Speech-to-Text Cognitive Impairment Screening With Machine Learning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for detecting cognitive impairment, particularly early signs, are not scalable and lack accuracy, especially when dealing with a growing and aging population with limited health resources.

Innovation Solution

A computer-implemented method using speech-to-text processing and machine learning models to analyze audio data from neuropsychological tests, calculating impairment probabilities based on multiple test variables and personal information, to accurately detect cognitive impairment.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If conventional neuropsychological tests with point-based scoring are used, then clinical practitioners can make diagnoses, but the method is not scalable to large populations with limited health resources

Engineering Contradiction:
Improvescreening capacityVSAvoidclinical consultation requirement
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system enables patients to complete neuropsychological tests independently using a computing device without requiring continuous clinical practitioner involvement. The automated speech-to-text processing and machine learning-based impairment probability calculation allow the system to serve itself by processing test responses and generating diagnostic indications autonomously.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical system of manual clinical assessment with an automated computational system. Speech-to-text engines convert patient utterances into text, machine learning models calculate impairment probabilities, and the system generates diagnostic indications automatically, substituting the manual evaluation process with computational processing.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If simple point-based scoring systems are used, then the testing process is straightforward, but detection accuracy for early cognitive impairment is insufficient

Engineering Contradiction:
Improvedetection accuracyVSAvoidanalysis system
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system transforms the simple point-based scoring parameter into a continuous impairment probability parameter calculated by machine learning models. Instead of discrete point totals, the system outputs probabilistic assessments that provide finer granularity and higher precision in detecting early cognitive impairment stages.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent combines multiple data sources including speech-to-text transcriptions, test variable calculations, personal information, and clinical practitioner assessments into a composite analysis framework. This multi-component approach integrates diverse information types to enhance detection accuracy beyond what any single method could achieve alone.

Inventive Principle:
Principle #40Composite materials

3Measurement precision

If multiple neuropsychological tests are administered to improve detection accuracy, then early signs of cognitive impairment can be detected, but the process becomes time-consuming and resource-intensive

Engineering Contradiction:
Improveearly detection accuracyVSAvoidtesting duration
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the analysis process into distinct components: speech-to-text processing, test variable calculation, machine learning impairment probability assessment, and final indication generation. This segmentation allows parallel processing of multiple tests and efficient resource utilization, reducing overall testing time while maintaining comprehensive evaluation.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system employs a universal machine learning framework that can process multiple different neuropsychological tests through a common impairment probability calculation mechanism. The same core processing architecture handles various test types, personal information, and speech data uniformly, improving efficiency compared to separate analysis methods for each test.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4256553B1Detection of cognitive impairment
Publication Date: 2025.09.24 ACCEXIBLE IMPACTO SL
  • EP4256553B1 patent drawingFigure 1
  • EP4256553B1 patent drawingFigure 2
  • EP4256553B1 patent drawingFigure 3

AI summary

A computer-implemented method (1) of detecting cognitive impairment comprising: receiving audio data (21) representing recorded utterances of a patient; processing the audio data using a speech-to-text engine (30) to produce a text transcription (301) of the recorded utterances; processing the text transcription to calculate (41) a plurality of test variables (411) associated with a neuropsychological test; calculating, by applying a trained detection model (51) on the plurality of test variables, an impairment probability (511) indicating a likelihood that the patient suffers from the cognitive impairment; and indicating that the patient suffers from the cognitive impairment if a final impairment probability based on the impairment probability is above a predetermined threshold, and indicating that the patient does not suffer from the cognitive impairment if the final impairment probability is below the predetermined threshold.