Tóth, Márton J.Blaskovics, TamásRuskó, LászlóDelso, GasparCsébfalvi, BalázsJan Bender and Arjan Kuijper and Tatiana von Landesberger and Holger Theisel and Philipp Urban2014-12-162014-12-162014978-3-905674-74-3https://doi.org/10.2312/vmv.20141279Recognition of body parts in three-dimensional medical images is an important task in many clinical applications. It can facilitate image segmentation, registration methods and it can be the first step of an automatic image-processing workflow. In this paper, we propose an automated method to classify the axial slices of threedimensional magnetic resonance image series according to the body part they belong to. We apply the Zernike transform to obtain feature vectors representing the structural information of the axial slices. Using machine learning tools statistical correlation is found between the extracted feature vectors and the position of the slices within the human body. The initial classification is filtered by a dynamic programming based error correction method that takes the correct sequence of anatomy regions into consideration to eliminate the false recognitions. Using our approach, different body regions can be recognized at high precision rate.I.4.8 [Image Processing and Computer Vision]Scene AnalysisObject recognition