Meeting Banner
Abstract #3006

A Computer-Aided Detection of Cerebral Microbleeds on Minimum Intensity Projection MR Images of SWI

Wei Bian1, 2, Christopher P. Hess2, Susan M. Chang3, Sarah J. Nelson1, 2, Janine M. Lupo2

1Graduate Program in Bioengineering, University of California San Francisco & Berkeley, San Francisco, CA, United States; 2Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, CA, United States; 3Neurological Surgery, University of California San Francisco, San Francisco, CA, United States

A computer-aided method is proposed to identify cerebral microbleeds on minimum intensity projected susceptibility-weighted MR images. It utilizes the 2D fast radial symmetry transform to initially detect nearly all possible putative CMBs. False positives are subsequently eliminated by examining geometric features measured after performing 3D region growing on the putative CMBs. The performance of the method was evaluated on 15 patients with total 420 CMBs induced by radiation treatment for resected gliomas. The evaluation showed that our method achieved a detection sensitivity of 88.3% using only 1 minute computation, which is much more sensitive and faster than previously published methods.

Abstract PDF Presentation Video

Keywords

absolute accurate achieved aided although arrow automated background binary biomedical brain candidates centroid cerebral chosen circular classifiers clinical comprises compromising computation computer computing confounding considered construct contrast core counted decreased designing detect detected detection diagram diameter directly discovery distinguished distorted distribution edges elevated eliminated empirically error evaluate evaluated examination examining extent false fast features fifteen final geometric geometrical gradient graduate gray greatest greatly grow growing highlight highlights highly human identified identify independently induced initial initially inspection intensity linear local locally magnitude mask measured mimics minute missed nearly need needed neurology next optimized others output outputs overall passed patient patients performance pixels posed positive positives potential prior priority processing program projected projection proposed radiation radiology radius raters remove reported review scanner scheme screening segmentation selected sensitivity slices smaller specificity step still structures subjected subsequence subsequently supported symmetry terms therapy thereafter transform treatment true truth typically undergo unit utilizing vessel vessels visual volume