Enhanced Machine Learning Framework for Chronic Gastritis Prediction from Electronic Nose Breath Sensor Data

Abstract

Chronic gastritis is a prevalent inflammatory condition of the gastric mucosa that, if not properly diagnosed, it can progress to ulcers or gastric cancer. Breath analysis using non-invasive electronic nose (e-nose) systems has developed as a promising diagnostic approach, detecting volatile organic compounds associated with the disease. This paper presents a hybrid Fuzzy Deep Convolutional Neural Network (F-DCNN) framework for gastritis prediction from e-nose sensor data. The proposed method integrates pre-processing, feature selection, deep convolutional feature extraction, and fuzzy inference to identify individual patterns. Experiments comparing SVM, k-NN, ANN, conventional DCNN, and F-DCNN show that F-DCNN achieves superior performance, with an accuracy of 96.23% and the highest AUC among all tested classifiers. The proposed F-DCNN algorithm achieves efficient processing, ensuring fast execution. These results highlight the potential of hybrid deep-fuzzy models in medical diagnostics, offering both high accuracy and interpretability.