DETAILED CORRESPONDANCE
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
This final office action on merits is in response to the communication received on 12/22/2025. Claims 2, 4-6, 10-11, 16-17 are cancelled. Claims 27-28 are new. Amendments to claims 1, 14, 18, and 20 are acknowledged and have been carefully considered. Claims 1, 3, 7-9, 12-15, 18-28 are pending and considered below.
Specification
The following title is suggested: AUTOMATED DETECTION AND MANAGEMENT OF VALVULAR HEART DISEASE USING MACHINE LEARNING
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, 3, 7-9, 12-15, 18-28 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1
Under step 1, the analysis is based on MPEP 2106.03, and claims 1, 3, 7-9, 12-13, 21-28 are drawn to a system, claims 14-15, and 18-19 are drawn to a method, and claim 20 is drawn to a non-transitory machine-readable storage medium. Thus, each claim, on its face, is directed to one of the statutory categories (i.e., useful process, machine, manufacture, or composition of matter) of 35 U.S.C. 101.
Step 2A Prong One
Claim 1 recites the limitations of repeatedly determining, over the time, updated severity measures for the new patients based on analysis of the tracked patient data as additional components of the new multi- modal data are received; repeatedly applying, responsive to the updated severity measures exceeding a threshold severity measure, the trained version of the ensemble-based outcomes forecasting model to the new multi-modal data for the new patients to predict the patient outcomes for the new patients; and repeatedly determining whether the new patients should proceed or not proceed with the cardiac valve procedure based on the patient outcomes predicted for the new patients and the updated severity measures. These limitations, as drafted, are processes that, under their broadest reasonable interpretation, cover performances of the limitations in the mind or by using a pen and paper. But for the “a memory that stores computer-executable instructions; and a processor, coupled to the memory, that executes the computer-executable instructions to perform operations” language, the claim encompasses a user reviewing patient data, evaluating severity measures, comparing the severity measures to a threshold, estimating potential outcomes, and deciding whether a patient should proceed with a cardiac valve procedure in their mind or by using a pen and paper. The mere nominal recitations of a memory that stores computer-executable instructions; and a processor, coupled to the memory, that executes the computer-executable instructions to perform operations do not take the claim limitations out of the mental process groupings. Thus, the claim recites a mental process which is an abstract idea.
Independent claims 14 and 20 recite identical or nearly identical steps with respect to claim 1 (and therefore also recite limitations that fall within this subject matter grouping of abstract ideas), and these claims are therefore determined to recite an abstract idea under the same analysis.
Under Step 2A Prong Two
The claimed limitations, as per claim 1, include:
a memory that stores computer-executable instructions; and
a processor, coupled to the memory, that executes the computer-executable instructions to perform operations, comprising:
training an ensemble-based outcomes forecasting model to predict patient outcomes resulting from undergoing a cardiac valve procedure using multi-modal training data for a plurality of different patients, wherein the training comprising separately training different machine learning sub-models of the ensemble-based outcomes forecasting model to predict preliminary patient outcome data and mapping the preliminary patient outcome data to the patient outcomes using a multilayer perceptron, resulting in a trained version of the ensemble- based outcomes forecasting model, wherein the patient outcomes include a length of stay and a readmission risk;
collecting, over time, new multi-modal data for new patients that are potential candidates for the cardiac valve procedure as the new multi-modal data is received at one or more electronic healthcare information systems, wherein the new multi-model data includes parameters represented in the multi-modal training data;
maintaining, in the memory, tracked patient data for each of the new patients, wherein the tracked patient data comprises patient identifiers, timestamps associated with receipt of the new multi-modal data, and previously determined severity measures representing a measure of severity of a cardiac condition of the respective new patients;
repeatedly determining, over the time, updated severity measures for the new patients based on analysis of the tracked patient data as additional components of the new multi- modal data are received;
repeatedly applying, responsive to the updated severity measures exceeding a threshold severity measure, the trained version of the ensemble-based outcomes forecasting model to the new multi-modal data for the new patients to predict the patient outcomes for the new patients;
repeatedly determining whether the new patients should proceed or not proceed with the cardiac valve procedure based on the patient outcomes predicted for the new patients and the updated severity measures; and
notifying a healthcare provider in response to a determination that a patient of the new patients should proceed with the cardiac valve procedure.
Examiner Note: underlined elements indicate additional elements of the claimed invention identified as performing the steps of the claimed invention.
The judicial exception expressed in claim 1 is not integrated into a practical application. The claim as a whole merely describes how to generally “apply” the concept of evaluating patient data to determine severity measures, predicting potential outcomes, and deciding whether a patient should proceed with a cardiac valve procedure in a computer environment. The claimed computer components (i.e., a memory that stores computer-executable instructions; a processor, coupled to the memory, that executes the computer-executable instructions to perform operations, comprising: training an ensemble-based outcomes forecasting model to predict patient outcomes resulting from undergoing a cardiac valve procedure using multi-modal training data for a plurality of different patients, wherein the training comprising separately training different machine learning sub-models of the ensemble-based outcomes forecasting model to predict preliminary patient outcome data and mapping the preliminary patient outcome data to the patient outcomes using a multilayer perceptron, resulting in a trained version of the ensemble- based outcomes forecasting model, wherein the patient outcomes include a length of stay and a readmission risk; and maintaining, in the memory, tracked patient data for each of the new patients, wherein the tracked patient data comprises patient identifiers, timestamps associated with receipt of the new multi-modal data, and previously determined severity measures representing a measure of severity of a cardiac condition of the respective new patients) are recited at a high level of generality and are merely invoked as tools to perform an existing process of analyzing patient information, estimating treatment outcomes, and determining whether medical intervention should occur. Simply implementing the abstract idea on a generic computer is not a practical application of the abstract idea. Accordingly, alone and in combination, these additional elements do not integrate the abstract idea into a practical application.
The judicial exception expressed in claim 1 is not integrated into a practical application. The claim recites the additional elements of collecting, over time, new multi-modal data for new patients that are potential candidates for the cardiac valve procedure as the new multi-modal data is received at one or more electronic healthcare information systems, wherein the new multi-model data includes parameters represented in the multi-modal training data; and notifying a healthcare provider in response to a determination that a patient of the new patients should proceed with the cardiac valve procedure. These limitations are recited at a high level of generality (i.e., as a general means of receiving patient information and communicating the results of the analysis), and amounts to mere data gathering and insignificant activity, which is a form of insignificant extra-solution activity. Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application. The claim is directed to an abstract idea.
Therefore, under step 2A, the claims are directed to the abstract idea, and require further analysis under Step 2B.
Under step 2B
Claim 1 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed with respect to Step 2A, the claim as a whole merely describes how to generally “apply” the concept of evaluating patient data to determine severity measures, predicting potential outcomes, and deciding whether a patient should proceed with a cardiac valve procedure in a computer environment. Thus, even when viewed as a whole, nothing in the claim adds significantly more (i.e., an inventive concept) to the abstract idea.
For claim 1, under step 2B, the additional elements of collecting, over time, new multi-modal data for new patients that are potential candidates for the cardiac valve procedure as the new multi-modal data is received at one or more electronic healthcare information systems, wherein the new multi-model data includes parameters represented in the multi-modal training data; and notifying a healthcare provider in response to a determination that a patient of the new patients should proceed with the cardiac valve procedure have been evaluated. The system comprising a processor, coupled to the memory performs a general function of receiving patient data for subsequent analysis, which represents a well-understood, routine, and conventional activity in the field of healthcare information processing and medical data management. The specification discloses that the processor is used in its ordinary capacity as a data input device and does not describe any improvement to the computer itself or to the functioning of the overall computer system (see [0089]). Also noted in Electric Power Group, LLC v. Alstom S.A., 830 F.3d 1350, 1354, 119 USPQ2d 1739, 1742 (Fed. Cir. 2016), merely collecting information for analysis without a technological improvement does not add significantly more to an abstract idea. The use of the system is no more than collecting information before performing analysis and providing the results of the analysis to a healthcare provider and does not integrate the abstract idea into a practical application. Additionally, as noted in In re Brown, 645 Fed. App'x 1014, 1016-1017 (Fed. Cir. 2016), merely notifying the user of the represents an insignificant application of the underlying mental process, as the transmission simply communicates the result of the determination and does not impose any meaningful limitation or add any technological improvement. Therefore, the claim does not recite an inventive concept and is not patent eligible.
Claims 3, 7-9, 12-13, 15, 18-19, and 21-28 recite no further additional elements, and only further narrow the abstract idea. The previously identified additional elements, individually and as a combination, do not integrate the narrowed abstract idea into a practical application for reasons similar to those explained above, and do not amount to significantly more than the narrowed abstract idea for reasons similar to those explained above.
Thus, as the dependent claims remain directed to a judicial exception, and as the additional elements of the claims do not amount to significantly more, the dependent claims are not patent eligible.
Therefore, the claims here fail to contain any additional element(s) or combination of additional elements that can be considered as significantly more and the claim is rejected under 35 U.S.C. 101 for lacking eligible 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.
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, 3, 7-9, 12-15, 18-28 are rejected under 35 U.S.C. 103 as being unpatentable over Zimmerman et al. (U.S. Patent Publication 2023/0245782 A1), referred to hereinafter as Zimmerman, in view of Otto et al. (Otto et al., "2020 ACC/AHA Guideline for the Management of Patients With Valvular Heart Disease: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines", Circulation, Vol. 143, No. 5, 2021, pages e90-e104), referred to hereinafter as Otto, and Nemati et al. (U.S. Patent Publication 2025/0095857 A1), referred to hereinafter as Nemati.
Regarding claim 1, Zimmerman teaches a system, comprising (Zimmerman [0025] “In yet another aspect, the present disclosure provides a system including at least one processor coupled to at least one memory including instructions.”):
a memory that stores computer-executable instructions (Zimmerman [0025] “In yet another aspect, the present disclosure provides a system including at least one processor coupled to at least one memory including instructions.”); and
a processor, coupled to the memory, that executes the computer-executable instructions to perform operations, comprising (Zimmerman [0025] “In yet another aspect, the present disclosure provides a system including at least one processor coupled to at least one memory including instructions.”):
training an ensemble-based outcomes forecasting model to predict patient outcomes using multi-modal training data for a plurality of different patients, wherein the training comprising separately training different machine learning sub-models of the ensemble-based outcomes forecasting model to predict preliminary patient outcome data and mapping the preliminary patient outcome data to the patient outcomes using a multilayer perceptron, resulting in a trained version of the ensemble- based outcomes forecasting model, wherein the patient outcomes include a length of stay and a readmission risk (Zimmerman [0304] “In one example, a composite model may include an ensemble of machine learning and/or deep learning models trained from heterogeneous data including but not limited to ECG, demographics, labs and vitals.”, Zimmerman [0382] “In some embodiments, the method may be diagnostic, whereby the clinical data can include outcome data, such as whether or not a patient developed atrial fibrillation (AF or Afib) in a time period following the day that the ECG was taken. In other embodiments, the clinical data may be used in a predictive sense, e.g., to determine based on that data a likelihood that the patient would develop Afib within a certain time period following the day that the ECG was taken.” and Zimmerman [0292] “The composite model may be any machine learning algorithm which analyzes an ECG waveform and basic demographic data (age/sex) readily available in a digital ECG file and consist of a deep neural network (DNN) or convolutional NN (CNN) trained on data from hundreds of thousands of patients and echocardiograms, and more than a million ECGs.” and Zimmerman [0206] “Moreover, additional variables may be added into a predictive model for purposes of both improving the prediction accuracy of the endpoints and identifying treatments which can positively impact the predicted bad outcome. For example, by extracting laboratory values (blood cholesterol measurements such as LDL/HDL/total cholesterol, blood counts such as hemoglobin/hematocrit/white blood cell count, blood chemistries such as glucose/sodium/potassium/liver and kidney function labs, and additional cardiovascular markers such as troponins and natriuretic peptides), vital signs (blood pressures, heart rate, respiratory rate, oxygen saturation), imaging metrics (such as cardiac ejection fractions, cardiac chamber volumes, heart muscle thickness, heart valve function), patient diagnoses (such as diabetes, chronic kidney disease, congenital heart defects, cancer, etc.) and treatments (including procedures, medications, referrals for services such as cardiac rehabilitation, dietary counseling, etc.), a model's accuracy may be improved. Some of these variables are “modifiable” risk factors that can then be used as inputs to the models to demonstrate the benefit of using a particular therapy. For example, a prediction may identify a patient as a 40% likelihood of developing atrial fibrillation in the next year, however, if the model was able to identify that the patient was taking a beta blocker, the predicted risk would drop to 20% based on the increased data available to the predictive model. In one example, demographic data 416 and patient data 1304 may be supplemented with these additional variables, such as the extracted laboratory values or modifiable risk factors.”);
collecting, over time, new multi-modal data for new patients that are potential candidates as the new multi-modal data is received at one or more electronic healthcare information systems, wherein the new multi-model data includes parameters represented in the multi-modal training data (Zimmerman [0140] “In some embodiments, the model 400 can additionally receive electronic health record (EHR) data points such as demographic data 416, which can include age and sex/gender as input features to the network, where sex can be encoded into binary values for both male and female, and age can be cast as a continuous numerical value corresponding to the date of acquisition for each 12-lead resting state ECG.”, Zimmerman [0328] “In one example, model inputs may include one or more patient features selected for their importance to the output high-risk label/prediction, including but not limited to demographic features such as age, sex, or other EHR-derived features, in addition to ECG-derived values. For example, LDL diagnostic testing values may account for the highest percentage of the risk assessment and be a required input. In another example, LDL values, ECGs, and age may collectively account for the highest percentage of the risk assessment and be required inputs. In yet another model, inputs may include LDL values, ECGs, age, blood pressure, HTN, AF status, BMI, diabetes, smoking, gender, and/or CHF. In an even more exhaustive implementation, a stroke prediction may include the top X features, where X is an integer. In one example, the integer, X, may be 20 features. Consistent with a 20 feature embodiment, a top 20 features may include: patient age, INDEX_CCI, STROKE_YN, AF_Target, Labs_AlC, Vitals_Weight, Vitals_Height, Demographics_SMOKER_FLG1, Anti_coag, Vitals_BMI, Labs_GLUCOSE, Labs_SODIUM, Vitals_BP_Systolic, Labs_LDL, Medications_ANTICOAGULANTS, Labs_HDL, Labs_HEMOGLOBIN, ECG_R_AXIS, and Echo_measurements_LAV_MOD_sp2. [0329] In some embodiments, all features having a contribution to the high-risk determination with greater than a weight of 1% may be included in the inputs to a composite model. Consistent with a weight inclusion model, features may be selected to include: Vitals_BMI, Vitals_BP_Diastolic, Vitals_BP_Systolic, Vitals_Heart_Rate, Vitals_Height, Vitals_Weight, Demographics_FRS, Demographics_PCE, INDEX_CCI, CHADSVASC_SCORE, CHADS_SCORE, Demographics_PT_AGE, Demographics_PT_RACE, Demographics_PT_SEX, Demographics_SMOKER_FLG, ICD_Phenotypes_AOR, ICD_Phenotypes_AOS, ICD_Phenotypes_IVS, ICD_Phenotypes_LEF, ICD_Phenotypes_MIR, ICD_Phenotypes_MIS, ICD_Phenotypes_PUR, ICD_Phenotypes_PUS, ICD_Phenotypes_TRR, ICD_Phenotypes_TRS, Labs_AlC, Labs_BILI, Labs_BNP, Labs_BUN, Labs_CHOLESTEROL Labs_CKMB, Labs_CREATININE, Labs_CRP, Labs_D_dimer, Labs_eGFR, Labs_GLUCOSE, Labs_HDL, Labs_HEMOGLOBIN, Labs_LDH, Labs_LDL, Labs_LYMPHOCYTES Labs_POTASSIUM Labs_PRO_BNP, Labs_SODIUM, Labs_Triglyceride, Labs_TROPONIN_I, Labs_TROPONIN_T, Labs_URIC_ACID, Labs_VLDL, Medications_ACE_INHIBITORS, Medications_ANGIOTENSIN_II_RECEPTOR_ANTAGONISTS, Medications_ANTICOAGULANTS, Medications_ANTIDIABETIC_MEDICATION, Medications_ANTIHYPERTENSIVE, Medications_DIGOXIN,Medications_ERX_EBBB_HEART_FAILUREMedications_ERX_SPIRONOLACTONE_EPLE R, ONE_HEART_FAILURE, Medications_LOOP_DIURETICS, ECG_Measurements, Echo_Measurements, ECG_Findings, HF_YN, HTN_YN, AGE_GTE_75_YN, DM_YN, STROKE_YN, VASC_DISC_YN, AGE_65_74_YN, and FEMALE_YN.”);
maintaining, in the memory, tracked patient data for each of the new patients, wherein the tracked patient data comprises patient identifiers, timestamps associated with receipt of the new multi-modal data, and previously determined severity measures representing a measure of severity of a cardiac condition of the respective new patients (Zimmerman [0034] “In one aspect, a method includes: receiving electrocardiogram data associated with a patient and an electrocardiogram configuration including a plurality of leads and a time interval, the electrocardiogram data comprising, for each lead included in the plurality of leads, voltage data associated with at least a portion of the time interval; receiving an age value associated with the patient; receiving a sex value associated with the patient; providing the age value, the sex value, and at least a portion of the electrocardiogram data to a trained model, the trained model being trained to generate a risk score based on input electrocardiogram data associated with the electrocardiogram configuration and supplementary information associated with the patient; receiving a risk score indicative of a likelihood the patient will suffer from aortic stenosis within a predetermined period of time from when the electrocardiogram data was generated; and outputting the risk score to at least one of a memory or a display for viewing by a medical practitioner or healthcare administrator.” Zimmerman [0192] “At 1316, the process can output the risk score to at least one of a memory (e.g., the memory 220 and/or the memory 240) or a display (e.g., the display 116, the display 208, and/or the display 228). In some embodiments, the display can be in view of a medical practitioner or healthcare administrator. In some embodiments, the process 1300 can generate and output a report based on the risk score. In some embodiments, the report can include the raw risk score and/or graphics related to the risk score. In some embodiments, the process 1300 can determine that the risk score is above a predetermined threshold associated with the condition (e.g., risk scores above the threshold can be indicative that the patient will suffer from the conditions within the predetermined time period). The process 1300 can then generate the report based on the determination that the risk score is above a predetermined threshold. In some embodiments, in response to determining that the risk score is above the predetermined threshold, the process 1300 can generate the report to include information (e.g., text) and/or links to sources (e.g., one or more hyperlinks) about treatments for the condition, causes of the condition, and/or other clinical information about the condition. In some embodiments, the process 1300 can generate the report from intermediate results stored in a standardized format, such as a standardized JavaScript Object Notation (JSON) format. The standardized format may also be converted to a different format for presentation to healthcare providers using format conversion software, such as for conversion into a healthcare providers' electronic health record system. In some embodiments, the process 1300 can generate the report to include name of the test, patient sex, patient date of birth, patient name, institution/physician name, and/or medical record number.”, Zimmerman [0117] “As shown in FIG. 1 , the computing device 104 can receive ECG data, such as 12-lead ECG data, and generate an AF, AS, CA, and/or SP risk score based on the ECG data. In some embodiments, the risk score can indicate a predicted risk of a patient developing the cardiac event within a predetermined time period from when the ECG was taken (e.g., three months, six months, one year, five years, ten years, etc.). In some embodiments, the computing device 104 can execute at least a portion of an ECG analysis application 132 to automatically generate the AF, AS, CA, and/or SP risk score.” and Zimmerman [0036] “The trained model further comprises: training a convolutional neural network on a plurality of patients, wherein the plurality of patients include at least patients having a recorded ECG within a diagnosis threshold and patients having a recorded ECG outside a diagnosis threshold; wherein the diagnosis threshold is compared against the time between the date of diagnosis of aortic stenosis and the date of the recorded ECG; and providing the trained convolutional neural network as the trained model. The trained model further comprises: refining the trained neural network using only the plurality of patients having the recorded ECG outside of the diagnosis threshold, wherein the diagnosis threshold is selected from a number of days.”);
the tracked patient data as additional components of the new multi- modal data are received (Zimmerman [0140] “In some embodiments, the model 400 can additionally receive electronic health record (EHR) data points such as demographic data 416, which can include age and sex/gender as input features to the network, where sex can be encoded into binary values for both male and female, and age can be cast as a continuous numerical value corresponding to the date of acquisition for each 12-lead resting state ECG.”, Zimmerman [0328] “In one example, model inputs may include one or more patient features selected for their importance to the output high-risk label/prediction, including but not limited to demographic features such as age, sex, or other EHR-derived features, in addition to ECG-derived values. For example, LDL diagnostic testing values may account for the highest percentage of the risk assessment and be a required input. In another example, LDL values, ECGs, and age may collectively account for the highest percentage of the risk assessment and be required inputs. In yet another model, inputs may include LDL values, ECGs, age, blood pressure, HTN, AF status, BMI, diabetes, smoking, gender, and/or CHF. In an even more exhaustive implementation, a stroke prediction may include the top X features, where X is an integer. In one example, the integer, X, may be 20 features. Consistent with a 20 feature embodiment, a top 20 features may include: patient age, INDEX_CCI, STROKE_YN, AF_Target, Labs_AlC, Vitals_Weight, Vitals_Height, Demographics_SMOKER_FLG1, Anti_coag, Vitals_BMI, Labs_GLUCOSE, Labs_SODIUM, Vitals_BP_Systolic, Labs_LDL, Medications_ANTICOAGULANTS, Labs_HDL, Labs_HEMOGLOBIN, ECG_R_AXIS, and Echo_measurements_LAV_MOD_sp2. [0329] In some embodiments, all features having a contribution to the high-risk determination with greater than a weight of 1% may be included in the inputs to a composite model. Consistent with a weight inclusion model, features may be selected to include: Vitals_BMI, Vitals_BP_Diastolic, Vitals_BP_Systolic, Vitals_Heart_Rate, Vitals_Height, Vitals_Weight, Demographics_FRS, Demographics_PCE, INDEX_CCI, CHADSVASC_SCORE, CHADS_SCORE, Demographics_PT_AGE, Demographics_PT_RACE, Demographics_PT_SEX, Demographics_SMOKER_FLG, ICD_Phenotypes_AOR, ICD_Phenotypes_AOS, ICD_Phenotypes_IVS, ICD_Phenotypes_LEF, ICD_Phenotypes_MIR, ICD_Phenotypes_MIS, ICD_Phenotypes_PUR, ICD_Phenotypes_PUS, ICD_Phenotypes_TRR, ICD_Phenotypes_TRS, Labs_AlC, Labs_BILI, Labs_BNP, Labs_BUN, Labs_CHOLESTEROL Labs_CKMB, Labs_CREATININE, Labs_CRP, Labs_D_dimer, Labs_eGFR, Labs_GLUCOSE, Labs_HDL, Labs_HEMOGLOBIN, Labs_LDH, Labs_LDL, Labs_LYMPHOCYTES Labs_POTASSIUM Labs_PRO_BNP, Labs_SODIUM, Labs_Triglyceride, Labs_TROPONIN_I, Labs_TROPONIN_T, Labs_URIC_ACID, Labs_VLDL, Medications_ACE_INHIBITORS, Medications_ANGIOTENSIN_II_RECEPTOR_ANTAGONISTS, Medications_ANTICOAGULANTS, Medications_ANTIDIABETIC_MEDICATION, Medications_ANTIHYPERTENSIVE, Medications_DIGOXIN,Medications_ERX_EBBB_HEART_FAILUREMedications_ERX_SPIRONOLACTONE_EPLE R, ONE_HEART_FAILURE, Medications_LOOP_DIURETICS, ECG_Measurements, Echo_Measurements, ECG_Findings, HF_YN, HTN_YN, AGE_GTE_75_YN, DM_YN, STROKE_YN, VASC_DISC_YN, AGE_65_74_YN, and FEMALE_YN.”);
the new multi-modal data for the new patients to predict the patient outcomes for the new patients (Zimmerman [0140] “In some embodiments, the model 400 can additionally receive electronic health record (EHR) data points such as demographic data 416, which can include age and sex/gender as input features to the network, where sex can be encoded into binary values for both male and female, and age can be cast as a continuous numerical value corresponding to the date of acquisition for each 12-lead resting state ECG.”, Zimmerman [0328] “In one example, model inputs may include one or more patient features selected for their importance to the output high-risk label/prediction, including but not limited to demographic features such as age, sex, or other EHR-derived features, in addition to ECG-derived values. For example, LDL diagnostic testing values may account for the highest percentage of the risk assessment and be a required input. In another example, LDL values, ECGs, and age may collectively account for the highest percentage of the risk assessment and be required inputs. In yet another model, inputs may include LDL values, ECGs, age, blood pressure, HTN, AF status, BMI, diabetes, smoking, gender, and/or CHF. In an even more exhaustive implementation, a stroke prediction may include the top X features, where X is an integer. In one example, the integer, X, may be 20 features.”);
repeatedly determining whether the new patients should proceed or not proceed based on the patient outcomes predicted for the new patients (Zimmerman [0111] “Atrial fibrillation (AF) is associated with substantial morbidity, especially when it goes undetected. If new onset AF can be predicted with high accuracy, screening methods could be used to find it early. The present disclosure provides a deep neural network that can predict new onset AF from a resting 12-lead electrocardiogram (ECG). The predicted new onset AF may assist medical practitioners (e.g., a cardiologist) in preventing AF-related adverse outcomes, such as stroke.”, Zimmerman [0328] “In one example, model inputs may include one or more patient features selected for their importance to the output high-risk label/prediction, including but not limited to demographic features such as age, sex, or other EHR-derived features, in addition to ECG-derived values., Zimmerman [0304] “In one example, a composite model may include an ensemble of machine learning and/or deep learning models trained from heterogeneous data including but not limited to ECG, demographics, labs and vitals.” and Zimmerman [0382] “In some embodiments, the method may be diagnostic, whereby the clinical data can include outcome data, such as whether or not a patient developed atrial fibrillation (AF or Afib) in a time period following the day that the ECG was taken. In other embodiments, the clinical data may be used in a predictive sense, e.g., to determine based on that data a likelihood that the patient would develop Afib within a certain time period following the day that the ECG was taken.”); and
notifying a healthcare provider in response to a determination that a patient of the new patients should proceed (Zimmerman [0213] “Outputs of the trained model may include the likelihood of a future adverse outcome (potential outcomes are listed in detail above) and potential interventions that may be performed to reduce the likelihood of the adverse outcome. An exemplary intervention that may be suggested includes notifying the attending physician that if a patient receives a beta blocker medication, their risk of hospitalization may decrease from 10% to 5%.”).
Zimmerman fails to explicitly teach patient outcomes resulting from undergoing a cardiac valve procedure; the cardiac valve procedure; repeatedly determining, over the time, updated severity measures for the new patients based on analysis; repeatedly applying, responsive to the updated severity measures exceeding a threshold severity measure, the trained version of the ensemble-based outcomes forecasting model; and the updated severity measures.
Otto teaches patient outcomes resulting from undergoing a cardiac valve procedure (Otto, page e101, “Figure 3. Choice of SAVR versus TAVI when AVR is indicated for valvular AS. Colors correspond to Table 2. *Approximate ages, based on US Actuarial Life Expectancy tables, are provided for guidance. The balance between expected patient longevity and valve durability varies continuously across the age range, with more durable valves preferred for patients with a longer life expectancy. Bioprosthetic valve durability is finite (with shorter durability for younger patients), whereas mechanical valves are very durable but require lifelong anticoagulation. Long-term (20-y) data on outcomes with surgical bioprosthetic valves are available; robust data on transcatheter bioprosthetic valves extend to only 5 years, leading to uncertainty about longer-term outcomes. The decision about valve type should be individualized on the basis of patient-specific factors that might affect expected longevity. †Placement of a transcatheter valve requires vascular anatomy that allows transfemoral delivery and the absence of aortic root dilation that would require surgical replacement. Valvular anatomy must be suitable for placement of the specific prosthetic valve, including annulus size and shape, leaflet number and calcification, and coronary ostial height. See ACC Expert Consensus Statement.20 AS indicates aortic stenosis; AVR, aortic valve replacement; LVEF, left ventricular ejection fraction; QOL, quality of life; SAVR, surgical aortic valve replacement; STS, Society of Thoracic Surgeons; TAVI, transcatheter aortic valve implantation; TF, transfemoral; and VKA, vitamin K antagonist.”); and
the cardiac valve procedure (Otto, page e101, “Figure 3. Choice of SAVR versus TAVI when AVR is indicated for valvular AS. Colors correspond to Table 2. *Approximate ages, based on US Actuarial Life Expectancy tables, are provided for guidance. The balance between expected patient longevity and valve durability varies continuously across the age range, with more durable valves preferred for patients with a longer life expectancy. Bioprosthetic valve durability is finite (with shorter durability for younger patients), whereas mechanical valves are very durable but require lifelong anticoagulation. Long-term (20-y) data on outcomes with surgical bioprosthetic valves are available; robust data on transcatheter bioprosthetic valves extend to only 5 years, leading to uncertainty about longer-term outcomes. The decision about valve type should be individualized on the basis of patient-specific factors that might affect expected longevity. †Placement of a transcatheter valve requires vascular anatomy that allows transfemoral delivery and the absence of aortic root dilation that would require surgical replacement. Valvular anatomy must be suitable for placement of the specific prosthetic valve, including annulus size and shape, leaflet number and calcification, and coronary ostial height. See ACC Expert Consensus Statement.20 AS indicates aortic stenosis; AVR, aortic valve replacement; LVEF, left ventricular ejection fraction; QOL, quality of life; SAVR, surgical aortic valve replacement; STS, Society of Thoracic Surgeons; TAVI, transcatheter aortic valve implantation; TF, transfemoral; and VKA, vitamin K antagonist.”).
Nemati teaches repeatedly determining, over the time, updated severity measures for the new patients based on analysis (Nemati [0068] “In some example embodiment, the decision analyzer (9) may identify the aforementioned most important features by systematically and iteratively altering the input features (e.g., using gradient descent) to change output of the predictor (5). For example, if the risk score (15) of the patient is 0.4 and the decision threshold applied by the stratification controller (3) is 0.5, a gradient descent approach may be applied in which the input features are altered to increase the risk score (15) above the 0.5 decision threshold. The features are then sorted according to the magnitude of their respective changes from the corresponding baseline values. The decision analyzer (9) (or some other logic unit) may use this information, in addition to information about the age of each feature, to determine which clinical observations are needed to improve model confidence. A set of features that are most likely to change the output of the decision analyzer (9) may be identified based on a ranking of the altered input features. Additional labs (22) providing the corresponding clinical observations may be ordered to reduce the prediction uncertainty of the decision analyzer (9) and, by corollary, reduce the likelihood of a false positive (18) or false negative (19) associated with the risk score (15).”);
repeatedly applying, responsive to the updated severity measures exceeding a threshold severity measure, the trained version of the ensemble-based outcomes forecasting model (Nemati [0006] “In another aspect, there is provided a method for machine learning enabled patient stratification. The method may include: applying a first machine learning model to determine, based at least on a clinical data of a patient, a risk score for the patient; in response to the risk score for the patient exceeding a first threshold, applying a second machine learning model to determine a first probability of the risk score being a false positive; in response to the risk score for the patient failing to exceed the first threshold, applying a third machine learning model to determine a second probability of the risk score being a false negative; and determining, based at least on the risk score, the first probability of the risk score being the false positive, and the second probability of the risk score being the false negative, one or more clinical recommendations for the patient.”); and
the updated severity measures (Nemati [0068] “In some example embodiment, the decision analyzer (9) may identify the aforementioned most important features by systematically and iteratively altering the input features (e.g., using gradient descent) to change output of the predictor (5). For example, if the risk score (15) of the patient is 0.4 and the decision threshold applied by the stratification controller (3) is 0.5, a gradient descent approach may be applied in which the input features are altered to increase the risk score (15) above the 0.5 decision threshold. The features are then sorted according to the magnitude of their respective changes from the corresponding baseline values. The decision analyzer (9) (or some other logic unit) may use this information, in addition to information about the age of each feature, to determine which clinical observations are needed to improve model confidence. A set of features that are most likely to change the output of the decision analyzer (9) may be identified based on a ranking of the altered input features. Additional labs (22) providing the corresponding clinical observations may be ordered to reduce the prediction uncertainty of the decision analyzer (9) and, by corollary, reduce the likelihood of a false positive (18) or false negative (19) associated with the risk score (15).”).
It would have been obvious to a person of ordinary skill in the art (PHOSITA) before the effective filing date of the invention to modify the predictive modeling system of Zimmerman to incorporate the threshold-based model evaluation and iterative risk assessment techniques taught by Nemati and to apply the resulting predictions within the clinical decision framework for cardiac valve procedures described by Otto. Zimmerman teaches machine learning models, including ensemble models trained on patient data such as ECG data, demographic information, laboratory values, and vital signs to predict patient outcomes and clinical risks. Nemati teaches generating a patient risk score using machine learning and conditionally applying additional predictive models when the risk score exceeds a threshold, as well as iteratively evaluating clinical features and obtaining additional patient observations to refine prediction confidence. A PHOSITA would have recognized that incorporating Nemati’s threshold-based evaluation and iterative patient data analysis into Zimmerman’s predictive modeling framework would improve the efficiency of patient risk assessment by enabling conditional model application and updated severity assessments as additional patient data becomes available.
Furthermore, Otto teaches that clinical decisions regarding cardiac valve procedures, such as surgical or transcatheter valve replacement, are made based on patient specific clinical factors and predicted patient outcomes. A PHOSITA would have been motivated to use predictive outcome models, such as those taught by Zimmerman and Nemati, to inform these clinical decision frameworks in order to assist healthcare providers in determining whether a patient should proceed with a cardiac valve procedure. The combination represents the predictable use of known machine learning patient risk prediction and clinical decision support techniques to improve treatment planning and patient outcome assessment, and therefore would have been obvious to a PHOSITA at the time of the invention.
Regarding claim 3, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein the patient outcomes further include at least one of: a change to one or more defined physiological parameters, potential adverse reactions, and a change to a severity measure representative of a measure of severity of a cardiac condition (Zimmerman [0205] “In instances where additional data may inform the model, extraction may include records from electronic health records having additional patient data such as patient status (alive/dead) which may be generated by combining each patient's most recent clinical encounters from the EHR and a regularly-updated death index registry. Patient status is used as an endpoint to determine predictions for 1-year mortality after an ECG, however, additional clinical outcomes may also be predicted, including, but not limited to, mortality at any interval (1, 2, 3 years, etc.); mortality associated with heart disease, cardiovascular disease, sudden cardiac death; hospitalization for cardiovascular disease; need for intensive care unit admission for cardiovascular disease; emergency department visit for cardiovascular disease; new onset of an abnormal heart rhythm such as atrial fibrillation; need for a heart transplant; need for an implantable cardiac device such as a pacemaker or defibrillator; need for mechanical circulatory support such as a left ventricular/right ventricular/biventricular assist device or a total artificial heart; need for a significant cardiac procedure such as percutaneous coronary intervention or coronary artery bypass graft/surgery; new stroke or transient ischemic attack; new acute coronary syndrome; or new onset of any form of cardiovascular disease such as heart failure; or the likelihood of diagnosis from other diseases which may be informed from an ECG.”.).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to modify Zimmerman’s predictive modeling system with Otto’s teachings of cardiac valve procedures so that the predicted outcomes further include physiological parameter changes, adverse reactions, and severity measures, because Zimmerman discloses predicting a range of clinical outcomes (mortality, ICU admission, etc.), and Otto teaches that such outcomes and severity assessments are central to managing cardiac valve patients; therefore a PHOSITA would have been motivated to combine these teachings to improve the clinical relevance of the predictions.
Regarding claim 7, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein the cardiac valve procedure comprises different types of procedures and wherein the operations further comprise further comprise:
determining a preferred procedure type of the different types of procedures for performing on the new patients based on the new multi-modal data, and wherein the patient outcomes predicted by the ensemble-based outcomes forecasting model for are a function of the preferred procedure type (Zimmerman [0187] “In some embodiments the process 1300 may also receive demographic data and/or other patient information associated with the patient. The demographic data can include an age value and a sex value of the patient or additional variables (e.g., race, weight, height, smoking status, etc.) for example from the electronic health record. In some embodiments, the process 1300 can receive one or more EHR data points. In some embodiments, the EHR data points can include laboratory values (blood cholesterol measurements such as LDL/HDL/total cholesterol, blood counts such as hemoglobin/hematocrit/white blood cell count, blood chemistries such as glucose/sodium/potassium/liver and kidney function labs, and additional cardiovascular markers such as troponins and natriuretic peptides), vital signs (blood pressures, heart rate, respiratory rate, oxygen saturation), imaging metrics (such as cardiac ejection fractions, cardiac chamber volumes, heart muscle thickness, heart valve function), patient diagnoses (such as diabetes, chronic kidney disease, congenital heart defects, cancer, etc.), treatments (including procedures, medications, referrals for services such as cardiac rehabilitation, dietary counseling, etc.), echo measurements, ICD codes, and/or care gaps.” Zimmerman [0304] “In one example, a composite model may include an ensemble of machine learning and/or deep learning models trained from heterogeneous data including but not limited to ECG, demographics, labs and vitals.”, Zimmerman [0382] “In some embodiments, the method may be diagnostic, whereby the clinical data can include outcome data, such as whether or not a patient developed atrial fibrillation (AF or Afib) in a time period following the day that the ECG was taken. In other embodiments, the clinical data may be used in a predictive sense, e.g., to determine based on that data a likelihood that the patient would develop Afib within a certain time period following the day that the ECG was taken.” and Otto, e100, “Figure 3. Choice of SAVR versus TAVI when AVR is indicated for valvular AS. Colors correspond to Table 2. *Approximate ages, based on US Actuarial Life Expectancy tables, are provided for guidance. The balance between expected patient longevity and valve durability varies continuously across the age range, with more durable valves preferred for patients with a longer life expectancy. Bioprosthetic valve durability is finite (with shorter durability for younger patients), whereas mechanical valves are very durable but require lifelong anticoagulation. Long-term (20-y) data on outcomes with surgical bioprosthetic valves are available; robust data on transcatheter bioprosthetic valves extend to only 5 years, leading to uncertainty about longer-term outcomes. The decision about valve type should be individualized on the basis of patient-specific factors that might affect expected longevity. †Placement of a transcatheter valve requires vascular anatomy that allows transfemoral delivery and the absence of aortic root dilation that would require surgical replacement. Valvular anatomy must be suitable for placement of the specific prosthetic valve, including annulus size and shape, leaflet number and calcification, and coronary ostial height.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to modify Zimmerman’s predictive model with Otto’s guideline based procedure selection so that the system determines a preferred procedure type and bases outcome predictions on that choice, because Zimmerman applies models to patient-specific features to generate specific predictions, and Otto teaches that patient outcomes vary depending on whether SAVR or TAVR is selected; thus combining these references would yield the predictable result of procedure specific outcome predictions.
Regarding claim 8, Zimmerman, Otto, and Nemati teach the invention in claim 7, as discussed above, and further teach wherein the cardiac value procedure comprises an aortic valve replacement procedure and wherein the different types of procedures comprise a surgical aortic valve replacement (SAVR) procedure type and a transcatheter aortic valve replacement (TAVR) procedure type (Otto, e100, “Figure 3. Choice of SAVR versus TAVI when AVR is indicated for valvular AS. Colors correspond to Table 2. *Approximate ages, based on US Actuarial Life Expectancy tables, are provided for guidance. The balance between expected patient longevity and valve durability varies continuously across the age range, with more durable valves preferred for patients with a longer life expectancy. Bioprosthetic valve durability is finite (with shorter durability for younger patients), whereas mechanical valves are very durable but require lifelong anticoagulation. Long-term (20-y) data on outcomes with surgical bioprosthetic valves are available; robust data on transcatheter bioprosthetic valves extend to only 5 years, leading to uncertainty about longer-term outcomes. The decision about valve type should be individualized on the basis of patient-specific factors that might affect expected longevity. †Placement of a transcatheter valve requires vascular anatomy that allows transfemoral delivery and the absence of aortic root dilation that would require surgical replacement. Valvular anatomy must be suitable for placement of the specific prosthetic valve, including annulus size and shape, leaflet number and calcification, and coronary ostial height.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention t to modify Zimmerman’s model with Otto’s detailed teachings on surgical aortic valve replacement (SAVR) and transcatheter aortic valve replacement (TAVR) so that the system distinguishes between these procedure types, because Zimmerman provides a flexible predictive modeling framework and Otto identifies SAVR and TAVR as the key alternatives in valve replacement; therefore it would have been routine to apply Zimmerman’s modeling to each known procedure type.
Regarding claim 9, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein the parameters selected from the group consisting of: annulus size, distance of coronary artery ostia from aortic valve annulus, severe aortic valve calcification, aneurysm of ascending aorta, position and tortuosity of ascending aorta, bicuspid aortic valve, access route tortuosity, small vessel diameter, left ventricular function, coronary artery disease, and electrocardiogram conduction (Otto, e100, “Figure 3. Choice of SAVR versus TAVI when AVR is indicated for valvular AS. Colors correspond to Table 2. *Approximate ages, based on US Actuarial Life Expectancy tables, are provided for guidance. The balance between expected patient longevity and valve durability varies continuously across the age range, with more durable valves preferred for patients with a longer life expectancy. Bioprosthetic valve durability is finite (with shorter durability for younger patients), whereas mechanical valves are very durable but require lifelong anticoagulation. Long-term (20-y) data on outcomes with surgical bioprosthetic valves are available; robust data on transcatheter bioprosthetic valves extend to only 5 years, leading to uncertainty about longer-term outcomes. The decision about valve type should be individualized on the basis of patient-specific factors that might affect expected longevity. †Placement of a transcatheter valve requires vascular anatomy that allows transfemoral delivery and the absence of aortic root dilation that would require surgical replacement. Valvular anatomy must be suitable for placement of the specific prosthetic valve, including annulus size and shape, leaflet number and calcification, and coronary ostial height.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to to modify Zimmerman’s predictive system with Otto’s discussion of valve anatomy so that the input parameters include annulus size, coronary ostia distance, calcification, aortic aneurysm, bicuspid valve morphology, vascular access characteristics, ventricular function, coronary artery disease, and conduction abnormalities, because Zimmerman teaches incorporating diverse multimodal features, and Otto highlights these anatomical and clinical parameters that help determine procedure planning; therefore a PHOSITA would have been motivated to include them as model inputs.
Regarding claim 13, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein the multi-modal training data comprises electronic medical record data, echocardiogram study findings data, and electrocardiogram study finding data (Zimmerman [0140] “In some embodiments, the model 400 can additionally receive electronic health record (EHR) data points such as demographic data 416, which can include age and sex/gender as input features to the network, where sex can be encoded into binary values for both male and female, and age can be cast as a continuous numerical value corresponding to the date of acquisition for each 12-lead resting state ECG. In some embodiments, other representations may be used, such as an age grouping 0-9 years, 10-19 years, 20-29 years, or other grouping sizes. In some embodiments, other demographic data such as race, smoking status, height, and/or weight may be included. In some embodiments, the EHR data points can include laboratory values, echo measurements, ICD codes, and/or care gaps. The EHR data points (e.g., demographic data, laboratory values, etc.) can be provided to the model 400 at a common location.” and Zimmerman [0142] “In some embodiments, the model 400 can be included in the trained models 136.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention tto modify Zimmerman’s ensemble model, which includes EHR, echo, and ECG data, with Otto’s context of cardiac valve procedures, so that the multimodal training data for valve outcome prediction includes EMR data, echocardiogram findings, and electrocardiogram findings, because both references identify these data sources as essential for assessing cardiac patients and combining them yields predictable improvement in predictive accuracy.
Claims 14 and 20 are analogous to claim 1, thus claims 14 and 20 similarly analyzed and rejected in a manner consistent with the rejection of claim 1.
Claim 15 is analogous to claim 3, thus claim 15 is similarly analyzed and rejected in a manner consistent with the rejection of claim 3.
Claim 18 is analogous to claim 7, thus claim 18 is similarly analyzed and rejected in a manner consistent with the rejection of claim 7.
Claim 19 is analogous to claim 8, thus claim 19 is similarly analyzed and rejected in a manner consistent with the rejection of claim 8.
Regarding claim 21, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein the multi-modal training data comprises echocardiogram (ECHO) data, electrocardiogram (ECG) data, laboratory data, electronic medical record (EMR) data, clinical reports and notes, patient reported symptoms, and medical monitoring device data (Zimmerman [0187] “The ECG data can include a first branch (e.g., “branch 1”) including leads I, II, V1, and V5, acquired from time (t)=0 (start of data acquisition) to t=5 seconds, a second branch (e.g., “branch 2”) including leads V1, V2, V3, II, and V5 from t=5 to t=7.5 seconds, and a third branch (e.g., “branch 3”) including leads V4, V5, V6, II, and V1 from t=7.5 to t=10 seconds as shown in FIG. 3. In some embodiments the process 1300 may also receive demographic data and/or other patient information associated with the patient. The demographic data can include an age value and a sex value of the patient or additional variables (e.g., race, weight, height, smoking status, etc.) for example from the electronic health record. In some embodiments, the process 1300 can receive one or more EHR data points. In some embodiments, the EHR data points can include laboratory values (blood cholesterol measurements such as LDL/HDL/total cholesterol, blood counts such as hemoglobin/hematocrit/white blood cell count, blood chemistries such as glucose/sodium/potassium/liver and kidney function labs, and additional cardiovascular markers such as troponins and natriuretic peptides), vital signs (blood pressures, heart rate, respiratory rate, oxygen saturation), imaging metrics (such as cardiac ejection fractions, cardiac chamber volumes, heart muscle thickness, heart valve function), patient diagnoses (such as diabetes, chronic kidney disease, congenital heart defects, cancer, etc.), treatments (including procedures, medications, referrals for services such as cardiac rehabilitation, dietary counseling, etc.), echo measurements, ICD codes, and/or care gaps.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to modify Zimmerman’s predictive modeling framework with Otto’s teachings on cardiac valve management so that the multimodal training data includes ECHO, ECG, laboratory, EMR, clinical reports and notes, patient-reported symptoms, and monitoring device data, because Zimmerman discloses incorporating these heterogeneous data sources into ensemble models, and Otto confirms their clinical importance in valve patients; therefore, combining the references would have yielded predictable improvements in outcome prediction.
Regarding claim 22, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein the multi-modal training data includes severity measures tracked for the plurality of different patients prior to their reception of the cardiac valve procedure, the severity measures indicating a measure of severity of a cardiac condition of the plurality of different patients (Zimmerman [0245] “Using 2,141,366 ECGs linked to structured echocardiography and electronic health record data from 461,466 adults, a machine learning composite model was trained to predict composite echocardiography-confirmed disease within 1 year. Seven exemplary diseases were included in the composite label: moderate or severe valvular disease (aortic stenosis or regurgitation, mitral stenosis (MS) or regurgitation, tricuspid regurgitation), reduced ejection fraction (EF)<50%, or interventricular septal thickness >15 mm.” and Zimmerman [0252] “A plurality of outcome labels (e.g., 7 outcome labels) using TTE reports, one for each disease outcome of interest (AS, AR, MR, MS, TR, reduced EF, increased IVS thickness). String matching was used on the reports to identify the presence of valvular stenosis or regurgitation, as well as the associated severity level (Table 5). Specifically, Table 5 includes a keyword list for assigning an abnormality and severity to each valve in an Echocardiography report.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to modify Zimmerman’s multimodal training with Otto’s severity based clinical framework so that the training data includes severity measures tracked prior to valve procedures, because Zimmerman discloses labeling outcomes by severity levels in echocardiography reports, and Otto teaches severity assessment as a central component of treatment decision-making; therefore, a PHOSITA would have been motivated to include such severity measures as model inputs.
Regarding claim 23, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein the training comprises extracting input parameters for the different machine learning sub-models from the multi-modal training data, wherein the input parameter comprise: the severity measures, cardiac disease type diagnosis, echocardiogram (ECHO) study findings, electrocardiogram (ECG) study findings, a measure of dyspnea on exertion, a measure of syncope, a measure of exertional angina, a measure of hypertension, a measure of coronary artery disease, a measure of atrial fibrillation, a measure of left ventricle systolic dysfunction, a measure of left ventricle diastolic dysfunction, a measure of mitral valve disease, a measure of pulmonary hypertension, behavioral activities, medications, demographic features, annulus size, distance of coronary artery ostia from aortic valve annulus, a measure of degree of severe aortic valve calcification, a measure of degree of aneurysm ofascending aorta, position and tortuosity of ascending aorta measures, bicuspid aortic valve characteristics, access route characteristics, and measures of left and right ventricular dysfunction (Zimmerman [0187] “In some embodiments, the EHR data points can include laboratory values (blood cholesterol measurements such as LDL/HDL/total cholesterol, blood counts such as hemoglobin/hematocrit/white blood cell count, blood chemistries such as glucose/sodium/potassium/liver and kidney function labs, and additional cardiovascular markers such as troponins and natriuretic peptides), vital signs (blood pressures, heart rate, respiratory rate, oxygen saturation), imaging metrics (such as cardiac ejection fractions, cardiac chamber volumes, heart muscle thickness, heart valve function), patient diagnoses (such as diabetes, chronic kidney disease, congenital heart defects, cancer, etc.), treatments (including procedures, medications, referrals for services such as cardiac rehabilitation, dietary counseling, etc.), echo measurements, ICD codes, and/or care gaps.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to modify Zimmerman’s approach to input feature extraction with Otto’s disclosures on valve disease assessment so that the extracted parameters include severity measures, disease diagnoses, ECHO and ECG findings, symptom measures, comorbidities, anatomical and procedural features, and demographic and behavioral factors, because Zimmerman teaches extracting a range of clinical parameters from EHR, echo, labs, and vitals, and Otto teaches that these specific parameters are determinative in valve procedure outcomes; therefore their combination would have been predictable.
Regarding claim 24, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein the different machine learning sub-models comprise two or more different types of models selected from the group consisting of: a random forest (RF) model type, an extreme gradient boosting model type, a non-linear kernel support vector machine (SVM) model type, and a neural network model type (Zimmerman [0137] “FIG. 4A is an exemplary embodiment of a model 400. Specifically, an architecture of the model 400 is shown. Artificial intelligence models referenced herein, including model 700 and model 724 discussed further below, may be gradient boosting models, random forest models, neural networks (NN), regression models, Naive Bayes models, or machine learning algorithms (MLA).”, Zimmerman [0264] Each CNN layer consisted of 16 kernels of size 5. The same network configuration was used to train one model per clinical outcome, resulting in 7 independently trained CNN models (FIG. 25B). Specifically, FIG. 25B displays a block diagram for a composite model that shows the classification pipeline for ECG trace and other EHR data. The output of each neural network (the triangles in FIG. 25B) applied to ECG trace data is concatenated to labs, vitals, and demographics to form a feature vector. The vector is the input to a classification pipeline (min-max scaling, mean imputation, and XGBoost classifier), which outputs a recommendation score for the patient.”, Zimmerman [0333] “Training may include using 5-fold cross-validation on a plurality of models such as two machine learning models using the baseline ECGs of approximately 92,848 resulting pairs to predict drug-induced LQT (>500 ms) in the on-drug ECGs. Artificial intelligence engines may be implemented, including, by example, a deep neural network using ECG voltage data and a gradient-boosted tree using the baseline QTc with age and sex as additional inputs to both models. Other models may include one or more inputs as described herein. Other combinations of folds, hold-out patients, validations, and number of models for comparison may be considered without departing from the methodology as described herein.” and Zimmerman [0162] “The AUROC and AUPRC of the POC DNN models for the prediction of new onset AF within 1 year in the holdout set (M0) were 0.83, 95% CI [0.83, 0.84] and 0.21 [0.20, 0.22], respectively, for DNN-ECG and 0.85 [0.84, 0.85] and 0.22 [0.21, 0.24], respectively, for DNN-ECG-AS. FIG. 7A is a bar chart of model performance as mean area under the receiver operating characteristic. FIG. 7B is a bar chart of model performance as mean area under the precision-recall curve. The bars represent the mean performance across the 5-fold cross-validation with error bars showing standard deviations. The circle represents the M0 model performance on the holdout set. The three bars represent model performance for (i) Extreme gradient boosting (XGB) model with age and sex as inputs; (ii) DNN model with ECG voltage-time traces as input and (iii) DNN model with ECG voltage-time traces, age and sex as inputs”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to modify Zimmerman’s ensemble predictive model to include sub-models comprising random forest, gradient boosting, SVM, and neural network types, because Zimmerman discloses using CNNs, gradient boosting, RF, and other ML algorithms, and Otto reinforces the need for robust modeling in valve patients; therefore, incorporating these known ML model types into sub-models would have been an obvious choice with predictable results.
Regarding claim 25, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein the multilayer perceptron comprises a two-layer multilayer perceptron (Zimmerman [0263] “In one instance, for the ECG trace models, a low-parameter convolutional neural network (CNN) was developed with 18,495 trainable parameters that consisted of six 1D CNN-Batch Normalization-ReLU (CBR) layer blocks followed by a two-layer multilayer perceptron and a final logistic output layer (Table 7).”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to modify Zimmerman’s model to implement the multilayer perceptron as a two-layer perceptron, because Zimmerman discloses using a two-layer MLP in its CNN-based predictive architecture; therefore, applying that architecture in the context of valve outcomes would have been a routine application of known modeling structures.
Regarding claim 26, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein the collecting comprises using a standardized healthcare information exchange protocol (Zimmerman [0290] “A health care provider may enroll one or more databases of patients into the system. Enrollment may include a backend, frontend, or other EMR integration. The databases may exist within the confinement of the EMR system or may be uploaded to the cloud as part of an information management system. A secure data exchange system may interface and/or liaison information between the databases of patient information and one or more databases of the system provider. Some healthcare providers may desire to keep their databases separate from the system for patient privacy and to protect their proprietary collections of data. Data may be ingested in an unstructured manner for abstraction and curation, in a structured manner, and/or may be normalized between one or more data or structure types. Data may also be securely exchanged between the structured and/or normalized data and one or more of the databases of patient information and one or more databases of the system provider.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to modify Zimmerman’s system to collect data using standardized healthcare information exchange protocols, because Zimmerman teaches EMR integration and secure exchange of structured patient data across systems, and Otto emphasizes the use of standardized patient data for clinical decision-making; therefore, employing standardized protocols would have been a predictable way to achieve secure data flow.
Regarding claim 27, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein repeatedly determining the updated severity measures comprises determining the updated severity measures based on at least one of an amount of change or a rate of change between the previously determined severity measures and subsequently determined severity measures for the respective new patients over time (Nemati [0006] “In another aspect, there is provided a method for machine learning enabled patient stratification. The method may include: applying a first machine learning model to determine, based at least on a clinical data of a patient, a risk score for the patient; in response to the risk score for the patient exceeding a first threshold, applying a second machine learning model to determine a first probability of the risk score being a false positive; in response to the risk score for the patient failing to exceed the first threshold, applying a third machine learning model to determine a second probability of the risk score being a false negative; and determining, based at least on the risk score, the first probability of the risk score being the false positive, and the second probability of the risk score being the false negative, one or more clinical recommendations for the patient.” and Nemati [0068] “In some example embodiment, the decision analyzer (9) may identify the aforementioned most important features by systematically and iteratively altering the input features (e.g., using gradient descent) to change output of the predictor (5). For example, if the risk score (15) of the patient is 0.4 and the decision threshold applied by the stratification controller (3) is 0.5, a gradient descent approach may be applied in which the input features are altered to increase the risk score (15) above the 0.5 decision threshold. The features are then sorted according to the magnitude of their respective changes from the corresponding baseline values. The decision analyzer (9) (or some other logic unit) may use this information, in addition to information about the age of each feature, to determine which clinical observations are needed to improve model confidence. A set of features that are most likely to change the output of the decision analyzer (9) may be identified based on a ranking of the altered input features. Additional labs (22) providing the corresponding clinical observations may be ordered to reduce the prediction uncertainty of the decision analyzer (9) and, by corollary, reduce the likelihood of a false positive (18) or false negative (19) associated with the risk score (15).”).
It would have been obvious to a person of ordinary skill in the art (PHOSITA) before the effective filing date of the invention to determine updated patient severity measures based on an amount of change or a rate of change between previously determined severity measures and subsequently determined severity measures over time. Nemati teaches generating patient risk scores using machine learning models and evaluating the risk score relative to a threshold in order to guide clinical recommendations. Nemati further teaches iteratively analyzing and modifying input features and evaluating the resulting changes in model output to identify features that significantly impact the risk score and improve prediction confidence. A PHOSITA would have recognized that analyzing the magnitude of changes in model outputs and input features over time implicitly involves evaluating changes in patient risk or severity measures. Therefore, it would have been obvious to determine updated severity measures based on the amount or rate of change between previously determined and subsequently determined severity measures in order to better track patient condition progression and improve the accuracy of clinical decision support.
Regarding claim 28, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein repeatedly applying the trained version of the ensemble-based outcomes forecasting model comprises refraining from applying the trained version of the ensemble-based outcomes forecasting model to the new multi-modal data when the updated severity measures do not exceed the threshold severity measure (Nemati [0006] “In another aspect, there is provided a method for machine learning enabled patient stratification. The method may include: applying a first machine learning model to determine, based at least on a clinical data of a patient, a risk score for the patient; in response to the risk score for the patient exceeding a first threshold, applying a second machine learning model to determine a first probability of the risk score being a false positive; in response to the risk score for the patient failing to exceed the first threshold, applying a third machine learning model to determine a second probability of the risk score being a false negative; and determining, based at least on the risk score, the first probability of the risk score being the false positive, and the second probability of the risk score being the false negative, one or more clinical recommendations for the patient.” and Nemati [0068] “In some example embodiment, the decision analyzer (9) may identify the aforementioned most important features by systematically and iteratively altering the input features (e.g., using gradient descent) to change output of the predictor (5). For example, if the risk score (15) of the patient is 0.4 and the decision threshold applied by the stratification controller (3) is 0.5, a gradient descent approach may be applied in which the input features are altered to increase the risk score (15) above the 0.5 decision threshold. The features are then sorted according to the magnitude of their respective changes from the corresponding baseline values. The decision analyzer (9) (or some other logic unit) may use this information, in addition to information about the age of each feature, to determine which clinical observations are needed to improve model confidence. A set of features that are most likely to change the output of the decision analyzer (9) may be identified based on a ranking of the altered input features. Additional labs (22) providing the corresponding clinical observations may be ordered to reduce the prediction uncertainty of the decision analyzer (9) and, by corollary, reduce the likelihood of a false positive (18) or false negative (19) associated with the risk score (15).”). and
Zimmerman [0140] “In some embodiments, the model 400 can additionally receive electronic health record (EHR) data points such as demographic data 416, which can include age and sex/gender as input features to the network, where sex can be encoded into binary values for both male and female, and age can be cast as a continuous numerical value corresponding to the date of acquisition for each 12-lead resting state ECG.”, Zimmerman [0328] “In one example, model inputs may include one or more patient features selected for their importance to the output high-risk label/prediction, including but not limited to demographic features such as age, sex, or other EHR-derived features, in addition to ECG-derived values. For example, LDL diagnostic testing values may account for the highest percentage of the risk assessment and be a required input. In another example, LDL values, ECGs, and age may collectively account for the highest percentage of the risk assessment and be required inputs. In yet another model, inputs may include LDL values, ECGs, age, blood pressure, HTN, AF status, BMI, diabetes, smoking, gender, and/or CHF. In an even more exhaustive implementation, a stroke prediction may include the top X features, where X is an integer. In one example, the integer, X, may be 20 features.”).
It would have been obvious to a person of ordinary skill in the art (PHOSITA) before the effective filing date of the invention to refrain from applying the trained predictive model when a severity or risk measure does not exceed a threshold. Nemati teaches determining a patient risk score using a machine learning model and conditionally applying additional predictive models when the risk score exceeds a threshold value. Nemati further teaches iterative analysis of clinical features to refine the prediction and improve model confidence. A PHOSITA would have understood that threshold model execution implicitly involves applying the predictive model when the threshold condition is satisfied and refraining from applying the model when the threshold condition is not satisfied. Incorporating this conditional model application into the predictive modeling system of Zimmerman would represent a predictable use of known machine learning decision logic to improve efficiency by avoiding unnecessary model execution when patient severity measures do not exceed the threshold.
Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable Zimmerman et al. (U.S. Patent Publication 2023/0245782 A1), referred to hereinafter as Zimmerman, in view of Otto et al. (Otto et al., "2020 ACC/AHA Guideline for the Management of Patients With Valvular Heart Disease: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines", Circulation, Vol. 143, No. 5, 2021, pages e90-e104), referred to hereinafter as Otto, and Nemati et al. (U.S. Patent Publication 2025/0095857 A1), referred to hereinafter as Nemati, and further in view of Andrie et al. (U.S. Publication 2013/0066647 A1) referred to hereinafter as Andrie.
Regarding claim 12, Zimmerman, Otto, and Nemati teach the invention in claim 1, as discussed above, and further teach wherein the operations further comprise, in response to the determination that patient should proceed with the cardiac valve procedure (Zimmerman [0111] “Atrial fibrillation (AF) is associated with substantial morbidity, especially when it goes undetected. If new onset AF can be predicted with high accuracy, screening methods could be used to find it early. The present disclosure provides a deep neural network that can predict new onset AF from a resting 12-lead electrocardiogram (ECG). The predicted new onset AF may assist medical practitioners (e.g., a cardiologist) in preventing AF-related adverse outcomes, such as stroke.”, Zimmerman [0328] “In one example, model inputs may include one or more patient features selected for their importance to the output high-risk label/prediction, including but not limited to demographic features such as age, sex, or other EHR-derived features, in addition to ECG-derived values., Zimmerman [0304] “In one example, a composite model may include an ensemble of machine learning and/or deep learning models trained from heterogeneous data including but not limited to ECG, demographics, labs and vitals.” and Zimmerman [0382] “In some embodiments, the method may be diagnostic, whereby the clinical data can include outcome data, such as whether or not a patient developed atrial fibrillation (AF or Afib) in a time period following the day that the ECG was taken. In other embodiments, the clinical data may be used in a predictive sense, e.g., to determine based on that data a likelihood that the patient would develop Afib within a certain time period following the day that the ECG was taken.” And Otto, page e101, “Figure 3. Choice of SAVR versus TAVI when AVR is indicated for valvular AS. Colors correspond to Table 2. *Approximate ages, based on US Actuarial Life Expectancy tables, are provided for guidance. The balance between expected patient longevity and valve durability varies continuously across the age range, with more durable valves preferred for patients with a longer life expectancy. Bioprosthetic valve durability is finite (with shorter durability for younger patients), whereas mechanical valves are very durable but require lifelong anticoagulation. Long-term (20-y) data on outcomes with surgical bioprosthetic valves are available; robust data on transcatheter bioprosthetic valves extend to only 5 years, leading to uncertainty about longer-term outcomes. The decision about valve type should be individualized on the basis of patient-specific factors that might affect expected longevity. †Placement of a transcatheter valve requires vascular anatomy that allows transfemoral delivery and the absence of aortic root dilation that would require surgical replacement. Valvular anatomy must be suitable for placement of the specific prosthetic valve, including annulus size and shape, leaflet number and calcification, and coronary ostial height. See ACC Expert Consensus Statement.20 AS indicates aortic stenosis; AVR, aortic valve replacement; LVEF, left ventricular ejection fraction; QOL, quality of life; SAVR, surgical aortic valve replacement; STS, Society of Thoracic Surgeons; TAVI, transcatheter aortic valve implantation; TF, transfemoral; and VKA, vitamin K antagonist.”): and
the cardiac valve procedure (Otto, page e101, “Figure 3. Choice of SAVR versus TAVI when AVR is indicated for valvular AS. Colors correspond to Table 2. *Approximate ages, based on US Actuarial Life Expectancy tables, are provided for guidance. The balance between expected patient longevity and valve durability varies continuously across the age range, with more durable valves preferred for patients with a longer life expectancy. Bioprosthetic valve durability is finite (with shorter durability for younger patients), whereas mechanical valves are very durable but require lifelong anticoagulation. Long-term (20-y) data on outcomes with surgical bioprosthetic valves are available; robust data on transcatheter bioprosthetic valves extend to only 5 years, leading to uncertainty about longer-term outcomes. The decision about valve type should be individualized on the basis of patient-specific factors that might affect expected longevity. †Placement of a transcatheter valve requires vascular anatomy that allows transfemoral delivery and the absence of aortic root dilation that would require surgical replacement. Valvular anatomy must be suitable for placement of the specific prosthetic valve, including annulus size and shape, leaflet number and calcification, and coronary ostial height. See ACC Expert Consensus Statement.20 AS indicates aortic stenosis; AVR, aortic valve replacement; LVEF, left ventricular ejection fraction; QOL, quality of life; SAVR, surgical aortic valve replacement; STS, Society of Thoracic Surgeons; TAVI, transcatheter aortic valve implantation; TF, transfemoral; and VKA, vitamin K antagonist.”).
Zimmerman, Otto, and Nemati fail to explicitly teach determining one or more medical supplies needed for the proceeding with the procedure for the patient; and ordering the one or more medical supplies using an electronic medical supply ordering system.
Andrie teaches determining one or more medical supplies needed for the proceeding with the procedure for the patient; and ordering the one or more medical supplies using an electronic medical supply ordering system (Andrie [0016] “In one embodiment, the inventory module can maintain the second database in real-time. The inventory module can be configured to display a list of items in the inventory and a current location of each item to a user, and/or to display list entries corresponding to items that are reserved for a surgery with a predetermined mark. The inventory module can also receive inventory quantities and locations from an RFID tracking system. In one embodiment, the inventory module orders the medical device required for the surgery from a supplier if the medical device required for the surgery is not available in the inventory. The inventory module can also automatically order replacements for items of the inventory that are consumed during the surgery and/or that are reserved for a surgery.”.).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to modify Zimmerman’s outcome forecasting system and Otto’s clinical guidance on valve procedures to incorporate Andrie’s teachings of electronic medical supply management. A PHOSTIA would have been motivated to integrate automated inventory and supply ordering into a clinical decision support system because once a determination is made that a patient should undergo a procedure, it is routine to ensure that the necessary supplies are identified and available. Andrie teaches systems that determine needed surgical items and order them electronically, which would predictably improve workflow efficiency and reduce delays in performing the cardiac valve procedure. The modification would have amounted to applying a known supply ordering technique to a clinical decision support context to yield predictable operational benefits.
Response to Arguments
Applicant’s arguments and amendments, see Remarks/Amendments submitted on 12/22/2025 with respect to the rejection of the claims have been carefully considered and is addressed below.
Specification
The following title is suggested: AUTOMATED DETECTION AND MANAGEMENT OF VALVULAR HEART DISEASE USING MACHINE LEARNING
Claim Rejections - 35 USC § 101
Applicant states that amended claim 1 does not recite a mental process but instead recites a specific computer-implemented architecture that maintains longitudinal patient data, repeatedly computes severity measures, and conditionally invokes an ensemble-based outcomes forecasting model based on a threshold severity measure. Applicant also states that this architecture improves computer operation by reducing unnecessary execution of computationally expensive machine learning models. The Examiner respectfully disagrees. Claim 1 recites limitations directed to analyzing patient data, evaluating severity measures, predicting patient outcomes, and determining whether a patient should proceed with a cardiac valve procedure. Under the broadest reasonable interpretation, these steps correspond to reviewing patient information, assessing severity, estimating possible outcomes, and making a treatment decision. Such activities constitute mental processes that can be performed in the human mind or with pen and paper and therefore fall within the abstract idea grouping.
Although the claim recites the use of an ensemble-based outcomes forecasting model and a multilayer perceptron, these limitations represent tools used to perform the abstract analysis of patient data and prediction of treatment outcomes. The courts have repeatedly held that collecting information, analyzing it, and reporting the results constitutes an abstract idea even when complex algorithms or computer models are employed (Electric Power Group, LLC v. Alstom S.A., 830 F.3d 1350, 1353–54 (Fed. Cir. 2016)). Accordingly, the recitation of machine learning models does not remove the claim from the mental process category of abstract ideas.
Applicant also states that the claim integrates the alleged abstract idea into a practical application by reciting a stateful system that tracks patient data and selectively invokes the forecasting model when a severity threshold is exceeded. However, the claim does not recite any improvement to computer functionality or to the operation of machine learning models themselves. Instead, the claim specifies that the predictive model is executed when a threshold severity measure is satisfied. The use of such threshold conditions represents decision logic that determines when the abstract analysis is applied and does not constitute a technological improvement to computer operation.
Applicant’s reliance on Enfish, McRO, and CardioNet is not persuasive. In Enfish, the claims were directed to a specific self-referential database structure that improved the way computers stored and retrieved data. In McRO, the claims recited specific rules that improved automated animation techniques. In CardioNet, the claims were directed to specific techniques for detecting cardiac arrhythmias from ECG signals. In contrast, the present claim does not introduce a new data structure, improve signal processing, or improve the functioning of machine learning models. Instead, the claim uses generic computer components to collect and analyze patient data and to generate a treatment recommendation.
Even when considered as an ordered combination, the additional elements do not amount to significantly more than the abstract idea. The claim recites generic computing components performing routine functions such as collecting patient data, storing tracked patient information, executing predictive models, and notifying a healthcare provider. These activities represent well-understood, routine, and conventional functions in healthcare information processing. As noted in Electric Power Group, collecting information, analyzing it, and reporting the results does not add significantly more to an abstract idea.
Therefore, Applicant’s amendments and arguments have been fully considered but are not persuasive. The rejection of claims 1, 3, 7-9, 12-15, and 18-28 under 35 U.S.C. § 101 is therefore maintained.
Claim Rejections - 35 USC § 103
Applicant states that Zimmerman and Otto fail to disclose or suggest maintaining tracked patient data including identifiers, timestamps, and previously determined severity measures, repeatedly determining updated severity measures over time, and repeatedly applying the predictive model responsive to updated severity measures exceeding a threshold. The Examiner respectfully disagrees. Zimmerman teaches receiving patient data including ECG signals and associated clinical information and generating predictive risk scores indicative of a likelihood that a patient will develop a cardiac condition within a specified time period. Zimmerman further teaches that patient data such as ECG measurements are associated with a time of acquisition and that resulting predictions may be output to memory or reports. Under the broadest reasonable interpretation, storing patient data and resulting risk scores in memory necessarily results in maintained patient records that include patient identifiers, temporal associations corresponding to the acquisition of clinical measurements, and previously determined risk scores representing measures of disease severity. Additionally, Zimmerman teaches generating predictive risk scores from multimodal clinical inputs obtained over time, and a person of ordinary skill in the art would have understood that newly acquired patient data would result in updated risk predictions when the model is applied to the updated data. Therefore, Zimmerman reasonably suggests maintaining tracked patient data and determining updated severity measures as additional patient data becomes available.
Applicant’s states that the references fail to disclose repeatedly applying a predictive model responsive to a severity threshold is also unpersuasive. Nemati teaches determining a patient risk score using machine learning and conditionally applying additional predictive models when the risk score exceeds a threshold, thereby demonstrating threshold-based execution of predictive models, and further teaches iterative analysis of patient features and refinement of predictions to improve model confidence as additional clinical observations are obtained. Otto teaches evaluating patient severity and clinical factors when determining whether a patient should undergo a cardiac valve procedure. A person of ordinary skill in the art would have found it obvious to incorporate Nemati’s threshold-based predictive model execution and iterative risk analysis into the predictive modeling framework of Zimmerman and to apply such predictions within the clinical decision framework described by Otto in order to assist healthcare providers in determining whether a patient should proceed with a cardiac valve procedure. Accordingly, the combination of Zimmerman, Nemati, and Otto suggests the claimed features, and the rejection under 35 U.S.C. §103 is therefore maintained.
Conclusion
The prior art made of record and not relied upon is considered pertinent to Applicant's disclosure.
Rapaka et al. (European Publication EP3404666A2) teaches a machine learning method that rapidly accesses emergency patient data to assist with diagnosis and triage, leveraging large datasets to improve prediction accuracy, and determine additional patient data is most important to obtain.
Dasi et al. (U.S. Patent Publication 2022/0392642 A1) teaches a computer-implemented method that retrieves a patient’s EMR, maps diagnosis to a treatment database to identify and rank likely medical interventions using predictive simulations, and provides the physician or EMR system with a ranked list of recommended options and associated outcome metrics.
Kilic et al. (Performance of a Machine Learning Algorithm in Predicting Outcomes of Aortic Valve Replacement, Ann Thorac Surg 2021;111:503-10, pgs. 1-8) teaches a study evaluated the performance of a machine learning (ML) algorithm in predicting outcomes of surgical aortic valve replacement (SAVR).
Hernandez-Suarez et al. (Machine Learning Prediction Models for In-Hospital Mortality After, Transcatheter Aortic Valve Replacement, JACC: Cardiovascular Interventions, 2019, Vol. 12, No. 14, pgs. 1-11) teaches a study sought to develop and compare an array of machine learning methods to predict in-hospital mortality after transcatheter aortic valve replacement (TAVR) in the United States.
Benedetto et al. (Machine learning improves mortality risk prediction after cardiac surgery: Systematic review and meta-analysis, The Journal of Thoracic and Cardiovascular Surgery, 2020, Volume 163, Number 6, pgs. 1-22) teaches systematic review and meta-analysis of studies comparing the discrimination accuracy between ML models versus LR in predicting operative mortality following cardiac surgery.
Amal et al. (Use of Multi-Modal Data and Machine Learning to Improve Cardiovascular Disease Care, Frontiers in Cardiovascular Medicine, 2022, Vol. 9, pgs. 1-11) teaches the state-of-the-art research that focuses on how the latest techniques in data fusion are providing scientific and clinical insights specific to the field of cardiovascular medicine. With these new data fusion capabilities, clinicians and researchers alike will advance the diagnosis and treatment of cardiovascular diseases (CVD) to deliver more timely, accurate, and precise patient care.
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 KYRA R LAGOY whose telephone number is (703)756-1773. The examiner can normally be reached Monday - Friday, 8:00 am - 5:00 pm EST.
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, Kambiz Abdi can be reached at (571)272-6702. 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.
/K.R.L./Examiner, Art Unit 3685
/KAMBIZ ABDI/Supervisory Patent Examiner, Art Unit 3685