Classification of Brain Hemorrhages in MRI Using Naïve Bayes- Probabilistic Kernel Approach

Literature Survey

In this section the overview of the existing concepts which led to this work is discussed. For the early detection of abnormal changes in tissues and organs, MRI is an important imaging technique. An automatic unsupervised segmentation method by integrating Dual tree complex wavelet transform with K-mean algorithm was proposed for Brain MR Image. The brain region was extracted by using expectation-maximization segmentation software (EMS), and intensity inhomogeneities were corrected. Finally, spatial constrained K-mean algorithm was proposed for automatic T1-weighted brain MR image segmentation. (Jingdan Zhang, 2014) The intensity inhomogeneities in medical images affect the efficiency of region-based image segmentation methods whereas the performance of level set segmentation gets affected by initialization and configuration of controlling parameters. Hence, to address this a hybrid method which integrates a local region-based level set method with a variation of fuzzy clustering was proposed. (Marayam Rastgarpour, 2014) The hemorrhage identification is difficult when the area of blood is small or clinician is not experienced. Therefore, for automatic hemorrhage segmentation many researchers have been working. A hybrid method for detecting brain hemorrhage regions from clinical head CT scans was implemented. This method could detect ICH and higher attenuation signal caused by subarachnoid hemorrhage (SAH) or intra-ventricle hemorrhage (IVH). (Yonghong Li, 2009)