AUTOMATIC DETECTION AND MULTI-CLASS CLASSIFICATION OF MANAGEMENT AND PREVENTION OF NEURODEGENERATIVE DISEASES STAGES USING PROGRESSIVE WASSERSTEIN GENERATIVE ADVERSARIAL NETWORK

Povzetek

Alzheimer diseased (AD) is a neurodegenerative disease that impacts on the memory and cognitive capacity of a person [1]. It is a non-curable disorder, which advances continuously. Therefore, the proposed framework introduces an innovative framework for classifying AD stages effectively. The proposed scheme comprises multiple steps: data procurement, pre-processing, feature extraction (FE), and classification. At first, the input neuroimaging data is acquired from a familiar database on neurodegenerative illness. Once pre-processed, a procedure called normalized gamma-based contrast-limited improved histogram equalization (NG-CLIHE) is presented in order to eliminate data artefacts. Extended concise empirical wavelet transform (Ex-CEWT) is then fed with pre-processed data in order to extract features. The obtained feature vectors are further optimized by uniting Ladybug Beetle Optimizer (LBO) with Salmon migration optimizer (SMO) that are referred to as improved Ladybug Beetle Optimization algorithm (Imp-LBOA) to optimize weight parameter of Ex-CEWT. Finally, the Progressive Wasserstein generative adversarial network (PW_GAN) is provided with the optimized features to implement a classification of several stages of AD into the categories of CN, SMC, EMCI, LMCI, MCI and AD. The suggested algorithm is implemented in MATLAB and the performance measures such as accuracy, precision, sensitivity, F1-Score, and the consumption of the computational time (TC) are compared. The proposed framework achieves the overall accuracy of 96.51, precision of 96.95, sensitivity of 96.47, F-measure of 96.96 and 191.36s respectively.