile storage device for storing application programs, system programs, and data and para 284 teaching various functions… are realized by operation processes performed by CPU in accordance with a program) and are similarly rejected.
Claim(s) 6 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Kashiwagi (US PG Pub 2023/0284983 A1) in view of Al-Shamaa (The Use of Hellinger Distance Undersampling Model to Improve the Classification of Disease Class in Imbalanced Medical Datasets), further in view of Mufti (Exploiting Machine Learning Algorithms and Methods for the Prediction of Agitated Delirium After Cardiac Surgery: Models Development and Validation Study) and Rodriguez (Rotation forest: A new classifier ensemble method).
Regarding Claim 6, this claim recites the limitations of Claim 1 and as to those limitations is rejected for the same basis and reasons as disclosed above. The combination of Kashiwagi, Al-Shamaa, and Mufti does not fully disclose the following limitation that Rodriguez discloses:
The method of claim 1, wherein the first feature and the n-th feature are different types of features . (Split F randomly into K subsets (K is a parameter of the algorithm). The subsets may be disjoint or intersecting. To maximize the chance for high diversity, we chose disjoint subsets. For simplicity, suppose that K is a factor of n so that each feature subset contains M=n/K features. Denote by Fi,j the jth subset of features for the training set of classifier Di. For every such subset, select randomly a nonempty subset of classes and then draw a bootstrap sample of objects, of size 75 percent of the data count. Run PCA using only the M features in Fi,j and the selected subset of X. Store the coefficients of the principal components, a(1)i,j,…,a(Mj)i,j, each of size M×1. Note that it is possible that some of the eigenvalues are zero, therefore, we may not have all M vectors and, so, Mj≤M.)
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 combination of the therapy selection support device (brain functional connectivity correlation value clustering device, method, system) as taught by Kashiwagi, the use of Hellinger distance undersampling model as taught by Al-Shamaa and the majority and minority classes as taught by Mufti with the Rotation Forest: A New Classifier Ensemble Method as taught by Rodriguez in order to promote diversity of the models by training models on different feature extracted subsets wherein ensemble learning methods combine multiple models to improve performance by exploiting their diversity.
As to claim 13, this claim is directed to the device implementing the method of claim 6 and further recites a memory configured to store instructions and one or more processors configured to execute the instructions to perform operations (e.g., see Kashiwagi Para 283 teaching a computer main body including: a processor, a RAM for temporarily storing instructions of an application program…a non-volatile storage device for storing application programs, system programs, and data and para 284 teaching various functions… are realized by operation processes performed by CPU in accordance with a program) and are similarly rejected.
Response to Arguments
Applicant’s arguments filed 5/1/2026 with respect to 35 U.S.C. § 101 have been fully considered, but are not persuasive.
The Applicant argues that the claims are not directed to a mental process because and quotes the MPEP that “claims do not recite a mental process when they do not contain limitations that can practically be performed in the human mind.” The Examiner respectfully disagrees and points to MPEP 2106.04(a)(2) (III) (C) that clarifies that claims that require a computer as a tool to perform the concept still recite a mental process. The training steps were not recited as part of the limitations that were found to recite a mental process. The specific limitations of the independent claims that were found to recite a mental process were the steps of, “receiving medical data of a patient, determining whether the patient has a medical condition using the medical data and a diagnostic model and transmitting or displaying information identifying the determination of whether the patient has the medical condition, and receiving majority class of samples corresponding to medical data of patients that do not have the medical condition and a minority class of samples corresponding to medical data of patients that do have the medical condition; determining sub-groups of the majority class of samples based on features of the majority class of samples; generating sub-group training datasets that each include respective samples of the sub-groups of the majority class of samples and samples of the minority class of samples.” These limitations disclose observations, evaluations, judgements, and/or opinions that a human being could perform mentally and thus the claims are directed to a mental process.
Further, the Applicant argues that the claims are not directed to certain methods of organizing human activity. The Examiner respectfully disagrees. MPEP 2106. 04(a)(2)(II) states that a claimed invention is directed to certain methods of organizing human activity if the identified claim elements contain limitations that encompass fundamental economic principles or practices, commercial or legal interactions, or managing personal behavior or relationships or interactions between people (including social activities, teaching, and following rules or instructions). The Examiner submits that the identified claim elements represent a series of rules or instructions that a person or persons, with or without the aid of a computer, would follow to determine whether a patient has a medical condition. The Applicant has not pointed to anything in the claims that fall outside of this characterization. The AI models were found to be an additional element and not a part of the abstract idea. The generation of “more balanced training datasets,” falls into the abstract idea because it describes rules or instructions a person or persons, with or without the aid of a computer, would follow to determine which sub-group to assign patients to based on shared features to determine whether a patient has a medical condition. Because the claim elements fall under a series of rules or instructions that a person or persons would follow to determine whether a patient has a medical condition., the claimed invention is directed to an abstract idea.
Additionally, the Applicant argues that the claims provide a practical application to the abstract idea.
Specifically, because the claims improve the technical field of training AI models using training datasets that “reduce an imbalance between a majority class of samples and minority class of samples.” The Examiner respectfully disagrees. MPEP 2106.04(d)(1) states “the word ‘improvements’ in the context of this consideration is limited to improvements to the functioning of a computer or any other technology/technical field, whether in Step 2A Prong Two or in Step 2B.” Here, there is no improvement to the computer (the processor of the device is performing as expected) nor is there an improvement to another technology. Because neither type of improvement is present in the claims, an improvement to technology is not present and there is no practical application.
The argument that the claims recite an improvement in the technical field of training AI models and to systems that use AI models by using training datasets that reduce an imbalance between a majority class of samples and a minority class of samples is not persuasive. The claims recite organizing the patient data into various sub groups dependent on features of the patients and, to paraphrase, a portion of the majority class samples are selected and paired with the entirety of the minority class samples to create a balanced dataset. The claims merely apply the AI model to the abstract idea has the AI model is performing as expected, it is trained and outputs data based on the training and input data it is given. The claims merely apply the AI model to the abstract idea. The AI model is performing as expected, it is trained and outputs data based on the data it is given. The problem of imbalanced datasets (wherein the size of the majority class of samples varies from the size of the minority class of samples) is not a problem caused by the processor or the AI models that are involved in the process. At best, Applicant’s identified problem is a business problem wherein the collected data is imbalanced.
Further, the Applicant argues that the specification discloses a technical problem associated with imbalanced training datasets. The Examiner respectfully disagrees. MPEP 2106.04(d)(1) and MPEP 2106.05(a) indicates that a practical application may be present where the claimed invention provides a technical solution to a technical problem. See, e.g., DDR Holdings, LLC. v. Hotels.com, L.P., 773 F.3d 1245, 1259 (Fed. Cir. 2014) (finding that claiming a website that retained the “look and feel” of a host webpage provided a technological solution to the problem of retention of website visitors by utilizing a website descriptor that emulated the “look and feel” of the host webpage, where the problem arose out of the internet and was thus a technical problem). Here, the Applicant’s argued problem is not a technological problem caused by the device comprising a memory and one or more processors or the AI models themselves (the technological environments to which the claims are confined). The problem of imbalanced datasets (wherein the size of the majority class of samples varies from the size of the minority class of samples) is not a problem caused by the processor or the AI models that are involved in the process. At best, Applicant’s identified problem is a business problem wherein the collected data is imbalanced. Because no technological problem is present, the claims do not provide a practical application.
The Applicant then argues that the claims provide significantly more in light of the alleged technical improvement. This argument is not persuasive as discussed above under the practical application argument. The claims determine which specific data to train the AI models on, however, it is not applying the trained data in a novel way. The claims merely apply the AI model to the abstract idea has the AI model is performing as expected, it is trained and outputs data based on the training and input data it is given. The problem of imbalanced datasets (wherein the size of the majority class of samples varies from the size of the minority class of samples) is not a problem caused by the processor or the AI models that are involved in the process. At best, Applicant’s identified problem is a business problem wherein the collected data is imbalanced.
Applicant’s arguments filed 5/1/2026 with respect to 35 U.S.C. § 103 have been fully
considered and are persuasive regarding the newly added limitations. Therefore, the previous 35 U.S.C.
§ 103 rejection has been withdrawn. However, upon further consideration, a new grounds of rejection under 35 U.S.C. § 103 necessitated by Applicant’s amendments as disclosed above.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SARA J MORICE DE VARGAS whose telephone number is (703)756-4608. The examiner can normally be reached M-F 8:30-5:30 pm.
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, Peter H. Choi can be reached at (469)295-9171. 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.
/SARA JESSICA MORICE DE VARGAS/Examiner, Art Unit 3681
/PETER H CHOI/Supervisory Patent Examiner, Art Unit 3681
Prosecution Insights