FuzzyGuard: A Novel Multimodal Neuro-Fuzzy Framework for COPD Early Diagnosis

Early detection is critical to effectively and efficiently managing chronic obstructive pulmonary disease (COPD) and improving patient outcomes. To enhance early COPD detection, this article presents a new multimodal neuro-fuzzy framework called “FuzzyGuard.” FuzzyGuard uses ensemble learning on various datasets, such as computed tomography (CT) scans and audio recordings of coughs and lung sounds, to guarantee comprehensive analysis and accurate diagnosis. The neuro-fuzzy random vector functional link (RVFL) is used in FuzzyGuard for clinical relevance evaluation and early COPD prediction. FuzzyGuard flexibility is increased by using RVFL to pick hyperparameter tuning parameters, including learning rate (η), momentum (μ), the number of epochs, and regularization coefficient. Using ensemble deep learning techniques, the FuzzyGuard-based framework extracts discriminating features from the chest diagnostic images and sputum samples from cough, as well as lung sound samples for COPD classification using an RVFL network, initialized with a random vector. FuzzyGuard’s excellent accuracy is demonstrated by the assessment, which yielded rates of 99.97% (CT-scan model), 96.98% (cough-based model), and 98.65% (output of FuzzyGuard) for early diagnosis of COPD based on early weighted sum-fusion approaches applied to cough and chest X-ray data. FuzzyGuard outperforms established benchmarks, marking a substantial leap in the diagnosis of COPD, and offers better patient treatment and respiratory health outcomes. © 2025 Elsevier B.V., All rights reserved.