SYSTEMS AND METHODS FOR DIAGNOSING A HEALTH CONDITION BASED ON PATIENT TIME SERIES DATA
Disclosed systems, methods, and computer readable media can diagnose a health condition based on patient time series data. For example, a method for diagnosing a health condition based on patient time series data includes identifying a training set of health records comprising a first set of patient time series data, training a neural network using the training set of health records, and executing the trained neural network model to diagnose a health condition based on a second set of patient time series data. In further examples, the first set of patient time series data and the second set of patient time series data can each comprise electrocardiogram data and the health condition can comprise pulmonary hypertension.
1 . A method for diagnosing a health condition based on patient time series data, wherein the method comprises:
receiving, using one or more hardware processors, patient time series data, wherein the patient time series data comprises an electrocardiogram (ECG) waveform from an adult patient at risk of heart failure;
identifying, using the one or more hardware processors, a trained neural network model which has been trained using a training set of health records, wherein the training set of health records comprised:
health records of patients who have been diagnosed with a health condition of interest; and
ECG waveforms of the patients correlated to positive diagnoses of the health condition of interest;
pre-processing, using the one or more hardware processors, the time series data, wherein preprocessing the time series data comprises:
segmenting, using the one or more hardware processors, the patient time series data comprising the ECG waveforms into at least a segment of the ECG waveform over at least a time window; and
inputting, using the one or more hardware processors, the patient time series data comprising the at least a segment of the ECG waveform into the trained neural network model;
executing, using the one or more hardware processors, the trained neural network model; and
predicting, using the one or more hardware processors, whether the adult patient at risk of heart failure is at risk from the health condition of interest as a function of the patient time series data comprising the at least a segment of the ECG waveform and the trained neural network model.
2 . The method of claim 1 , further comprising outputting, using the one or more hardware processors and the patient time series data comprising the ECG waveform, a numerical score representative of risk from the health condition of interest for the adult patient at risk of heart failure.
3 . The method of claim 1 , wherein predicting whether the adult patient at risk of heart failure is at risk from the health condition of interest comprises a binary prediction of either “positive” or “negative”.
4 . The method of claim 1 , wherein identifying the trained neural network model comprises selecting, using the one or more hardware processors, the trained neural network model from a plurality of trained neural network models, as a function of performance of the trained neural network model with a cohort common to the adult patient at risk of heart failure.
5 . The method of claim 4 , further comprising recommending, using the one or more hardware processors, an intervention as a function of the predicted heart condition.
6 . The method of claim 1 , wherein the training set was filtered based upon age of the patients.
7 . The method of claim 1 , wherein at least a portion of the patient health records of the training set were comprehensively assessed by a physician.
8 . The method of claim 1 , wherein pre-processing the time series data further comprises:
segmenting, using the one or more hardware processors, the patient time series data comprising the ECG waveforms into a first segment of the ECG waveform over a first time window;
segmenting, using the one or more hardware processors, the patient time series data comprising the ECG waveforms into a second segment of the ECG waveform over a second time window;
inputting, using the one or more hardware processors, the first segment and the second segment of the ECG waveform into the trained neural network model.
9 . The method of claim 8 , further comprising:
executing, using the one or more hardware processors, the trained neural network model; and
outputting, using the one or more hardware processors and the first segment of the ECG waveform, a first output representative of risk from the health condition of interest for the adult patient at risk of heart failure, from the trained neural network model;
outputting, using the one or more hardware processors and the second segment of the ECG waveform, a first output representative of risk from the health condition of interest for the adult patient at risk of heart failure, from the trained neural network model;
aggregating, using the one or more hardware processors, an aggregated output as a function of the first output and the second output; and
predicting, using the one or more hardware processors, whether the adult patient at risk of heart failure is at risk from the health condition of interest as a function of the aggregated output.
10 . The method of claim 1 , wherein the training set comprised a first set of health records associated with patients diagnosed with the health condition of interest and a second set of health records associated with patients not diagnosed with the health condition of interest.
11 . The method of claim 1 , wherein the ECG waveforms of the patients in the training set comprised diagnostic ECG waveforms that were captured within a predetermined amount of time of a date on which the patients received the positive diagnoses for the health condition of interest.
12 . The method of claim 1 , wherein the ECG waveforms of the patients in the training set comprised preemptive ECG waveforms that were captured at least a predetermined amount of time before a date on which the patients received the positive diagnoses for the health condition of interest.
13 . The method of claim 1 , wherein the ECG waveforms of the patients in the training set were captured while the patients were not challenged by exercise.
14 . A system for diagnosing a health condition based on patient time series data, wherein the system comprises:
a non-transitory memory; and
one or more hardware processors configured to read instructions from the non-transitory memory that, when executed, cause the one or more processors to perform operations comprising:
receive patient time series data, wherein the patient time series data comprises an electrocardiogram (ECG) waveform from an adult patient at risk of heart failure;
identify a trained neural network model which has been trained using a training set of health records, wherein the training set of health records comprised:
health records of patients who have been diagnosed with a health condition of interest; and
ECG waveforms of the patients correlated to positive diagnoses of the health condition of interest;
pre-process the time series data, wherein preprocessing the time series data comprises:
segmenting, using the one or more hardware processors, the patient time series data comprising the ECG waveforms into at least a segment of the ECG waveform over at least a time window; and
input the patient time series data comprising the at least a segment of the ECG waveform into the trained neural network model;
execute the trained neural network model; and
predict whether the adult patient at risk of heart failure is at risk from the health condition of interest as a function of the patient time series data comprising the at least a segment of the ECG waveform and the trained neural network model.
15 . The system of claim 14 , wherein the instructions, when executed, cause the one or more hardware processors to perform additional operations comprising output, using the patient time series data comprising the ECG waveform, a numerical score representative of risk from the health condition of interest for the adult patient at risk of heart failure.
16 . The system of claim 14 , wherein predicting whether the adult patient at risk of heart failure is at risk from the health condition of interest comprises a binary prediction of either “positive” or “negative”.
17 . The system of claim 14 , wherein identifying the trained neural network model comprises selecting, using the one or more hardware processors, the trained neural network model from a plurality of trained neural network models, as a function of performance of the trained neural network model with a cohort common to the adult patient at risk of heart failure.
18 . The system of claim 14 , wherein the instructions, when executed, cause the one or more hardware processors to perform additional operations comprising recommend an intervention as a function of the predicted heart condition.
19 . The system of claim 14 , wherein the training set was filtered based upon age of the patients.
20 . The system of claim 14 , wherein at least a portion of the patient health records of the training set were comprehensively assessed by a physician.
21 . The system of claim 14 , wherein pre-processing the time series data further comprises:
segmenting, using the one or more hardware processors, the patient time series data comprising the ECG waveforms into a first segment of the ECG waveform over a first time window;
segmenting, using the one or more hardware processors, the patient time series data comprising the ECG waveforms into a second segment of the ECG waveform over a second time window; and
inputting, using the one or more hardware processors, the first segment and the second segment of the ECG waveform into the trained neural network model.
22 . The system of claim 21 , wherein the instructions, when executed, cause the one or more hardware processors to perform additional operations comprising:
execute the trained neural network model; and
output, using the first segment of the ECG waveform, a first output representative of risk from the health condition of interest for the adult patient at risk of heart failure, from the trained neural network model;
output, using the second segment of the ECG waveform, a first output representative of risk from the health condition of interest for the adult patient at risk of heart failure, from the trained neural network model;
aggregate an aggregated output as a function of the first output and the second output; and
predict whether the adult patient at risk of heart failure is at risk from the health condition of interest as a function of the aggregated output.
23 . The system of claim 14 , wherein the training set comprised a first set of health records associated with patients diagnosed with the health condition of interest and a second set of health records associated with patients not diagnosed with the health condition of interest.
24 . The system of claim 14 , wherein the ECG waveforms of the patients in the training set comprised diagnostic ECG waveforms that were captured within a predetermined amount of time of a date on which the patients received the positive diagnoses for the health condition of interest.
25 . The system of claim 14 , wherein the ECG waveforms of the patients in the training set comprised preemptive ECG waveforms that were captured at least a predetermined amount of time before a date on which the patients received the positive diagnoses for the health condition of interest.
26 . The system of claim 14 , wherein the ECG waveforms of the patients in the training set were captured while the patients were not challenged by exercise.
27 . A system for diagnosing a health condition based on patient time series data, wherein the system comprises:
a non-transitory memory; and
one or more hardware processors configured to read instructions from the non-transitory memory that, when executed, cause the one or more processors to perform operations comprising:
receive patient time series data, wherein the patient time series data comprises an electrocardiogram (ECG) waveform from an adult patient at risk of heart failure;
identify a trained neural network model which has been trained using a training set of health records, wherein the training set of health records comprised:
health records of patients who have been diagnosed with a health condition of interest;
ECG waveforms of the patients correlated to positive diagnoses of the health condition of interest;
the training set was filtered based upon age of the patients;
at least a portion of the patient health records of the training set were comprehensively assessed by a physician;
the training set comprised a first set of health records associated with patients diagnosed with the health condition of interest and a second set of health records associated with patients not diagnosed with the health condition of interest; and
the ECG waveforms of the patients in the training set were captured while the patients were not challenged by exercise;
pre-process the time series data, wherein preprocessing the time series data comprises:
segmenting, using the one or more hardware processors, the patient time series data comprising the ECG waveforms into at least a segment of the ECG waveform over at least a time window; and
input the patient time series data comprising the at least a segment of the ECG waveform into the trained neural network model;
execute the trained neural network model; and
predict whether the adult patient at risk of heart failure is at risk from the health condition of interest as a function of the patient time series data comprising the at least a segment of the ECG waveform and the trained neural network model wherein predicting whether the adult patient at risk of heart failure is at risk from the health condition of interest comprises a binary prediction of either “positive” or “negative”; and
recommend an intervention as a function of the predicted heart condition.