Predicting stroke symptoms based on integration of video, audio, and blood pressure data

A computer-implemented method for predicting stroke symptoms for a subject, the method comprising:

• detecting physiological data corresponding to the subject, the physiological data comprising medical data captured in near real time;
• generating, based on the physiological data, a classifier configured to generate an indication of the subject experiencing one or more of a plurality of predefined stroke symptoms, wherein the classifier is trained based on trained physiological data comprising data from multiple subjects and augmented versions of the trained physiological data;
• and
• predicting, using the classifier, a likelihood that the subject is experiencing or will imminently experience one or more of the plurality of predefined stroke symptoms;
• and
• in response to determining that the likelihood is above a threshold level associated with one or more of the plurality of predefined stroke symptoms, causing generation of an alert for the subject and providing the alert to a device associated with the subject and an alert for one or more emergency contacts corresponding to the subject.

A system for predicting stroke symptoms for a subject, the system comprising:

• at least one processor;
• and
• memory storing instructions that, when executed by the at least one processor, cause the system to execute instructions comprising:
• receiving audial data representing one or more utterances of the subject and image data representing one or more images of at least a portion of a face or body of the subject;
• detecting physiological data corresponding to the subject, the physiological data comprising gait data or balance data;
• generating, based on outputs from trained machine learning models, a classifier configured to generate an indication of the subject experiencing one or more of a plurality of predefined stroke symptoms, wherein the outputs are generated by machine learning models trained on the detected physiological data and on external data comprising at least one of internet data, clinical data, and medical record data;
• predicting, using the classifier, a likelihood that the subject is experiencing or will imminently experience one or more of the plurality of predefined stroke symptoms;
• and
• in response to determining that the likelihood is above a threshold level associated with one or more of the plurality of predefined stroke symptoms, causing generation of an alert for the subject and providing the alert to a device associated with the subject or providing the alert for one or more emergency contacts corresponding to the subject.

A non-transitory computer-readable storage medium storing instructions that, when executed by at least one processor, cause the at least one processor to predict stroke symptoms for a subject by executing instructions comprising:

• receiving audial data representing one or more utterances of the subject and image data representing one or more images of at least a portion of a face or body of the subject;
• detecting physiological data corresponding to the subject;
• augmenting the detected physiological data to emulate one or more of a plurality of stroke symptoms for the subject;
• generating, based on outputs from trained machine learning models, a classifier configured to generate an indication of the subject experiencing one or more of a plurality of predefined stroke symptoms, wherein the outputs are generated by machine learning models trained on augmented physiological data generated to emulate one or more of the plurality of stroke symptoms and on external data comprising at least one of internet data, clinical data, and medical record data;
• predicting, using the classifier, a likelihood that the subject is experiencing or will imminently experience one or more of the plurality of predefined stroke symptoms;
• and
• in response to determining that the likelihood is above a threshold level associated with one or more of the plurality of predefined stroke symptoms, causing generation of an alert for the subject and providing the alert to a device associated with the subject or providing the alert for one or more emergency contacts corresponding to the subject.