My research is focused on on the early diagnosis of Alzheimer's disease (AD), and predicting of patients with mild cognitive impairment (MCI), a common precursor to AD, will undergo further decline. Early diagnosis will make new treatments more effective, as they can disrupt the disease process, as well as allowing for smaller and faster clinical trials.
Doing this involves the use of a wide variety of biomarkers and machine learning algorithms. As it is one of the structures to most strongly affected by the disease process, the shape of the hippocampus, determined from structural MRI scans, can be used to accurately classify patients as AD or healthy using a support vector machine classifier. This can also be used to predict AD in patients with MCI . More recently, we have introduced Gaussian process classification to the problem, which gives naturally probabilistic predictions based on grey matter maps of the brain . We have since shown that these predictions can be improved by combining the grey matter maps with PET data and genetic biomarkers in a multikernel framework . This is illustrated below.
My most recent work concerns modelling continuous predictors of conversion. We have found that an approach where we try to predict the brain atrophy of patients over a 1-year time span produces a better marker of conversion to AD than using labels denoting which diagnostic group individuals belong to . Currently an expanded version of this method is being submitted for publication. We plan to further study which continuous measures provide the best prediction of conversion to AD, and whether results can be improved by simultaneously predicting several in a multi-task regression. We also plan to use new feature extraction techniques combining voxel level features and anatomical regions derived from multi-atlas automatic parcelation.
1. J. Young, G. Ridgway, K. Leung, and S. Ourselin, 'Classification of Alzheimer's disease patients with hippocampal shape, wrapper based feature selection and support vector machine', 2012, vol. 8314, p. 83140Q–83140Q–7.
2. J. Young, M. Modat, M. J. Cardoso, J. Ashburner, and S. Ourselin, 'Classification of Alzheimer's disease patients and controls with Gaussian processes', in 2012 9th IEEE International Symposium on Biomedical Imaging (ISBI), 2012, pp. 1523 –1526.
3. J. Young, M. Modat, M. J. Cardoso, A. Mendelson, D. Cash, and S. Ourselin, 'Accurate multimodal probabilistic prediction of conversion to Alzheimer's disease in patients with mild cognitive impairment', NeuroImage: Clinical, vol. 2, pp. 735–745, 2013.
4. J. Young, M. Modat, M. J. Cardoso, A. Mendelson, J. Ashburner, and S. Ourselin, 'An oblique approach to prediction of conversion to Alzheimer's Disease with multikernel Gaussian processes', in MLINI 2013: 3rd NIPS Workshop on Machine Learning and Intrpretation in Neuroimaging. (Runner-up best paper)