Personalized communication in a digital therapy platform
An example digital therapy platform is disclosed that provides personalized, real-time feedback to patients during sessions. The platform collects real-time performance data during the sessions, including metrics such as range of motion, pelvic floor movements, exercise completion rates, and the accuracy of movements. The digital therapy platform may also retrieve historical data from past sessions to provide a comprehensive overview of the patient's context. The digital therapy platform dynamically generates structured prompts based on the analyzed data and provides these prompts to a large language model (LLM). The LLM generates personalized messages tailored to the patient's specific context. The digital therapy platform delivers the personalized messages to the patient through a user interface.
1. A computer-implemented method for generating personalized messages in a digital therapy platform, the method comprising:
rendering and displaying, at a user device of a patient, a user interface of the digital therapy platform to present instructions for performing an exercise involving one or more body parts;
capturing, by one or more sensors associated with the user device, motion data for the patient during a session of the digital therapy platform while the patient performs the exercise and while maintaining display of the user interface;
processing, by at least one processor, the motion data while the session is in progress to track motion of the one or more body parts and generate real-time performance data for the patient;
displaying, while the session is in progress, at least one of the motion data or the real-time performance data in the user interface;
automatically accessing, by the at least one processor, input data comprising at least one of the real-time performance data or historical data associated with the patient;
automatically and dynamically generating, by the at least one processor, a prompt for a Large Language Model (LLM) based on the input data, the generating of the prompt for the LLM comprising generating a prompt data structure and updating at least part of the prompt data structure in real-time, while the session is in progress, based on at least one of the real-time performance data or the historical data;
automatically generating, using the LLM, a personalized message based on the prompt; and
automatically causing real-time presentation of the personalized message to the patient, the real-time presentation comprising at least one of;
transmitting, by audio hardware of the user device, audio output to deliver the personalized message: or
displaying the personalized message in the user interface.
2. The method of claim 1 , wherein the real-time performance data includes quantitative metrics comprising at least one of range of motion, movement of pelvic floor muscles, exercise completion, or movement accuracy.
3. The method of claim 1 , wherein the historical data comprises at least one of quantitative metrics indicative of historical performance from a previous session, information describing a goal of the patient, or information describing a baseline condition of the patient.
4. The method of claim 1 , wherein dynamically generating the prompt further comprises employing a rules-based engine to prioritize a first type of the input data over a second type of the input data based on quantitative significance.
5. The method of claim 1 , wherein the personalized message includes feedback on progress of the patient relative to a goal.
6. The method of claim 1 , further comprising:
detecting completion of the exercise, wherein the personalized message is generated in real-time in response to detecting the completion of the exercise.
7. The method of claim 1 , further comprising receiving and responding, via the LLM, to a follow-up question from the patient in natural language based on a contextual understanding of the session.
8. The method of claim 1 , wherein the LLM is fine-tuned based on a dataset comprising examples of sessions.
9. The method of claim 1 , wherein the personalized message is adapted based on an analysis of performance trends of the patient observed over multiple sessions.
10. The method of claim 1 , wherein the digital therapy platform is configured to store the personalized message for review by a healthcare provider in a structured database.
11. The method of claim 1 , wherein the digital therapy platform is configured to adjust at least one of difficulty or intensity of subsequent exercises based on an analysis of the personalized message.
12. The method of claim 1 , wherein the personalized message comprises a personalized cue for performing an exercise during the session, the personalized cue being presented to the patient in real-time based on an analysis of the real-time performance data.
13. The method of claim 1 , wherein dynamically generating the prompt includes utilizing one or more machine learning algorithms to determine the prompt data structure based on a combination of the real-time performance data and the historical data.
14. The method of claim 1 , wherein the historical data comprises at least one of patient-specific historical data related to historical performance of the patient or population-based historical data related to historical performance of a group of patients.
15. The method of claim 1 , further comprising:
receiving user input via the user device of the patient; and
in response to receiving the user input, automatically adjusting one or more subsequent exercises of the session.
16. The method of claim 1 , wherein the prompt data structure comprises a base prompt structure, and wherein the dynamic generation of the prompt comprises:
identifying a scenario type from among a plurality of scenario types;
generating the base prompt structure based on the scenario type; and
dynamically integrating at least some of the input data into the base prompt structure.
17. The method of claim 1 , wherein the prompt data structure comprises:
a system prompt that includes an instruction;
a scenario descriptor; and
a user prompt that includes at least some of the input data.
18. The method of claim 1 , wherein the one or more sensors comprise a camera of the user device, the capturing of the motion data comprises capturing, by the camera, images of the one or more body parts while the patient performs the exercise, and the processing of the motion data comprises processing the images in real-time using one or more computer vision algorithms to track the motion of the one or more body parts appearing in the images while the session is in progress.
19. A system comprising:
at least one processor; and
at least one memory storing instructions that, when executed by the at least one processor, configure the system to perform operations comprising:
rendering and displaying, at a user device of a patient, a user interface of a digital therapy platform to present instructions for performing an exercise involving one or more body parts;
capturing, by one or more sensors associated with the user device, motion data for the patient during a session of the digital therapy platform while the patient performs the exercise and while maintaining display of the user interface;
processing the motion data while the session is in progress to track motion of the one or more body parts and generate real-time performance data for the patient;
displaying, while the session is in progress, at least one of the motion data or the real-time performance data in the user interface;
automatically accessing input data comprising at least one of the real-time performance data or historical data associated with the patient;
automatically and dynamically generating a prompt for a Large Language Model (LLM) based on the input data, the generating of the prompt for the LLM comprising generating a prompt data structure and updating at least part of the prompt data structure in real-time, while the session is in progress, based on at least one of the real-time performance data or the historical data;
automatically generating, using the LLM, a personalized message based on the prompt; and
automatically causing real-time presentation of the personalized message to the patient, the real-time presentation comprising at least one of:
transmitting, by audio hardware of the user device, audio output to deliver the personalized message: or
displaying the personalized message in the user interface.
20. A non-transitory machine-readable storage medium, the machine-readable storage medium including instructions that when executed by a computer system, cause the computer system to perform operations comprising:
rendering and displaying, at a user device of a patient, a user interface of a digital therapy platform to present instructions for performing an exercise involving one or more body parts;
capturing, one or more sensors associated with the user device, motion data for the patient during a session of the digital therapy platform while the patient performs the exercise and while maintaining display of the user interface;
processing the motion data while the session is in progress to track motion of the one or more body parts and generate real-time performance data for the patient;
displaying, while the session is in progress, at least one of the motion data or the real-time performance data in the user interface;
automatically accessing input data comprising at least one of the real-time performance data or historical data associated with the patient;
automatically and dynamically generating a prompt for a Large Language Model (LLM) based on the input data, the generating of the prompt for the LLM comprising generating a prompt data structure and updating at least part of the prompt data structure in real-time, while the session is in progress, based on at least one of the real-time performance data or the historical data;
automatically generating, using the LLM, a personalized message based on the prompt; and
automatically causing real-time presentation of the personalized message to the patient, the real-time presentation comprising at least one of:
transmitting, by audio hardware of the user device, audio output to deliver the personalized message: or
displaying the personalized message in the user interface.