Methods and systems for engineering visual features from biophysical signals for use in characterizing physiological systems

A system comprising:

• a processor;
• and
• a memory having instructions stored thereon, wherein execution of the instructions by the processor cause the processor to:
• simultaneously acquire, via one or more surface sensors placed on a subject while the subject is at rest, a biophysical signal data set of a subject including at least one cardiac signal and photoplethysmogram (PPG) signal;
• generate a phase space model of the biophysical signal data set;
• generate an orthogonal projection of the phase space model of the biophysical signal data set, wherein the orthogonal projection defines a plurality of orthogonal planes;
• determine based at least in part on the orthogonal projection of the phase space model, values of a plurality of visual-predictor associated properties of a plurality of waveform regions of interest in the biophysical signal data set, wherein the visual-predictor associated properties characterize properties of repetitive cycles between the at least one cardiac signal and PPG signal, and wherein the plurality of waveform regions of interest comprises an atrial repolarization (AR) waveform region, a ventricular repolarization (VR) waveform region, and a ventricular depolarization (VD) waveform region;
• determine, by assessing morphologies and variability associated with the plurality of waveform regions of interest, an estimated value for presence of a metric associated with a cardiac disease state or abnormal condition based, in part, on an application of the determined values of the one or more visual-predictor associated properties to a trained visual-based machine learning model;
• and
• generate or modify a treatment parameter for the subject in response to determining based on the estimated value that the subject has the cardiac disease state or abnormal condition, wherein the subject is treated with surgery or medication based, at least in part, on the treatment parameter.

A non-transitory computer readable medium having instructions stored thereon, wherein execution of the instructions by a processor causes the processor to:

• simultaneously acquire, via one or more surface sensors placed on a subject while the subject is at rest, a biophysical signal data set of a subject including at least one cardiac signal and photoplethysmogram (PPG) signal;
• generate a phase space model of the biophysical signal data set;
• generate an orthogonal projection of the phase space model of the biophysical signal data set, wherein the orthogonal projection defines a plurality of orthogonal planes;
• determine based at least in part on the orthogonal projection of the phase space model, values of a plurality of visual-predictor associated properties of a plurality of waveform regions of interest in the biophysical signal data set, wherein the visual-predictor associated properties characterize properties of repetitive cycles between the at least one cardiac signal and PPG signal, and wherein the plurality of waveform regions of interest comprises an atrial repolarization (AR) waveform region, a ventricular repolarization (VR) waveform region, and a ventricular depolarization (VD) waveform region;
• determine, by assessing morphologies and variability associated with the plurality of waveform regions of interest, an estimated value for presence of a metric associated with a cardiac disease state or abnormal condition based, in part, on an application of the determined values of the one or more visual-predictor associated properties to a trained visual-based machine learning model;
• and
• generate or modify a treatment parameter for the subject in response to determining based on the estimated value that the subject has the cardiac disease state or abnormal condition, wherein the subject is treated with surgery or medication based, at least in part, on the treatment parameter.