Voice Biomarker Modeling for T2DM Prediction With Reduced Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current voice analysis systems for predicting type-2 diabetes mellitus (T2DM) face challenges in efficiently identifying vocal changes between non-diabetic and diabetic individuals, particularly in age- and BMI-matched populations, and require significant processing capabilities for real-time analysis on various platforms while minimizing patient burden.
Innovation Solution
A computer-implemented method using voice biomarkers, such as pitch, intensity, and jitter features, combined with age and BMI data, to predict T2DM status through improved recording schedules and reduced sample numbers, utilizing mobile applications for voice recording and predictive models like Logistic Regression, Naïve Bayes, and Support Vector Machines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If voice analysis systems use comprehensive voice biomarker features and predictive models for T2DM detection, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts and analyzes only the most relevant voice biomarker features (pitch, intensity, jitter, shimmer) from the complete voice signal, rather than processing all possible acoustic parameters. This selective extraction approach maintains high detection accuracy while reducing computational complexity and processing requirements
Solution Approach 2:
The system performs preliminary processing of voice samples by extracting key biomarker features before applying predictive models. This pre-processing step simplifies the subsequent analysis by preparing optimized feature sets that reduce computational burden while preserving diagnostic accuracy
2Measurement precision
If voice recordings are collected more frequently and in greater numbers, then measurement precision is improved, but loss of time and patient burden increase
Solution Approach 1:
The patent determines that a limited number of voice recordings (rather than continuous or frequent recording) are sufficient to achieve high prediction accuracy. By identifying the minimum necessary sample size through statistical analysis, the system achieves adequate measurement precision without excessive time investment or patient burden
Solution Approach 2:
The system optimizes recording parameters by determining optimal intervals and durations for voice sample collection. By adjusting these parameters based on statistical requirements rather than using maximum possible sampling, the system achieves high accuracy while minimizing time loss and patient burden
Data Source
AI summary
Provided is a computer-implemented device for generating a type-Il (T2DM) diabetic status prediction, including: a memory comprising a diabetic status prediction model; and a processor in communication with the memory, the processor configured to: receive a voice sample from the subject; extract at least one voice biomarker feature value from the voice sample for at least one predetermined voice biomarker feature; determine the type-II (T2DM) diabetic status prediction for the subject based on the at least one voice biomarker feature value and the diabetic status prediction model; and output, to an output device, the type-II (T2DM) diabetic status prediction for the subject or an output based on the diabetic status prediction.