Related Experiment Video
Updated: Nov 17, 2025
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
Breath analysis based early gastric cancer classification from deep stacked sparse autoencoder neural network
Muhammad Aqeel Aslam1, Cuili Xue1, Yunsheng Chen1,2
1Institute of Nano Biomedicine and Engineering, Shanghai Engineering Research Center for Intelligent Instrument for Diagnosis and Therapy, Department of Instrument Science & Engineering, School of Electronic Information and Electrical Engineering, Yantai Information Technology Research Institute of Shanghai Jiao Tong University, Shanghai Jiao Tong University, 800 Dongchuan Road, Shanghai, 200240, People's Republic of China.
This study introduces a deep learning model for early gastric cancer detection using breath analysis. The computer-aided diagnosis system achieved high accuracy, showing promise for clinical applications.
Area of Science:
- Medical Diagnostics
- Artificial Intelligence in Medicine
- Biomarker Discovery
Background:
- Gastric cancer poses a significant global health challenge, with early detection crucial for improving patient survival rates.
- Current diagnostic methods can be invasive or lack sensitivity for early-stage disease.
- Computer-aided diagnosis (CAD) systems offer a promising avenue for enhancing disease detection accuracy and efficiency.
Purpose of the Study:
- To develop and evaluate a novel deep learning-based computer-aided diagnosis (CAD) system for the early detection of gastric cancer.
- To utilize breath sample analysis for non-invasive gastric cancer screening.
- To extract discriminative features from unlabeled breath data using a stacked sparse autoencoder.
Main Methods:
- A stacked sparse autoencoder was employed for unsupervised feature extraction from breath sample spectra.
- A Softmax classifier was integrated with the autoencoder for gastric cancer classification.
- The deep stacked sparse autoencoder neural network was fine-tuned using labeled data for improved reliability.
Main Results:
- The proposed CAD system achieved high accuracy: 98.7% for advanced gastric cancer and 97.3% for early gastric cancer detection.
- The model demonstrated excellent performance in recall, precision, and F-score values.
- The system effectively identified fifty spectral peaks to distinguish between early gastric cancer, advanced gastric cancer, and healthy individuals.
Conclusions:
- The developed deep stacked sparse autoencoder neural network is a reliable and effective tool for gastric cancer detection via breath analysis.
- This non-invasive CAD system shows significant potential for clinical application in early gastric cancer screening.
- The method successfully extracts discriminative features from unlabeled breath data, reducing the gap between input and output for accurate diagnosis.
