Abstract #5293

A fully unsupervised method for spinal cord lesion segmentation in Multiple Sclerosis

Carole Hélène Sudre^1,2, Ferran Prados^1,3, Rosanna Cortese³, Marios Yiannakas³, Hugh Kearney³, Olga Ciccarelli³, Sébastien Ourselin^1,2, Claudia Angela Gandini Wheeler-Kingshott^3,4,5, and M. Jorge Cardoso^1,2

¹Translational Imaging Group, CMIC, Department of Medical Physics and Biomedical Engineering, University College London, London, United Kingdom, ²UCL Institute of Neurology, Dementia Research Centre, University College London, London, United Kingdom, ³Queen Square MS Centre, UCL Institute of Neurology, Faculty of Brain Sciences, University College London, London, United Kingdom, ⁴Department of Brain and Behavioural Sciences, University of Pavia, Pavia, Italy, ⁵Brain MRI 3T Research Centre, C. Mondino National Neurological Institute, Pavia, Italy

The presence of focal lesions in the spinal cord is an important diagnostic criteria for Multiple Sclerosis (MS). Accurate estimation of lesion volume is important for monitoring disease progression over time. However, manual and automated lesion segmentation for volume estimation remain challenging, since they rely respectively on the skills of the rater or on the automated criteria set within the algorithms. In this work, we present an adaptation to the spinal cord, of a fully unsupervised hierarchical model selection framework that automatically detects abnormality tissue patterns without any a priori knowledge on pathology location.

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

Join Here