CT image labeling using simulated annealing algorithm

Segmentation of computed tomography (CT) head images is required by many image analysis procedures for quantitative measurements of human spontaneous intracerebral brain hemorrhage (ICH). In this work we describe a stochastic method for segmentation of CT head images based on simulated annealing (SA). In the proposed method, the segmentation problem is defined as the pixel labeling problem with labels for this particular application set to: background, skull and ICH, and brain tissue. The proposed method is based on the Maximum A-Posteriori (MAP) estimation of the unknown pixel labels. A Markov random field (MRF) model has been used for the posterior distribution. The MAP estimation of the segmented image has been determined using the simulated annealing algorithm. Experimental results have demonstrated good results and proved the usability of the method.