Performance evaluation of segmentation methods for brain CT images based hemorrhage detection

For the patients with neurological disturbance and head injury, brain hemorrhage may be considered as one of the primary causes. It is the leading cause of human deaths in the world following heart disease and cancer. The rupturing of blood vessels present inside the brain results in brain hemorrhage. It consists of bleeding outside or inside of the brain substance. Computed Tomography (CT) is one of the medical imaging methods used for the detection of brain hemorrhages because of its wide availability, quick to perform, less expensive, and depicts good bony details. However as the quality of the CT image depends the dose of X-rays, worst tissue contrast, intensity inhomogeneity, and irregular boundaries cause significant difficulties in the detection of hemorrhage from brain CT images. Thus in this paper the performance of various segmentation methods namely thresholding, region growing, fuzzy clustering and active contour based methods are evaluated for hemorrhage detection from brain CT images. The results of these methods are compared on the basis of manually delineated hemorrhagic regions by the subject expert.