Patient Speech Health Assessment with Expert Embeddings

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current methods for detecting health conditions in individuals are limited, leading to undiagnosed and untreated conditions due to the impracticality of clinician-driven data analysis, lacking scalability and consistency.

Innovation Solution

A computer-implemented method using natural language processing and machine learning to generate sentence embeddings from patient utterances, comparing them to crowdsourced expert knowledge and clinical questionnaires for objective health assessment.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If clinician-driven data analysis is used for health condition detection, then diagnostic accuracy may be improved, but scalability and consistency deteriorate

Engineering Contradiction:

Improvediagnostic accuracyVSAvoidscalability

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The system enables patients to conduct self-assessment of health conditions through automated analysis of their spoken responses. The natural language processing system processes patient utterances independently without requiring clinician intervention for each assessment, allowing patients to monitor their own health status while maintaining diagnostic accuracy through expert-crowdsourced knowledge bases.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical system of manual clinician data analysis with an automated natural language processing system. The computer-based system transcribes, processes, and analyzes patient speech using machine learning models trained on crowdsourced expert knowledge, substituting human clinician effort with automated computational analysis that maintains consistency and scalability.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If clinician-driven data analysis is used for health condition detection, then diagnostic accuracy may be improved, but consistency deteriorates

Engineering Contradiction:

Improvediagnostic accuracyVSAvoidconsistency

Core Design Contradiction:

Measurement precisionVSStability of the object's composition

Solution Approach 1:

The system transforms the variable human clinician assessment process into a standardized computational process with fixed parameters. The natural language processing system applies consistent algorithms, thresholds, and evaluation criteria to all patient assessments, eliminating variability introduced by different clinicians while maintaining diagnostic accuracy through carefully engineered processing parameters.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The system creates a replicable digital model of expert clinical assessment through crowdsourced knowledge bases. Multiple expert assessments are aggregated into standardized sentence embeddings that serve as consistent reference models, allowing the system to replicate expert-level diagnostic consistency across all patients without requiring actual expert involvement in each case.

Inventive Principle:

Principle #26Copying

3Productivity

If automated natural language processing is used for health assessment, then scalability is improved, but measurement precision may deteriorate

Engineering Contradiction:

ImprovescalabilityVSAvoiddiagnostic accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The system performs preliminary action by pre-processing and structuring expert medical knowledge into standardized sentence embeddings before actual patient assessment. Crowdsourced expert knowledge is aggregated, processed, and stored as reference models in advance, enabling the scalable automated system to maintain measurement precision by comparing patient responses against pre-established expert-derived criteria rather than attempting to replicate complex clinical reasoning in real-time.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12417828B2Expert crowdsourcing for health assessment learning from speech in the digital healthcare era

Publication Date: 2025.09.16 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

Objective health assessment is provided. An utterance of a patient is received in response to a question being presented to the patient. A transcription is generated of the utterance. A set of sentence embeddings is generated from the transcription of the utterance. A plurality of sentence embeddings corresponding to characteristics of a health condition is retrieved. Similarity is measured between the set of sentence embeddings generated from the transcription of the utterance and the plurality of sentence embeddings corresponding to the characteristics of the health condition. A result of a health assessment of the patient is sent to a healthcare professional based on the similarity between the set of sentence embeddings generated from the transcription of the utterance and the plurality of sentence embeddings corresponding to the characteristics of the health condition.