Clinical study on the application of a high-sensitivity electronic nose on thin-film gas sensor array technology combined with deep learning algorithm for early non-invasive diagnosis of chronic atrophic gastritis

Chronic atrophic gastritis (CAG) is a common digestive disorder often diagnosed late due to its nonspecific symptoms. Our team developed a high-sensitivity electronic nose (HSe-nose) using thin-film gas sensor array technology for early, non-invasive CAG diagnosis by detecting breath odor changes. It directly analyzes original breath samples, unlike traditional ones. With ppb level sensitivity, it generates odor fingerprints, enhancing classification. It’s user-friendly, non-invasive, and can replace gastroscopy and biopsy, with up to 0.1 ppm sensitivity.

The research involved 596 participants from two hospitals, and after applying exclusion criteria, 522 samples were analyzed. Machine learning and pattern recognition methods were used, with the Random Forest algorithm and SMOTE showing the highest classification accuracy, distinguishing CAG patients from healthy controls with an accuracy of 0.9682.

Further analysis with deep learning algorithms revealed significant differences in exhaled odor profiles between CAG and chronic non-atrophic gastritis (CNAG) patients, and between CAG and CAG with intestinal metaplasia (CAG-IM) patients, with accuracies of 85.57 % and 93.75 % respectively. Specific volatile organic compounds (VOCs) such as H₂S, triethylamine, methane, and formic acid were identified as potential CAG markers, while benzene, toluene, xylene, ethylacetate, and isopropanol were found in CAG-IM cases.

The study concludes that the electronic nose is a promising tool for the early and non-invasive diagnosis of CAG, providing a cost-effective, rapid method. The identified VOCs could shed light on the pathophysiology of CAG and its progression to gastric cancer.