Classifying clinical images: a new aid to dementia diagnosis with implications for treatment strategies

Richard Frackowiak
Department of Clinical Neuroscience, Service of Neurology, CHUV University Hospital, Lausanne, Switzerland

The application of modern computerized automated techniques for analyzing structural and functional brain images have led to real advances in the application of advanced neuroimaging to clinical practice in ways undreamed of only a short time ago. One major advance is the development of image classification techniques that puts diagnosis of the individual at the centre of the enterprise. This approach is currently based on machine learning techniques, some of the best results being obtained with support vector machines (SVM). There is a lot of activity in this area at present and new methods of analysis and results are constantly being reported. MR scanner manufacturers are becoming interested in translating these encouraging results into potential products.

Thus, neuroimaging techniques, in addition to their traditional diagnostic role are currently expanding understanding of the structural and functional changes that occur in dementia. Further research may allow identification of early pathological signs of AD, before clinical symptoms are evident, providing the opportunity to test preventative therapies.

The lecture will review imaging in Alzheimer’s disease and other neurodegenerative diseases and attempt to project into the future how the field will develop. Additionally obstacles to such developments will be highlighted and approaches to validating image classification as a diagnostic tool and a means of monitoring treatment discussed.

Keywords: machine learning, neurodegeneration, neuroimaging, support vector machines