Development of AI tools for identifying neurodegenerative diseases Lead Investigator: Timothy Rittman Institution : University of Cambridge E-Mail : tr332@medschl.cam.ac.uk Proposal ID : 1587 Proposal Description: We are interested in developing novel AI algorithms for identifying neurodegenerative diseases. There is currently a lack of well validated tools for identifying neurodegenerative tools based on specific brain regions. We have two immediate projects: 1. Identifying Alzheimer's disease using volumetric and cortical thickness data. We have trained a deep learning algorithm to identify the structural neuroimaging signature of Alzheimer's disease. A key feature of this algorithm is the use of node dropout to generate a measure of certainty for the results obtained. We will validate the performance of this method in the NACC dataset. 2. Developing an AI model for volumetric analysis. We have applied a novel 3D-UNET model for accurate volumetric analysis of specific brain regions that are relevant to neurodegenerative disorders. Initially, we will focus on the hippocampus and the parietal lobe (for Alzheimer's disease), and specific regions of the frontal lobe (for Frontotemporal Dementia), and mid-brain pons ratio (for Progressive supranuclear Palsy). We have developed this in our local memory clinic dataset and wish to validate our findings in the NACC dataset as an independent cohort. We will assess the value of the volumetric measurements to predict diagnosis, baseline cognitive score and cognitive decline. We expect the outputs to be both academic publications and open source neuroimaging AI tools to be used as biomarkers in neurodegenerative diseases.