Chronic Obstructive Pulmonary Disease Detection Using Electronic Nose Breath Sensor Data and a Machine Learning Framework

Abstract

Chronic Obstructive Pulmonary Disease (COPD) is a progressive respiratory disorder that significantly affects health, leading to reduced quality of life and increased mortality. Conventional diagnostic techniques for COPD are often invasive, expensive, and time-consuming. In this study, we propose a non-invasive diagnostic approach using breath-based Volatile Organic Compounds (VOCs), which reflect underlying metabolic processes associated with disease. To enhance detection accuracy, optimal feature selection is performed using a hybrid Scatter Search–Aquila Optimization (SS-AO) algorithm. The optimal features obtained by the hybrid Scatter Search–Aquila Optimization (SS-AO) algorithm are classified using the proposed Adaptive Neuro-Fuzzy Inference System with Caputo Fractional Gradient Descent (ANFIS-CFGD). In this framework, ANFIS effectively models nonlinear relationships, while CFCD enhances learning stability and convergence, thereby reducing misclassification rates and improving diagnostic accuracy. Experimental results demonstrate that ANFIS-CFGD achieves an accuracy of 94% and outperforms existing methods across multiple performance metrics, thus confirming its potential as an efficient tool for the early diagnosis of COPD from breath analysis.