Biomarker identification

1. A computer-implemented method comprising:

during a session with a subject, presenting a first stimulus to the subject;

obtaining subject response data comprising audio data or audio-video data of the subject's response to the first stimulus, over a defined time period;

generating a first subject descriptor data object and a second subject descriptor data object based on the subject response data,

wherein the first subject descriptor data object comprises a first series of values based on a first segmentation of the defined time period,

wherein the second subject descriptor data object comprises a second series of values based on a second segmentation of the defined time period, wherein the second segmentation is different from the first segmentation, and

wherein the first series of values and the second series of values characterize action of the subject in response to the first stimulus;

deriving, from the first subject descriptor data object and the second subject descriptor data object, a plurality of biomarkers characterizing a behavior of the subject in response to the first stimulus, wherein values of the plurality of biomarkers are indicative of a severity of a disease in the subject;

based on the values of the plurality of biomarkers, identifying a presence of the disease in the subject;

in response to the identification of the presence of the disease in the subject, selecting an additional stimulus for presentation to the subject, the additional stimulus tailored to induce a response indicative of the severity of the disease;

presenting the additional stimulus to the subject during the session;

determining a value of an additional biomarker based on a response to the additional stimulus; and

training a machine learning model using, as inputs of the training,

the values of the plurality of biomarkers and the value of the additional biomarker, and

data characterizing a disease status of the subject,

wherein the trained machine learning model is configured to output disease severity levels based on biomarker data;

providing, as input to the trained machine learning model, biomarkers characterizing a response of a second subject to at least one stimulus provided to the second subject; and

executing the trained machine learning model to obtain, as output of the trained machine learning model, a disease severity level for the second subject as a function of the biomarkers provided as input to the trained machine learning model.

2. The computer-implemented method of claim 1 , comprising:

obtaining a selection by the subject of a presented answer choice in response to a directed question,

wherein the disease severity level is based on the plurality of biomarkers, the value of the additional biomarker, and the selection.

3. The computer-implemented method of claim 1 , comprising:

based on the first subject descriptor data object and the second subject descriptor data object, identifying an untruthful response in the response.

4. The computer-implemented method of claim 1 , wherein the response to the additional stimulus is associated with multiple biomarkers derivable from the response, and wherein the method comprises:

based on the identification of the presence of the disease in the subject, selecting the additional biomarker from among the multiple biomarkers as a biomarker that is relevant to the disease; and

determining the value of the additional biomarker without determining values of other biomarkers of the multiple biomarkers.

5. The computer-implemented method of claim 1 , wherein the first stimulus is a positively valenced question, and wherein the additional stimulus is a negatively valenced question.

6. The computer-implemented method of claim 1 , wherein the first stimulus and the additional stimulus are two different types selected from the following stimulus types: a request to make a face having a specified emotion, a request to make a specified sound, and an open-ended question.

7. The computer-implemented method of claim 1 , wherein the first subject descriptor data object characterizes facial movement of the subject, wherein the subject response data comprises the audio-video data, and

wherein generating the first subject descriptor data object comprises:

for each frame of the audio-video data, generating an initial data object comprising an indication of (i) whether an action unit is present and (ii) an expressivity of the action unit; and

combining the initial data objects corresponding to each frame, to obtain the first subject descriptor data object.

8. The computer-implemented method of claim 1 , wherein the first subject descriptor data object characterizes verbal acoustic characteristics of the response, and

wherein the second subject descriptor data object characterizes speech content of words of the response.

9. The computer-implemented method of claim 8 , wherein generating the second subject descriptor data object comprises:

extracting transcribed speech from the subject response data, the transcribed speech comprising transcribed words; and

segmenting the defined time period based on the transcribed words.

10. The method of claim 1 , wherein generating the first subject descriptor data object comprises decomposing audio in the audio data or the audio-video data into constituent frequencies.

11. The computer-implemented method of claim 1 , wherein the subject response data comprises the audio-video data, wherein the audio-video data includes facial movements of the subject,

wherein the first subject descriptor data object comprises a first time series indicating emotional responses of the subject based on the facial movements,

wherein the second subject descriptor data object comprises a second time series indicating facial tremor of the subject based on the facial movements, and

wherein deriving the plurality of biomarkers comprises:

deriving a first biomarker based on the first time series, and

deriving a second biomarker based on the second time series.

12. The computer-implemented method of claim 1 , wherein the plurality of biomarkers comprise biomarkers of two different types selected from the following biomarker types: facial behavior, verbal, and speech.

13. The computer-implemented method of claim 1 , wherein generating the first subject descriptor data object comprises applying a machine learning process to the subject response data.

14. The computer-implemented method of claim 1 ,

wherein the first segmentation comprises a segmentation based on equal-sized sets of a first number of frames of the audio data or the audio-video data, and

wherein the second segmentation comprises a segmentation based on equal-sized sets of a second number of frames of the audio data or the audio-video data, the first number different from the second number.

15. The computer-implemented method of claim 1 , comprising:

sending, to a controller controlling a display unit, instructions to present the additional stimulus to the subject using the display unit.

16. The computer-implemented method of claim 1 , wherein the first stimulus comprises an instruction to the subject to take a medication, and

wherein the subject response data comprises video data showing the subject taking the medication.

17. The computer-implemented method of claim 1 , wherein the first stimulus is presented to the subject by a display of a device,

wherein a processor of the device controls the display, and

wherein a camera of the device records the subject response data.

18. The computer-implemented method of claim 1 , comprising:

receiving stimulus data, the stimulus data comprising

respective types of a plurality of stimuli presented to the subject, and

time portions corresponding to subject response corresponding to each of the plurality of stimuli; and

deriving the plurality of biomarkers based on the respective types of the plurality of stimuli.

19. The computer-implemented method of claim 1 ,

wherein the first segmentation comprises a segmentation based on equal-sized sets of one or more frames of the audio data or the audio-video data, and

wherein the second segmentation comprises a segmentation based on words or phrases of the response.

20. A system comprising:

a display;

a controller, the controller configured to send instructions to the display to present a first stimulus to a subject during a session;

a sensor configured to record subject response data comprising audio data or audio-video data of the subject's response to the first stimulus over a defined time period; and

a computer, the computer configured to perform operations comprising:

obtaining the subject response data,

generating a first subject descriptor data object and a second subject descriptor data object based on the subject response data,

wherein the first subject descriptor data object comprises a first series of values based on a first segmentation of the defined time period,

wherein the second subject descriptor data object comprises a second series of values based on a second segmentation of the defined time period, wherein the second segmentation is different from the first segmentation, and

wherein the first series of values and the second series of values characterize action of the subject in response to the first stimulus;

deriving, from the first subject descriptor data object and the second subject descriptor data object, a plurality of biomarkers characterizing a behavior of the subject in response to the first stimulus, wherein values of the plurality of biomarkers are indicative of a severity of a disease in the subject;

based on the values of the plurality of biomarkers, identifying a presence of the disease in the subject;

in response to the identification of the presence of the disease in the subject, selecting an additional stimulus for presentation to the subject, the additional stimulus tailored to induce a response indicative of the severity of the disease;

presenting the additional stimulus to the subject during the session;

determining a value of an additional biomarker based on a response to the additional stimulus; and

training a machine learning model using, as inputs of the training,

the values of the plurality of biomarkers and the value of the additional biomarker, and

data characterizing a disease status of the subject,

wherein the trained machine learning model is configured to output disease severity levels based on biomarker data;

providing, as input to the trained machine learning model, biomarkers characterizing a response of a second subject to at least one stimulus provided to the second subject; and

executing the trained machine learning model to obtain, as output of the trained machine learning model, a disease severity level for the second subject as a function of the biomarkers provided as input to the trained machine learning model.

21. The system of claim 20 , wherein the operations comprise:

sending an instruction to the controller to have the display present the additional stimulus to the subject.

22. The system of claim 20 , wherein the display, the controller, and the sensor are integrated into a portable device.

23. The system of claim 20 , wherein the operations comprise:

obtaining a selection by the subject of a presented answer choice in response to a directed question,

wherein the disease severity level is based on the plurality of biomarkers, the value of the additional biomarker, and the selection.

24. The system of claim 20 , wherein generating the second subject descriptor data object comprises:

extracting transcribed speech from the subject response data, the transcribed speech comprising transcribed words; and

segmenting the defined time period based on the transcribed words, so that the second subject descriptor data object characterizes speech content of words of the response.

25. The system of claim 20 , wherein the first stimulus and the additional stimulus are two different types selected from the following stimulus types: a request to make a face having a specified emotion, a request to make a specified sound, and an open-ended question.

26. The system of claim 20 , wherein the operations comprise:

receiving, from the controller, stimulus data, the stimulus data comprising

respective types of a plurality of stimuli presented to the subject, and

time portions corresponding to subject response corresponding to each of the plurality of stimuli; and

deriving the plurality of biomarkers based on the respective types of the plurality of stimuli.