Virtually monitoring blood pressure levels in a patient using machine learning and digital twin technology

A non-transitory computer readable medium storing instructions for predicting a blood pressure level of a patient encoded thereon that, when executed by a processor, cause the processor to:

• access, by a health management platform from a data store, a plurality of biosignals recorded for the patient, the plurality of biosignals comprising one or more of:
• 1) sensor data collected during a time period by one or more wearable sensors worn by the patient, and 2) lab test data collected during the time period;
• encode the 1) sensor data and 2) lab test data into a vector representation, the vector representation comprising static features and sequential features, wherein a static feature represents biological data collected at a single point in time, a sequential feature represents biological data collected at multiple points during the time period and includes a time dependent variable, and the encoding comprises:
• dividing the sequential feature based on a set of time intervals, and
• for each sequential feature, labelling the sequential feature to define (1) a time interval from the set of time intervals and (2) a diastolic blood pressure measurement in the time interval, and (3) a systolic blood pressure measurement in the time interval;
• apply a long-term prediction model to the vector representation to generate a prediction of the blood pressure of the patient over the time period, the long term prediction model generating the prediction by:
• determining a sequential blood pressure estimation by passing sequential input data through a series of recurrent neural network layers, the recurrent neural network layers:
• determining a long-term diastolic blood pressure prediction for the patient during the time period based on the vector representation;
• and
• determining a long-term systolic blood pressure prediction for the patient during the time period based on the vector representation;
• determining a static blood pressure estimation by passing static input data through a stack of dense neural network layers, the dense neural network layers:
• determining a static blood pressure prediction the patient during the time period based on the vector representation;
• and
• concatenating the static blood pressure estimation and the sequential blood pressure estimation in neural network layers to determine the prediction of the blood pressure of the patient over the time period;
• wherein the long-term prediction model is trained based on a training dataset of true blood pressure measurements recorded for the patient during an initialization period and historical biosignals recorded for the patient that contributed to each true blood pressure measurement;
• and
• generate, for display on a mobile device, a notification describing the long-term diastolic blood pressure prediction and the long-term systolic blood pressure prediction determined for the patient.

A system comprising:

• one or more processors;
• and
• a non-transitory computer readable medium storing instructions for predicting a blood pressure level of a patient encoded thereon that, when executed by a processor, cause the processor to:
• access, by a health management platform from a data store, a plurality of biosignals recorded for the patient, the plurality of biosignals comprising one or more of:
• 1) sensor data collected during a time period by one or more wearable sensors worn by the patient, and 2) lab test data collected during the time period;
• encode the 1) sensor data and 2) lab test data into a vector representation, the vector representation comprising static features and sequential features, wherein a static feature represents biological data collected at a single point in time, a sequential feature represents biological data collected at multiple points during the time period and includes a time dependent variable, and the encoding comprises:
• dividing the sequential feature based on a set of time intervals, and
• for each sequential feature, labelling the sequential feature to define (1) a time interval from the set of time intervals and (2) a diastolic blood pressure measurement in the time interval, and (3) a systolic blood pressure measurement in the time interval;
• apply a long-term prediction model to the vector representation to generate a prediction of the blood pressure of the patient over the time period, the long term prediction model generating the prediction by:
• determining a sequential blood pressure estimation by passing sequential input data through a series of recurrent neural network layers, the recurrent neural network layers:
• determining a long-term diastolic blood pressure prediction for the patient during the time period based on the vector representation;
• and
• determining a long-term systolic blood pressure prediction for the patient during the time period based on the vector representation;
• determining a static blood pressure estimation by passing static input data through a stack of dense neural network layers, the dense neural network layers:
• determining a static blood pressure prediction the patient during the time period based on the vector representation;
• and
• concatenating the static blood pressure estimation and the sequential blood pressure estimation in neural network layers to determine the prediction of the blood pressure of the patient over the time period;
• wherein the long-term prediction model is trained based on a training dataset of true blood pressure measurements recorded for the patient during an initialization period and historical biosignals recorded for the patient that contributed to each true blood pressure measurement;
• and
• generate, for display on a mobile device, a notification describing the long-term diastolic blood pressure prediction and the long-term systolic blood pressure prediction determined for the patient.