Multi-modal fusion based deep learning network for effective diagnosis of Alzheimers disease

Alzheimers Disease (AD) is a prevalent, irreversible, chronic, and degenerative disorder that leads to deterioration of cognitive functions caused by anatomical and functional alteration in the brain. Diagnosis of AD at prodromal stage is critical. Mostly, single data modality such as MRI and PET is used to make predictions in AD studies. However, the functional and structural data fusion can improve the accuracy and provide a holistic view of AD staging analysis. To achieve this objective, we propose a novel multi-modal fusion-based method. At first, we performed an optimal fusion of MRI and PET by harnessingDemon Algorithm and discrete wavelet transform (DWT). Then, the fused images were classified using ResNet-50 and support vector machine (SVM). Experiments on the Alzheimers disease neuroimaging initiative (ADNI) dataset show accuracy 97.88%,sensitivity 96.03%, specificity 98.81%, precision 97.58%, and f-score 96.80%. The proposed model will be beneficial for health professionals in accurately diagnosing AD.