Brain tissue magnetic resonance imaging segmentation using anisotropic textural features

One of the most useful diagnostic tests for brain diseases is magnetic resonance imaging (MRI). It is a demanding task to segment brain tissue into cerebrospinal fluid (CSF), gray matter (GM), and white matter (WM) for early diagnosis of brain diseases and their causes based on MRI images. This study proposed a novel CSF, GM, and WM segmentation method employing just one modality (FLAIR image). The proposed segmentation is based on an anisotropic textural analysis of brain MRIs. For this purpose, the gray level co-occurrence matrix (GLCM) and curvelet transform were combined. The curvelet transform is an anisotropic multi-resolution method that was fully exploited for brain tissue segmentation. In addition to the information richness of GLCM features, the Relief method was utilized to achieve the best feature subset. Finally, support vector machine (SVM) and fuzzy C-means (FCM) were applied to recognize each pixel's label. FCM provided better segmentation results for CSF, GM, and WM with more selected features than SVM. Furthermore, FCM could track the area changes of scan sequences more accurately than SVM. Our segmentation framework involves analyzing an anisotropic curvelet of the statistical features, feature selection, clustering, and a classification-based method for segmentation. The proposed method outperforms well compared to other methods implemented on the MRBrainS18 challenge dataset. This study's outcomes can automatically detect the area of the brain tissue for all scans and capture the variations, reducing the specialist's burden of evaluating each scan and improving the performance.