Design and Implementation of Alzheimer's Disease Detection using cGAN and CNN

The importance of diagnosing Alzheimer's disease (AD) and Mild Cognitive Impairment (MCI), its prodromal form, for patient care, has made it a new study focus because of AD's high prevalence and huge social impact on health. Existing research have used machine learning approaches to diagnose AD early by looking at numerous biomarkers. However, in the bulk of current research, the diagnosis performance is hampered by the inability to efficiently extract high-level brain features that indicate the primary abnormalities of anatomical brain regions connected to AD. In order to diagnose AD brain pictures, the suggested model combines a deep convolutional neural network (CNN) and a conditional generative adversarial network (cGAN).The dataset is preprocessed to create a clear representation of the brain's grey matter, and the cGAN is utilised to create more training instances for the model to improve accuracy. According to the limitations of the discriminator, the generator in cGAN-ADD learns to include the illness category feedback from the classifier into the 2D-brain slice image reconstruction process for image improvement in the training phase. In the prediction phase, the stacked CNN layers extract high-level brain properties from category-enhanced 2D-brain slice pictures. The classifier gets the collected brain features in order to output the posterior probability of disease conditions (Normal, AD, Mild and Very Mild). Tensorflow is used to train the model, and Keras is used to evaluate classification accuracy. The model is trained and tested on ADNI dataset.