Automated prediction system for Alzheimer detection based on deep residual autoencoder and support vector machine

Alzheimer's disease (AD) is a type of neurological disorder and is a most frequent cause of dementia across the world. The area of medical imaging has created an advancement in computing power and the use of cutting-edge methods for the analysis, interpretation, processing, and display of images. An advanced imaging method known as Magnetic Resonance Imaging (MRI) produce a high-quality image of the brain, for diagnostic purposes. The proposed method is used for detecting Alzheimer based on multiscale pooling residual autoencoder and Support Vector Machine. The dataset images from Kaggle and ADNI datasets are considered for Alzheimer detection. The image pre-processing is performed using modified optimal curvelet thresholding and a hybrid enhancement approach Octagon histogram equalization with black and white Stretching. Multi scale pooling Residual autoencoder architecture is used for extracting white matter from the image. Support Vector Machine (SVM), Extreme Learning Machine (ELM) and Knearest neighbors algorithm (KNN) is used for classification. SVM shows the overall accuracy for AD classification as 99.77 % for Kaggle and 98.21 % for ADNI dataset.