Classification of Mild Cognitive Impairment and Alzheimer’s Disease from Magnetic Resonance Images using Deep Learning

Mild cognitive impairment (MCI) is an early stage of Alzheimer's disease (AD). Since AD is unlikely to modify its related and intrinsic decay, early diagnosis is crucial, which gives patients a chance to rearrange their lives. Identifying the MCI level is essential, as it assists with further treatment and preliminary steps to control the forward progression towards AD. Brain tissue segmentation is an important aspect of clinical diagnostic tools, yielding excellent results compared to conventional segmentation methods or individual modalities. In this work, the Meta-Heuristic Markov Random Field Segmentation method is used to separate brain tissue, followed by the extraction of intensity, texture, and shape-based features of the segmented Cerebral Spinal Fluid (CSF) and Grey Matter (GM). With the proposed technique, the pre-processing of the input image improves quality and reduces noise. The segmentation is based on multilevel thresholding using Particle Swarm Optimization (PSO) and further improving using Markov Random Field model. Segmented brain tissue (GM and CSF) are used to extract features on shape, intensity, and texture. The four-layer deep neural network is used to classify features. The proposed method is tested with the standard dataset from Alzheimer's disease-neuro imaging (ADNI). The proposed method achieved a high accuracy rate of 97.5%. A comparison with the previous work yielded results that demonstrate this method's superiority in terms of the classification of AD and MCI.