Neuroimaging-driven machine learning models for early diagnosis of Alzheimer's disease Lead Investigator: Vijaya Kolachalama Institution : E-Mail : vkola@bu.edu Proposal ID : 955 Proposal Description: We are seeking NACC data to improve early detection of AD by developing cutting-edge quantitative tools from uniquely rich imaging datasets. Our goal is to leverage this dataset and develop unbiased, image-driven machine learning models for early, accurate diagnosis of AD. To accomplish this, we will: Aim I: Develop a deep learning model based on convolutional neural network (CNN) that will differentiate between individuals with no cognitive impairment (CI) at baseline who progress to mild CI and those who do not, over a 5-year period. This model will also predict which of those with psychometrically defined mild CI at baseline progress to AD in 5 years and which do not. We will leverage CNN architecture to associate MRI images from the NACC cohort at baseline with the NPA performed at the 5-year follow-up time. For CNN, pixel-level information at baseline will serve as the input and the categorical cognitive status at 5 years will be the output.