A Hybrid Approach to Detection of Brain Hemorrhage Candidates from Clinical Head CT Scans

In this paper we present an approach for detecting brain hemorrhage regions from clinical head computed tomography (CT) scans. Firstly, non-brain tissues are removed by thresholding based on Fuzzy C-means (FCM) clustering. Then, thresholding based on maximum entropy is employed for the candidate hemorrhage region detection. Finally, non-hemorrhage regions and other normal artifacts are differentiated from hemorrhage regions by a knowledge-based classification system. The approach has been validated against 30 clinical brain CT images and compared with Otsu thresholding as well as hierarchical FCM thresholding.