Meeting Banner
Abstract #0621

Automatic Detection of Cerebral Microbleed in SWI Using Radon Transform

Amir Fazlollahi1, 2, Fabrice Meriaudau2, Luca Giancardo, 23, Patricia M. Desmond4, Victor L. Villemagne5, Christopher C. Rowe5, Paul Yates5, Olivier Salvado1, Bourgeat Pierrick1, the AIBL Research Group6

1The Australian E-Health Research Centre-BioMedIA, Brisbane, QLD, Australia; 2Laboratoire Le2I, Universit de Bourgogne, Le Creusot, France; 3Istituto Italiano di Tecnologia, Genoa, Italy; 4Department of Radiology, The Melbourne Brain Centre at Royal Melbourne Hospital, University of Melbourne, Melbourne, VIC, Australia; 5Department of Nuclear Medicine and Centre for PET, Austin Hospital, Melbourne, VIC, Australia; 6http://www.aibl.csiro.au/, Australia, Australia

Since presence and number of cerebral microbleeds (CMBs) have come to attention as a potential biomarker, an automated scheme to improve visualization is required. In this work, a new approach of CMB identification in SWIs is presented and compared to visual rating. The method relies on two main steps: a 3D anisotropic multi-scale approach that extracts size and centre of all potential CMBs within the image, and feature extraction using the Radon Transform for final classification using a random forest classifier. The novelty of the technique consists in combining Radon transform and multiscale analysis to obtain robust feature descriptors.

Abstract PDF Presentation Video

Keywords

accumulating achieved aging appear applicable around automated bias bottom brain cerebral chain classification classifier combining computed contrast core cross defining described descriptor descriptors detection deviation deviations diagnostic dimensional discard disease efficient eigenvalues even expected expert exposed extraction extracts false feature features field filter final fixed forest future global health hospital hypo identification identify improved improvement intensity inter invariance irregular leave leaving lifestyle local machine manifestation mask matrix maximization medicine midas model necessarily negative neurology noise normalization normalized nuclear objects oriented overall partial past patch performance poorly population positive positives potential previously problematic process producing promising proposed published radial radiology radon random rating read reader relatively required resolution rotation royal scale scheme screening section sections segmentation selected sensitivity shape side since slices specificity spherical step still strength structures subject subjects subset suffer summarized susceptibility symmetry table takes terms threshold touching trained transform true unified variability vary vessel vessels victor viewing visual volume