Abstract #3268

Machine Learning to Improve Compatibility of Multi Centre Arterial Spin Labelling Data to Characterise Dementia Spatial Perfusion Abnormality

Jack Highton¹, Rebecca Steketee², Rozanna Meijboom³, Marion Smits², Ross Paterson⁴, Nick Fox⁴, Alexander Foulkes⁵, Catherine Slattery⁵, Jonathan Schott⁵, Enrico De Vita⁶, and David L Thomas^7,8,9

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, ⁴Institute of Neurology, University College London, London, United Kingdom, ⁵Dementia Research Centre, University College London, London, United Kingdom, ⁶King's College London, London, United Kingdom, ⁷University College London, London, United Kingdom, ⁸Neuroradiological Academic Unit, Department of Brain Repair and Rehabilitation, Institute of Neurology,, University College London, London, United Kingdom, ⁹Wellcome Centre for Human Neuroimaging, Institute of Neurology, University College London, London, United Kingdom

Arterial Spin Labelling (ASL) is a non-invasive MRI method to measure cerebral blood flow (CBF). Here, Support Vector Machine (SVM) based machine learning was used to identify spatial patterns of CBF abnormality in patients with Alzheimer’s Disease from two cohorts scanned at different centers. Support Vector Machine Regression models were found to be more accurate than conventional SVMs previously used for dementia classification. Then, motivated by the lack of ASL standardisation, an SVM based method was used to improve the compatibility of the two studies by removing differences in spatial patterns likely caused by differences in hardware and acquisition protocol.

This abstract and the presentation materials are available to members only; a login is required.

Join Here