Granted Patent B2

US 12,376,790 · App. 18/393,360 · Granted Aug 5, 2025

Inventors: Jahangir Mohammed (Los Gatos, CA); Mulk Abdul Maluk Mohamed (Tiruchirappalli, IN); Terrence Chun Yin Poon (Foster City, CA); Wei Tian (Sunnyvale, CA)

Assignee: Twin Health, Inc.

A61B5/4866A61B5/1118A61B5/14532A61B5/4833A61B5/486A61B5/6801A61B5/6802A61B5/7267A61B5/7275A61B5/742A61B5/7475A61B5/749G06N20/00G16H10/40G16H10/60G16H15/00G16H20/10G16H20/60G16H40/67G16H50/20G16H50/30G16H50/50A61B5/0205

Abstract

Claims (117)

1. A method for generating a personalized metabolic treatment recommendation for a patient, the method comprising:

accessing, from a data store, by a health management platform, a personalized metabolic profile of the patient and a plurality of biosignals recorded for the patient during a current time period;

receiving, by user input to an application on a patient device, patient data recorded during the current time period;

determining a representation of a metabolic state of the patient during the current time period based on the plurality of biosignals;

determining a prediction of the metabolic state of the patient during the current time period based on the patient data;

comparing the representation of the metabolic state and the prediction of the metabolic state;

generating a patient-specific treatment recommendation based on the comparison, wherein generating the patient-specific treatment recommendation comprises:

iteratively applying a model to the representation of the metabolic state of the patient during the current time period to categorize the patient among a plurality of cohorts, wherein each cohort specifies a set of metabolic parameters, and the plurality of cohorts forms a hierarchical structure in which each layer of the cohorts represents a categorization of patients with an added metabolic parameter compared to cohorts of a higher layer;

receiving an output from the model comprising a categorized cohort that the patient belongs to, wherein the set of metabolic parameters of the categorized cohort match with the metabolic state of the patient;

determining a representative metabolic state profile for the categorized cohort based on mean metabolic profiles of patients in the categorized cohort, wherein representative metabolic state profiles for different cohorts are different; and

generating the patient-specific treatment recommendation for the patient based at least in part on the representative metabolic profile of the categorized cohort; and

providing, for display to the patient, the patient-specific treatment recommendation to the patient device.

2. The method of claim 1 , wherein iteratively applying the model comprises:

applying a combination of binary rules to the representation of the metabolic state of the patient during the current time period;

assigning the patient to a candidate cohort comprising a plurality of sub-cohorts based on the combination of binary rules;

applying an additional binary rule to the representation of the metabolic state of the patient during the current time period, wherein applying the additional binary rule comprising selecting the additional binary rule based on the assignment of the candidate cohort; and

assigning the patient to one of the plurality of sub-cohorts of the candidate cohort.

3. The method of claim 1 , wherein generating the patient-specific treatment recommendation further comprises:

continuously monitoring the metabolic state of the patient; and

responsive to determining a change to the metabolic state of the patient, re-assigning the patient to a different cohort.

4. The method of claim 1 , wherein patient data is recorded as a voice recording, the method further comprising:

applying a natural language processing model to transcribe the voice recording;

for a food item in the voice recording that is not recognized by the natural language processing model, generating a notification for display on the patient device, the notification prompting the patient to label the food item via an application on the patient device; and

responsive to receiving the label of the food item, training the natural language processing model to recognize the food item in future voice recordings based on the voice recording and the labeled food item.

5. The method of claim 1 , wherein the patient-specific treatment recommendation comprises one or more of:

a medication regimen with one or more medications and a schedule for taking the one or more medications;

a schedule for consuming one or more food items;

a schedule for consuming one or more micronutrient and biota nutrient supplements; and

a record of one or more lifestyle adjustments to improve the metabolic state of the patient.

6. The method of claim 1 , wherein generating the patient-specific treatment recommendation comprises:

generating a plurality of candidate recommendations for improving the metabolic state of the patient, each candidate recommendation outlining a unique food regimen, medication schedule, and set of lifestyle adjustments;

ranking the plurality of candidate recommendations based on an improvement of each candidate recommendation on the metabolic state of the patient; and

generating the patient-specific treatment recommendation based on a highest ranked candidate recommendation.

7. The method of claim 1 , further comprising:

accessing a history of previous metabolic states for the patient, each of the previous metabolic states determined during a time period preceding the current time period;

identifying changes between each previous metabolic state and the representation of the metabolic state determined at a conclusion of the current time period based on changes in one or more of the plurality of biosignals; and

responsive to identifying changes in metabolic states leading to the representation of the metabolic state determined at the conclusion of the current time period, generating, for display on the patient device, a graphical user interface modeling and tracking the changes in the metabolic states.

8. The method of claim 1 , further comprising:

generating a digital twin of the patient based on the recorded patient data in the current time period, and the plurality of biosignals, wherein the digital twin comprises:

a health dimension determined based on the plurality of biosignals and entries in a timeline of recorded patient data describing nutrition, sleep, and exercise; and

a happiness dimension determined based on entries in the recorded patient data describing taste preferences and lifestyle satisfaction; and

updating the digital twin based on biosignals and patient data recorded after a conclusion of the current time period.

9. The method of claim 1 , wherein generating the patient-specific treatment recommendation comprises:

applying a set of medication rules codified in a binary format to the representative metabolic profile to generate a medication recommendation.

10. A system comprising:

one or more wearable sensors worn by a patient, each of the one or more wearable sensors configured to collected sensor data during a current time period;

an application stored on a patient device that presents metabolic insights generated for the patient; and

a non-transitory computer readable medium storing instructions for generating a personalized metabolic treatment recommendation for a patient encoded thereon that, when executed by a processor, cause the processor to:

access, from a data store, by a health management platform, a personalized metabolic profile of the patient and a plurality of biosignals recorded for the patient during a current time period;

receive, by user input to an application on a patient device, patient data recorded during the current time period;

determine a representation of a metabolic state of the patient during the current time period based on the plurality of biosignals;

determine a prediction of the metabolic state of the patient during the current time period based on the patient data;

compare the representation of the metabolic state and the prediction of the metabolic state;

generate a patient-specific treatment recommendation based on the comparison, wherein generating the patient-specific treatment recommendation comprises:

iteratively applying a model to the representation of the metabolic state of the patient during the current time period to categorize the patient among a plurality of cohorts, wherein each cohort specifies a set of metabolic parameters, and the plurality of cohorts forms a hierarchical structure in which each layer of the cohorts represents a categorization of patients with an added metabolic parameter compared to cohorts of a higher layer;

receiving an output from the model comprising a categorized cohort that the patient belongs to, wherein the set of metabolic parameters of the categorized cohort match with the metabolic state of the patient;

determining a representative metabolic state profile for the categorized cohort based on mean metabolic profiles of patients in the categorized cohort, wherein representative metabolic state profiles for different cohorts are different; and

generating the patient-specific treatment recommendation for the patient based at least in part on the representative metabolic profile of the categorized cohort; and

provide, for display to the patient, the patient-specific treatment recommendation to the patient device.

11. The system of claim 10 , wherein the instructions for iteratively applying the model further cause the processor to:

apply a combination of binary rules to the representation of the metabolic state of the patient during the current time period;

assign the patient to a candidate cohort comprising a plurality of sub-cohorts based on the combination of binary rules;

apply an additional binary rule to the representation of the metabolic state of the patient during the current time period, wherein applying the additional binary rule comprising selecting the additional binary rule based on the assignment of the candidate cohort; and

assign the patient to one of the plurality of sub-cohorts of the candidate cohort.

12. The system of claim 10 , wherein the instructions for generating the patient-specific treatment recommendation further cause the processor to:

continuously monitor the metabolic state of the patient; and

responsive to determining a change to the metabolic state of the patient, re-assigning the patient to a different cohort.

13. The system of claim 10 , wherein the patient data is recorded as a voice recording, the instructions further causing the processor to:

apply a natural language processing model to transcribe the voice recording;

for a food item in the voice recording that is not recognized by the natural language processing model, generate a notification for display on the patient device, the notification prompting the patient to label the food item via an application on the patient device; and

responsive to receiving the label of the food item, train the natural language processing model to recognize the food item in future voice recordings based on the voice recording and the labeled food item.

14. The system of claim 10 , wherein the instructions further cause the processor to:

access a history of previous metabolic states for the patient, each of the previous metabolic states determined during a time period preceding the current time period;

identify changes between each previous metabolic state and the representation of the metabolic state determined at a conclusion of the current time period based on changes in one or more of the plurality of biosignals; and

responsive to identifying changes in metabolic states leading to the representation of the metabolic state determined at the conclusion of the current time period, generate, for display on the patient device, a graphical user interface modeling and tracking the changes in the metabolic states.

15. The system of claim 10 , wherein the instructions further cause the processor to:

generate a digital twin of the patient based on the current time period of recorded patient data, and the plurality of biosignals, wherein the digital twin comprises:

a health dimension determined based on the plurality of biosignals and entries in the recorded patient data describing nutrition, sleep, and exercise; and

a happiness dimension determined based on entries in a timeline of recorded patient data describing taste preferences and lifestyle satisfaction; and

update the digital twin based on biosignals and patient data recorded after a conclusion of the current time period.

16. The system of claim 10 , wherein the instructions to generate the patient-specific treatment recommendation further cause the processor to:

apply a set of medication rules codified in a binary format to the representative metabolic profile to generate a medication recommendation.

17. A non-transitory computer readable storage medium, storing instructions for generating a personalized metabolic treatment recommendation for a patient encoded thereon that, when executed by a processor, cause the processor to:

access, from a data store, by a health management platform, a personalized metabolic profile of the patient and a plurality of biosignals recorded for the patient during a current time period;

receive, by user input to an application on a patient device, patient data recorded during the current time period;

determine a representation of a metabolic state of the patient during the current time period based on the plurality of biosignals;

determine a prediction of the metabolic state of the patient during the current time period based on the patient data;

compare the representation of the metabolic state and the prediction of the metabolic state;

generate a patient-specific treatment recommendation based on the comparison, wherein generating the patient-specific treatment recommendation comprises:

iteratively applying a model to the representation of the metabolic state of the patient during the current time period to categorize the patient among a plurality of cohorts, wherein each cohort specifies a set of metabolic parameters, and the plurality of cohorts forms a hierarchical structure in which each layer of the cohorts represents a categorization of patients with an added metabolic parameter compared to cohorts of a higher layer;

receiving an output from the model comprising a categorized cohort that the patient belongs to, wherein the set of metabolic parameters of the categorized cohort match with the metabolic state of the patient;

determining a representative metabolic state profile for the categorized cohort based on mean metabolic profiles of patients in the categorized cohort, wherein representative metabolic state profiles for different cohorts are different; and

generating the patient-specific treatment recommendation for the patient based at least in part on the representative metabolic profile of the categorized cohort; and

provide, for display to the patient, the patient-specific treatment recommendation to the patient device.

18. The non-transitory computer readable storage medium of claim 17 , wherein the instructions for iteratively applying the model further cause the processor to:

apply a combination of binary rules to the representation of the metabolic state of the patient during the current time period;

assign the patient to a candidate cohort comprising a plurality of sub-cohorts based on the combination of binary rules;

apply an additional binary rule to the representation of the metabolic state of the patient during the current time period, wherein applying the additional binary rule comprising selecting the additional binary rule based on the assignment of the candidate cohort; and

assign the patient to one of the plurality of sub-cohorts of the candidate cohort.

19. The non-transitory computer readable storage medium of claim 17 , wherein the instructions for iteratively applying the model further cause the processor to:

continuously monitor the metabolic state of the patient; and

responsive to determining a change to the metabolic state of the patient, re-assigning the patient to a different cohort.

20. The non-transitory computer readable storage medium of claim 17 , wherein the patient data is recorded as a voice recording, the instructions further causing the processor to:

apply a natural language processing model to transcribe the voice recording;

for a food item in the voice recording that is not recognized by the natural language processing model, generate a notification for display on the patient device, the notification prompting the patient to label the food item via an application on the patient device; and

responsive to receiving the label of the food item, train the natural language processing model to recognize the food item in future voice recordings based on the voice recording and the labeled food item.

21. The non-transitory computer readable storage medium of claim 17 , wherein the instructions further cause the processor to:

access a history of previous metabolic states for the patient, each of the previous metabolic states determined during a time period preceding the current time period;

identify changes between each previous metabolic state and the representation of the metabolic state determined at a conclusion of the current time period based on changes in one or more of the plurality of biosignals; and

responsive to identifying changes in metabolic states leading to the representation of the metabolic state determined at the conclusion of the current time period, generate, for display on the patient device, a graphical user interface modeling and tracking the changes in the metabolic states.

22. The non-transitory computer readable storage medium of claim 17 , wherein the instructions further cause the processor to:

generate a digital twin of the patient based on the current time period of recorded patient data, and the plurality of biosignals, wherein the digital twin comprises:

a health dimension determined based on the plurality of biosignals and entries in the recorded patient data describing nutrition, sleep, and exercise; and

a happiness dimension determined based on entries in a timeline of recorded patient data describing taste preferences and lifestyle satisfaction; and

update the digital twin based on biosignals and patient data recorded after a conclusion of the current time period.

23. The non-transitory computer readable storage medium of claim 17 , wherein the instructions to generate the patient-specific treatment recommendation further cause the processor to:

apply a set of medication rules codified in a binary format to the representative metabolic profile to generate a medication recommendation.

Assignments (1)

Priority Claims (2)

Continuity (4)

References Cited (56)