Related Experiment Video
Updated: Jul 9, 2025
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.1K
Exhaled breath signal analysis for diabetes detection: an optimized deep learning approach
Anita Gade1, V Vijaya Baskar2, John Panneerselvam3
1Department of Electronics Engineering, Sathyabama Institute of Science and Technology, Chennai, India.
Computer Methods in Biomechanics and Biomedical Engineering
|December 8, 2023
Summary
This study introduces a novel deep learning system for breath analysis, achieving over 98% accuracy in diabetes detection. The hybrid model significantly outperforms traditional methods, offering a promising advancement in non-invasive diagnostic tools.
Area of Science:
- Biomedical Engineering
- Artificial Intelligence
- Computational Biology
Background:
- Diabetes mellitus poses a significant global health challenge, necessitating advanced diagnostic tools.
- Breath analysis offers a non-invasive method for disease detection and monitoring.
- Current diagnostic methods can be invasive or lack real-time capabilities.
Purpose of the Study:
- To develop a flexible deep learning system for accurate diabetes detection using breath analysis.
- To extract and utilize relevant breath signal features for enhanced diagnostic performance.
- To optimize a hybrid deep learning model for improved accuracy and efficiency.
Main Methods:
- Pre-processing of raw breath signals to enhance data quality.
- Extraction of key features including Improved IMFCC, BFCC, DWT, peak detection, QT, and PR intervals.
- Development of a hybrid classifier using an optimized Deep Belief Network (DBN) and Bidirectional Gated Recurrent Unit (BI-GRU) model, fine-tuned with a novel Sine Customized by Marine Predators (SCMP) algorithm.
Main Results:
- The proposed hybrid deep learning model achieved an accuracy exceeding 98% in diabetes detection.
- The model demonstrated significant accuracy improvements over traditional methods like CNN+LSTM, CNN, LSTM, RNN, SVM, RF, and DBN.
- Specific improvements ranged from 18.2% to 57% at a 60% learning percentage.
Conclusions:
- The developed deep learning system provides a highly accurate and efficient method for diabetes detection via breath analysis.
- The novel hybrid model, incorporating SCMP-tuned DBN and BI-GRU, represents a significant advancement in non-invasive diagnostics.
- This approach holds potential for early diabetes screening and management.
