Hough transform detection of the longitudinal fissure in tomographic head images

Abstract

A technique is presented for automatic detection of the longitudinal fissure in tomographic scans of the brain. The technique utilizes the planar nature of the fissure and is a three-dimensional variant of the Hough transform principle. Algorithmic and computational aspects of the technique are discussed. Results and performance on coronal and transaxial magnetic resonance data show that the algorithm is robust with respect to variations in image contrast in the data and to slight anatomic anomalies. A crucial resolution requirement in the data for accurate parameter estimations is a sufficient number of slices covering the whole brain. The Sobel magnitude edge operator, used for preprocessing, proved adequate for magnetic resonance scans with positive and negative brain/cerebrospinal fluid contrast.