The document discusses a study that aimed to classify Alzheimer's disease using deep learning on T1-weighted brain MRI images. Specifically, it sought to 1) build a dataset from the ADNI database combining MRI images with clinical data, 2) train a deep neural network to classify images as Cognitive Normal or Alzheimer's, and 3) evaluate techniques for addressing class imbalance. The researchers explored the ADNI data, finding correlations between diagnostic labels and attributes. They then trained a ResNet model for binary classification but faced class imbalance issues given the rarity of Alzheimer's cases. To address this, they evaluated random and stratified undersampling of the majority class as well as oversampling the minority class using a WGAN-GP for synthetic image generation.