An Optimized Computer-Aided Diagnosis System for MRI Classification

Computer-aided diagnosis (CAD) stands out as a prominent area of research within diagnostic radiology and medical imaging. This paper introduces an optimized CAD system employing the Multi-Objective Optimization (MOO) approach for the classification of brain Magnetic Resonance Images (MRIs). The proposed methodology leverages an Optimized Kernel Principal Component Analysis (OKPCA) combined with an Artificial Neural Network (ANN), referred to as the OKPCA-ANN scheme, to discern between pathological and normal brain images. The study unfolds across three key stages: initial data acquisition and preprocessing, followed by the dimensionality reduction phase, and concluding with the classification stage. To enhance the generalizability of the scheme, we incorporate the k-fold cross-validation technique, facilitating adaptation to diverse benchmarks. Our dataset encompasses images representing seven brain diseases, namely Huntington's disease, sarcoma, glioma, Alzheimer's disease, Alzheimer's disease plus visual agnosia, meningioma, and Pick's disease, all sourced from the ‘Harvard Medical School’ website. The experimental results underscore the exceptional efficiency of the proposed technique within CAD systems, affirming its robustness and potential for application in various medical imaging scenarios.