Enhancing Rheumatoid Arthritis Detection using Bio-Inspired Feature Selection of GOA Model with DBN Classification

Rheumatoid arthritis (RA) is a debilitating autoimmune disease where early and precise diagnosis is crucial for effective intervention. This study introduces an advanced GOAT Optimisation Algorithm (GOA) for feature selection combined with a Deep Belief Network (DBN) to significantly improve RA detection accuracy. Unlike conventional methods, GOA incorporates adaptive mechanisms for enhanced exploration-exploitation balance, enabling more efficient selection of discriminative biomarkers from high-dimensional clinical data. The optimised features are then classified using a DBN with hierarchical feature learning capabilities. Evaluated on a comprehensive RA dataset, our GOAT-DBN framework achieves 98.7% accuracy, 98.2% sensitivity, and 99.1% specificity, surpassing existing approaches like PSO-ANN (96.7%) and standard GOA-DBN (98.2%). Notably, GOAT reduces feature dimensionality by 72% while improving diagnostic reliability, demonstrating its superiority in handling complex medical data patterns. The proposed system not only provides a clinically viable tool for early RA diagnosis but also establishes GOAT as a superior bio-inspired optimiser for medical feature selection problems.