Deep learning based multi-stage classification of Alzheimer’s disease using T1-weighted MRI images

In great majority of instances, dementia is caused primarily by Alzheimer's disease(AD). AD is a brain degeneration disorder that cannot be reversed that causes individuals to gradually lose the cognitive and memory capacities. Alzheimer's disease has no known cure, although early detection of the symptoms allows physicians to prescribe preventative medication, which slows the progression of disease. This study seeks to introduce an automated system that uses structural MRI to distinguish individuals having Alzheimer's disease, progressive mild cognitive impairment(pMCI), stable mild cognitive impairment(sMCI) and are cognitively normal(CN). T1-weighted 1.5 tesla structural MRI scans of 342 cohorts on baseline are collected from ADNI database, preprocessed and classified using CNN, VGG-16, VGG-19 and ResNet-50 classification models. Best testing accuracy of 89.80% 79.80% precision, 79.14% recall and 79.14% F-1 score is achieved by VGG-19 among the four models for classification of CN, sMCI, pMCI and AD classes. sMRI scans are clearer, detailed and cost-effective than other imaging modalities like Positron emission tomography, Single-photon emission computed tomography. Results demonstrates viability of 2D images for the training of models for early identification of AD.