A Segmentation Framework for MR Images Combining with Tissue Characteristics

Magnetic Resonance (MR) brain tissue segmentation plays an important role in assisting diagnosis, treatment planning and surgical navigation. In this paper, a framework combining with T1 mapping is proposed to segment white matter (WM), gray matter (GM) and cerebrospinal fluid (CSF). For the T1 parameters are intrinsic MR properties of tissue, it can be used as prior knowledge for MR images segmentation. The pro-posed framework is mainly divided into two parts: one is T1 mapping acquisition and preprocessing, another is prediction and reclassification of mislabeled voxels based on support vector machine (SVM). Evaluated by Dice coefficient, the experiment results are significantly improved above 15 percent for brain tissue segmentation. In a conclusion, T1 mapping can be seen as objective prior information for MR images segmentation.