Fully Automatic Method for 3D T1-Weighted Brain Magnetic Resonance Images Segmentation

Accurate segmentation of brain MR images is of interest for many brain disorders. However, due to several factors such noise, imaging artefacts, intrinsic tissue variation and partial volume effects, brain extraction and tissue segmentation remains a challenging task. So, in this paper, a full automatic method for segmentation of anatomical 3D brain MR images is proposed. The method consists of many steps. First, noise reduction by median filtering is done; second segmentation of brain/non-brain tissue is performed by using a Threshold Morphologic Brain Extraction method (TMBE). Then initial centroids estimation by gray level histogram analysis is executed, this stage yield to a Modified version of Fuzzy C-means Algorithm (MFCM) that is used for MRI tissue segmentation. Finally 3D visualisation of the three clusters (CSF, GM and WM) is performed. The efficiency of the proposed method is demonstrated by extensive segmentation experiments using simulated and real MR images. A confrontation of the method with similar methods of the literature has been undertaken trough different performance measures. The MFCM for tissue segmentation introduce a gain in rapidity of convergence of about 70%.