ECG-based cardiovascular disease detection systems and related methods
A method for determining cardiology disease risk from electrocardiogram trace data and clinical data includes receiving electrocardiogram trace data associated with a patient, receiving the patient's clinical data, providing both sets of data to a trained machine learning composite model that is trained to evaluate the data with respect to each disease of a set of cardiology diseases including three or more of cardiac amyloidosis, aortic stenosis, aortic regurgitation, mitral stenosis, mitral regurgitation, tricuspid regurgitation, abnormal reduced ejection fraction, or abnormal interventricular septal thickness, generating, by the model and based on the evaluation, a composite risk score reflecting a likelihood of the patient being diagnosed with one or more of the cardiology diseases within a predetermined period of time from when the electrocardiogram trace data was generated, and outputting the composite risk score to at least one of a memory or a display.
1 . A method for evaluating electronic health data for a subject, the method comprising:
receiving, from a first source comprising at least one electronic health record, electronic health data associated with a subject;
generating at a first layer of a machine learning model, one or more labels, the one or more labels being generated based at least in part on the electronic health data, each label corresponding to a respective disease state;
generating, based at least in part on a first portion of the electronic health data and the one or more labels, indications of risk of two or more cardiac diseases;
identifying, based at least in part on the indications of risk, one or more of a monitoring or management procedure related to the disease state corresponding to the generated one or more labels; and
sending, to a recipient, a report, the report including the one or more of the monitoring or management procedures.
2 . The method of claim 1 , wherein generating the indications of risk includes providing to a second layer of the machine learning model, the first portion of the electronic health data and the one or more labels, the second layer being trained to generate the indications of risk based on electronic health data and associated labels,
wherein the indications of risk are generated at the second layer based on the first portion of the electronic health data and the one or more labels.
3 . The method of claim 1 , further comprising:
generating a notification to a recipient if the indications of risk exceed a first threshold,
wherein the notification includes the one or more of the monitoring or management procedures.
4 . The method of claim 3 , wherein the notification includes a timing interval in which to perform the one or more of the monitoring or management procedures.
5 . The method of claim 1 , wherein one of the one or more monitoring or management procedures is a diagnostic test for at least one cardiac disease of the two or more cardiac diseases.
6 . The method of claim 1 , wherein one of the one or more monitoring or management procedures is one of additional monitoring, a physical examination, an echocardiogram, or an echocardiograph.
7 . The method of claim 1 , further comprising:
providing at least a portion of the electronic health data to the first layer.
8 . The method of claim 1 , wherein the one or more labels include a condition label, the condition label including an abnormality and a severity.
9 . The method of claim 8 , wherein the one or more labels include a condition label corresponding to one or more cardiac valves of the subject.
10 . The method of claim 1 , further comprising:
receiving, from a second source, second health data of the subject, wherein the indications of risk are generated based at least in part on the second health data.
11 . The method of claim 10 , wherein the second source includes a plurality of subject data records, and wherein the second health data comprises electrocardiogram trace data of the subject.
12 . The method of claim 10 , wherein the first source is associated with a first entity, and wherein the second source is associated with a second entity.
13 . The method of claim 2 , further comprising, for each of the one or more monitoring or management procedures, generating, by the second layer, a clinical score associated with the one or more monitoring or management procedures based at least in part on the one or more labels.
14 . The method of claim 2 , further comprising, generating, by the second layer, and based at least in part on the one or more labels and a second portion of the electronic health data, second indications of risk, the second indications of risk being associated with a second one or more monitoring or management procedures for the subject, wherein the second portion of the electronic health data is not included in the first portion of the electronic health data.
15 . The method of claim 1 , wherein the electronic health data includes one or more of vitals data, laboratory data, comorbidities data, or demographic information of the subject.
16 . The method of claim 15 , wherein the vitals data comprises one or more of body mass index, systolic blood pressure, diastolic blood pressure, heart rate, height, weight, or smoking status,
wherein the laboratory data comprises one or more of A1C, bilirubin, blood urea nitrogen, cholesterol, creatine kinase myocardial band, creatinine, C-reactive protein, D-dimer, glucose, high-density lipoprotein, hemoglobin, high-density lipoprotein, lactate dehydrogenase, lymphocytes, potassium, pro B-type natriuretic peptide, sodium, troponin I and T, triglyceride, uric acid, very low-density lipoprotein, or estimated glomerular filtration rate, and
wherein the comorbidities data comprises one or more of heart failure, prior myocardial infarction, diabetes mellitus, chronic obstructive pulmonary disease, renal failure, prior echocardiogram, coronary artery disease, or hypertension.
17 . The method of claim 1 , further comprising determining a positive predictive value of the machine learning model for one or more subject populations, the subject populations being defined by one or more of a race, ethnicity, sex, or age.
18 . The method of claim 1 , further comprising:
training the machine learning model using a first subject population, a majority of the subject in the first subject population having a first demographic characteristic;
evaluating a predictive value of the machine learning model for subjects of a second subject population, a majority of the subjects in the second subject population not having the first demographic characteristic; and
if the predictive value of the machine learning model for the second subject population is below a predictive value threshold, modifying the model to at least partially account for differences between subjects of the first subject population and the second subject population.
19 . A system for evaluating electronic health data for a subject, the system comprising:
a computer including a processing device, the processing device configured to:
receive, from a first source comprising at least one electronic health record, electronic health data associated with a subject;
generate at a first layer of a machine learning model, one or more labels, the one or more labels being generated based at least in part on the electronic health data, each label corresponding to a respective disease state;
generate, based at least in part on a first portion of the electronic health data and the one or more labels, indications of risk of two or more cardiac diseases;
identify, based at least in part on the indications of risk, one or more of a monitoring or management procedure related to the disease state corresponding to the generated one or more labels; and
send, to a recipient, a report, the report including the one or more of the monitoring or management procedures.
20 . A non-transitory computer readable medium, comprising instructions for causing a computer to:
receive, from a first source comprising at least one electronic health record, electronic health data associated with a subject;
generate at a first layer of a machine learning model, one or more labels, the one or more labels being generated based at least in part on the electronic health data, each label corresponding to a respective disease state;
generate, based at least in part on a first portion of the electronic health data and the one or more labels, indications of risk of two or more cardiac diseases;
identify, based at least in part on the indications of risk, one or more of a monitoring or management procedure related to the disease state corresponding to the generated one or more labels; and
send, to a recipient, a report, the report including the one or more of the monitoring or management procedures.