Granted Patent B2

US 12,417,828 · App. 18/325,244 · Granted Sep 16, 2025

Inventors: Raquel Norel (New York, NY); Guillermo Cecchi (New York, NY)

Assignee: International Business Machines Corporation

G16H15/00G10L15/26G16H10/20

Abstract

Claims (95)

1. A computer-implemented method for objective health assessment, the computer-implemented method comprising:

training, by a computer, a machine learning model (MLM) utilizing a plurality of audio recordings received from a plurality of subject matter experts or specially trained actors speaking as a typical patient with a health condition wherein the plurality of audio recordings include a first set of sentences containing a first set of characteristics of the health condition that are pathognomic of the health condition and associates prosody related to the health condition;

retrieving, by the computer, a set of clinical health assessment questionnaires corresponding to the health condition from a set of remote databases;

extracting, by the computer, a second set of sentences containing a second set of characteristics of the health condition from the set of clinical health assessment questionnaires;

generating, by the computer using natural language processing, a plurality of sentence embeddings corresponding to a combined set of characteristics of the health condition based on the first set of sentences and the second set of sentences, wherein the plurality of sentence embeddings are real-valued vectors that encode meaning of the sentences in a vector space;

storing the combined set of characteristics of the health condition and the plurality of sentence embeddings in a storage device;

establishing, by the computer, a connection with a client device of a patient using patient-authorized contact information to perform a health assessment of the patient;

retrieving, by the computer, a set of open-ended questions from the set of clinical health assessment questionnaires corresponding to the health condition;

presenting, by the computer via a display or a speaker of the client device, an open-ended question from the set of open-ended questions to the patient;

receiving, by the computer via the client device, an utterance of the patient in response to the open-ended question being presented to the patient;

generating, by the computer, using natural language understanding a transcription of the utterance of the patient responding to the open-ended question;

generating, by the computer using the trained MLM, a set of sentence embeddings from the transcription of the utterance made by the patient;

retrieving, by the computer, the plurality of sentence embeddings corresponding to the combined set of characteristics of a health condition from the storage device;

measuring, by the computer, similarity between the set of sentence embeddings generated from the transcription of the utterance made by the patient and the plurality of sentence embeddings corresponding to the combined set of characteristics of the health condition stored in the storage device

and their relationship with aspects of clinical symptomatology of the health condition to determine results of a health assessment of the patient; and

sending, by the computer, via a network to a client device of healthcare professional the results of the health assessment of the patient based on the similarity between the set of sentence embeddings generated from the transcription of the utterance made by the patient and the plurality of sentence embeddings corresponding to the characteristics of the health condition.

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

receiving, by the computer, an input to generate the plurality of sentence embeddings corresponding to the combined set of characteristics of the health condition;

sending, by the computer, a request for the first set of characteristics of the health condition to the plurality of subject matter experts who have specialized knowledge regarding the health condition; and

receiving, by the computer, the first set of sentences containing the first set of characteristics of the health condition from the plurality of subject matter experts who have the specialized knowledge regarding the health condition.

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

retrieving, by the computer, a number of clinical health assessment questionnaires corresponding to the health condition, wherein the number is at least two; and

extracting, by the computer, a third set of sentences containing the characteristics of the health condition from the number of clinical health assessment questionnaires corresponding to the health condition.

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

retrieving, by the computer, a set of questions to perform the health assessment of the patient;

selecting, by the computer, the question from the set of questions; and

presenting, by the computer, the question to the patient.

5. The computer-implemented method of claim 1 , wherein the characteristics are distinctive traits, qualities, or attributes that are pathognomonic of the health condition.

6. The computer-implemented method of claim 1 , wherein the computer utilizes a similarity function to measure the similarity between the set of sentence embeddings generated from the transcription of the utterance made by the patient and the plurality of sentence embeddings corresponding to the characteristics of the health condition.

7. The computer-implemented method of claim 6 , wherein the similarity function is cosine similarity.

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

training the MLM to utilize prosody by recognizing and identifying elements of expressiveness selected from a group consisting of intonation, stress, tone, and rhythm of speech.

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

sending the results of the health assessment of the patient to healthcare professional via an approach selected from a group consisting of email, text message, social media post, and page.

10. A computer system for objective health assessment, the computer system comprising:

a communication fabric;

a storage device connected to the communication fabric, wherein the storage device stores program instructions; and

a processor connected to the communication fabric, wherein the processor executes the program instructions to:

train a machine learning model (MLM) utilizing a plurality of audio recordings received from subject matter experts or specially trained actors speaking as a typical patient with a health condition wherein the plurality of audio recordings include a first set of sentences containing a first set of characteristics of the health condition that are pathognomic of the health condition and associates prosody related to the health condition;

retrieve a set of clinical health assessment questionnaires corresponding to the health condition from a set of remote databases;

extract a second set of sentences containing a second set of characteristics of the health condition from the set of clinical health assessment questionnaires;

generate using natural language processing, a plurality of sentence embeddings corresponding to a combined set of characteristics of the health condition based on the first set of sentences and the second set of sentences, wherein the plurality of sentence embeddings are real-valued vectors that encode meaning of the sentences in a vector space;

store the combined set of characteristics of the health condition and the plurality of sentence embeddings in a storage device;

establish a connection with a client device of a patient using patient-authorized contact information to perform a health assessment of the patient;

retrieve a set of open-ended questions from the set of clinical health assessment questionnaires corresponding to the health condition;

present via a display or a speaker of the client device, an open-ended question from the set of open-ended questions to the patient;

receive via the client device an utterance of the patient in response to the open-ended question being presented to the patient;

generate using natural language understanding a transcription of the utterance of the patient responding to the open-ended question;

generate using the trained MLM a set of sentence embeddings from the transcription of the utterance made by the patient;

retrieve the plurality of sentence embeddings corresponding to the combined set of characteristics of the health condition from the storage device;

measure similarity between the set of sentence embeddings generated from the transcription of the utterance made by the patient and the plurality of sentence embeddings corresponding to the combined set of characteristics of the health condition stored in the storage device

and their relationship with aspects of clinical symptomatology of the health condition to determine results of a health assessment of the patient; and

send via a network to a client device of healthcare professional the results of the health assessment of the patient based on the similarity between the set of sentence embeddings generated from the transcription of the utterance made by the patient and the plurality of sentence embeddings corresponding to the characteristics of the health condition.

11. The computer system of claim 10 , wherein the processor further executes the program instructions to:

receive an input to generate the plurality of sentence embeddings corresponding to the combined set of characteristics of the health condition;

send a request for the first set of characteristics of the health condition to the plurality of subject matter experts who have specialized knowledge regarding the health condition; and

receive the first set of sentences containing the first set of characteristics of the health condition from the plurality of subject matter experts who have the specialized knowledge regarding the health condition.

12. The computer system of claim 10 , wherein the processor further executes the program instructions to:

retrieve a number of clinical health assessment questionnaires corresponding to the health condition, wherein the number is at least two; and

extract a third set of sentences containing the characteristics of the health condition from the number of clinical health assessment questionnaires corresponding to the health condition.

13. The computer system of claim 10 , wherein the processor further executes the program instructions to:

train the MLM to utilize prosody by recognizing and identifying elements of expressiveness selected from a group consisting of intonation, stress, tone, and rhythm of speech.

14. The computer system of claim 10 , wherein the processor further executes the program instructions to:

send the results of the health assessment of the patient to healthcare professional via an approach selected from a group consisting of email, text message, social media post, and page.

15. A computer program product for objective health assessment, the computer program product comprising a non-transitory computer-readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to:

train, a machine learning model (MLM) utilizing a plurality of audio recordings received from subject matter experts or specially trained actors speaking as a typical patient with a health condition wherein the plurality of audio recordings include a first set of sentences containing a first set of characteristics of the health condition that are pathognomic of the health condition and associates prosody related to the health condition;

retrieve a set of clinical health assessment questionnaires corresponding to the health condition from a set of remote databases;

extract a second set of sentences containing a second set of characteristics of the health condition from the set of clinical health assessment questionnaires;

generate using natural language processing, a plurality of sentence embeddings corresponding to a combined set of characteristics of the health condition based on the first set of sentences and the second set of sentences, wherein the plurality of sentence embeddings are real-valued vectors that encode meaning of the sentences in a vector space;

store the combined set of characteristics of the health condition and the plurality of sentence embeddings in a storage device;

establish a connection with a client device of a patient using patient-authorized contact information to perform a health assessment of the patient;

retrieve a set of open-ended questions from the set of clinical health assessment questionnaires corresponding to the health condition;

present via a display or a speaker of the client device, an open-ended question from the set of open-ended questions to the patient;

receive via the client device an utterance of the patient in response to the open-ended question being presented to the patient;

generate using natural language understanding a transcription of the utterance of the patient responding to the open-ended question;

generate using the trained MLM a set of sentence embeddings from the transcription of the utterance made by the patient;

retrieve the plurality of sentence embeddings corresponding to the combined set of characteristics of a health condition from the storage device;

measure similarity between the set of sentence embeddings generated from the transcription of the utterance made by the patient and the plurality of sentence embeddings corresponding to the combined set of characteristics of the health condition stored in the storage device

and their relationship with aspects of clinical symptomatology of the health condition to determine results of a health assessment of the patient; and

send via a network to a client device of healthcare professional the results of the health assessment of the patient based on the similarity between the set of sentence embeddings generated from the transcription of the utterance made by the patient and the plurality of sentence embeddings corresponding to the characteristics of the health condition.

16. The computer program product of claim 15 , wherein the program instructions further cause the computer to:

receive an input to generate the plurality of sentence embeddings corresponding to the combined set of characteristics of the health condition;

send a request for the first set of characteristics of the health condition to the plurality of subject matter experts who have specialized knowledge regarding the health condition; and

receive the first set of sentences containing the first set of characteristics of the health condition from the plurality of subject matter experts who have the specialized knowledge regarding the health condition.

17. The computer program product of claim 15 , wherein the program instructions further cause the computer to:

retrieve a number of clinical health assessment questionnaires corresponding to the health condition, wherein the number is at least two; and

extract a third set of sentences containing the characteristics of the health condition from the number of clinical health assessment questionnaires corresponding to the health condition.

18. The computer program product of claim 15 , wherein the program instructions further cause the computer to:

retrieve a set of questions to perform the health assessment of the patient;

select the question from the set of questions; and

present the question to the patient.

19. The computer program product of claim 15 , wherein the program instructions further cause the computer to:

train the MLM to utilize prosody by recognizing and identifying elements of expressiveness selected from a group consisting of intonation, stress, tone, and rhythm of speech.

20. The computer program product of claim 15 , wherein the program instructions further cause the computer to:

send the results of the health assessment of the patient to healthcare professional via an approach selected from a group consisting of email, text message, social media post, and page.

Assignments (1)

Continuity (1)

References Cited (24)