Jack Highton1, Rebecca Steketee2, Rozanna Meijboom3, Marion Smits2, Enrico De Vita4, Jonathan Schott5, and David L Thomas6,7,8
1Department of Medical Physics and Biomedical Engineering, University College London, London, United Kingdom, 2Department of Radiology and Nuclear Medicine, Erasmus Medical Centre, Rotterdam, Netherlands, Rotterdam, Netherlands, 3Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, United Kingdom, 4King's College London, London, United Kingdom, 5University College London, London, United Kingdom, 6Dementia Research Centre, University College London, London, United Kingdom, 7Neuroradiological Academic Unit, Department of Brain Repair and Rehabilitation, Institute of Neurology, University College London, London, United Kingdom, 8Wellcome Centre for Human Neuroimaging, Institute of Neurology,, University College London, London, United Kingdom
Partial volume correction (PVC) is often applied to analyse cerebral blood flow (CBF) in dementia, in order to isolate perfusion changes from atrophy of grey matter. But PVC inevitably results in spatial smoothing of CBF, which may weaken the informative spatial pattern of CBF changes.
Here, machine learning was used to detect Fronto-Temporal Dementia (FTD), considering spatial patterns in CBF from Arterial Spin Labelling, with and without PVC. Also, multimodal voxel concordance analysis untangled the spatial relationship between GM atrophy and CBF. This information was used with CBF PVC to improve the FTD detection accuracy without the atrophy signature.
