Aim: We aim to develop deep learning techniques to diagnose Alzheimer’s Disease (AD) based on existing neuroimaging data.
Social/scientific motivation: AD is a chronic neurodegenerative disease that usually starts slowly, worsens over time, and is fatal. It is the cause of 60% to 70% of all cases of dementia. As the disease worsens, symptoms can include articulation problems, disorientation, rapid changes in mood, avolition, not managing self care, and behavioural issues. The patient’s change over time is often mistaken as normal ageing, such as withdrawing from group and family activities. The cause of AD is under active study. No treatments can stop or reverse its progression, though some may temporarily reduce symptoms, but only if AD is diagnosed early. Currently, AD can only be diagnosed with certainty post mortem by doing an autopsy of the brain.
Scientific background: AI solutions have been developed and used for a long time already. Expert systems are an example that has been in use with varying levels of success for years in the fields of science, technology, and medicine. Over the past decade Deep learning has been used to outperform all other methods and in some cases even humans in tasks, such as image recognition. Supervised learning of deep neural networks, however, require large sets of labelled data. Labelled neuroimaging data is scarce and more data efficient learning strategies are needed for detection of early onset of AD.
My motivation: We believe that if we can successfully train a predictive machine learning model with a limited set of data we can first develop an early-detection AD to improve current treatment and then develop a general prediction model. We hope that this will help to detect AD.
Members: Jose Ramon Chang, Paul Tsai