Automated segmentation of a ventricle boundary from CT brain image based on naïve Bayes classifier

The ventricle is filled with cerebrospinal fluid (CSF) in the brain. Some brain diseases are caused by changing of the ventricle shape or volume. The ventricle shape and volume are used to diagnose patients who have brain diseases. This paper proposes an algorithm of digital image processing for segmentation of a ventricle from CT brain images. The process starts with normalizing the CT brain images and extracts the region of interest using profile of gray level. In the segmentation step, we apply Bayesian segmentation to classify intensities into 3 classes: white matter, gray matter, and CSF. The proposed algorithm segments the area of CSF that is obtained by the posterior probability from Bayes´ rule. Finally, the ventricle is evaluated with the relatively ground truth from a neurologists. Our experimental results from the proposed algorithm reveal a low error of 3.14% and a standard deviation of 1.41.