Convolutional Neural Network using Multi-modal imaging for dementia

To date, deep learning studies have successfully used image-to-image neural networks, including U-Net architectures, to segment MRI images and quantify brain atrophy based on segmentation maps. We will adapt image generation deep learning networks for a novel application: to predict the future cortical atrophy rate in elderly populations with and without cognitive impairment, using image-derived cortical atrophy rate as an anatomical biomarker.