Alzheimer Detection System Using Hybrid Deep Convolutional Neural Network

Abstract

Alzheimer’s disease of the sixth leading causes of death in the United States of America is projected to grow to the third place of all causes of death for the elderly soon to cancer and heart decease. Timely detection and prevention are crucial to it. AD detection is based on multiple medical examinations which all lead to extensive multivariate heterogeneous data. This factor makes manual comparison, evaluation, and analysis hardly possible. The hereby study proposes a new approach to the detection of AD at the earliest stage hybrid deep learning algorithms. Several feature extraction and selection draw possible features. The method involves InceptionV3 and DenseNet for both pre-processing and classification tasks, while MobileNet enables data pre-processing and object detection. Experimental results with 100 epochs and 15 hidden layers show InceptionV3 has an accuracy of 98%, which outperforms other models available. The comparative analysis with other CNN models endorses the proposed method, achieving the highest performance across the board from our system.