Early identification of AD and MCI given their daily and demographic information Lead Investigator: Nan Xu Institution : Shanghai Jiao Tong University E-Mail : xunannancy@gmail.com Proposal ID : 1060 Proposal Description: This application for NACC data is for research purpose. My colleagues and I am interested in how to distinguish the AD patients from normal controls and further group them into different subtypes by analyzing their clinical and cognitive characterizations. We hope to discover the relations between the clinical variables and the diagnosed AD subtypes with the help of deep learning approaches. If the classification task could be completed with satisfactory results in the dataset provided by NACC, then it should of great significance in reality. After a routine physical measurement followed by our model?s analysis, the elderly with mild or even none symptoms can be aware of the probability of suffering from AD and its four subtypes (typical AD, limbic-predominant AD, hippocampal-sparing AD and no atrophy AD) at an early stage. Those who have higher probabilities of having AD would be recommended to receive a list of further tests to find out if their performance meets the criteria proposed in NINCD/ADRDA or not. That could make sense for early AD detection and provide suggestions for the physicians to make decision if the further professional but expensive tests are necessary or not.