Wavelet-based Alzheimer’s Disease Detection using Edge-Weighted Local Binary Pattern

Alzheimer’s disease is a neurodegenerative syndrome that influences the people worldwide. As the disease progresses, it reduces the cognitive skills in individuals and makes them unable to perform their everyday tasks. In the proposed method a new feature extraction technique is proposed which captures the important biomarkers for early Alzheimer’s disease detection. In proposed technique Dual tree complex wavelet transform has been applied on 2-D MR slices to obtain six sub-band images. Further Edge weighted Local Binary Pattern has been applied on sub-band images to obtain edge prominent texture patterns. The dimensionality of these patterns is reduced by using Principal Component Analysis. The extracted features are then used for Alzheimer’s disease classification. Experimental results show that edge weighted Local Binary Pattern gives better classification results than normal Local Binary Pattern as edge weighted Local Binary Pattern provides the local gradient variation information. Different models have been compared in the paper to test the potential of proposed method. The proposed method has been implemented on Open Access Series of Imaging Studies dataset and remarkable results have been obtained in terms of accuracy of 94.2%, sensitivity of 96%, and specificity of 93%. The proposed method surpasses the performance of latest methods existing in literature for Alzheimer’s Disease detection.