Automatic brain hemorrhage segmentation and classification algorithm based on weighted grayscale histogram feature in a hierarchical classification structure

Brain hemorrhage is the first cause of death in ages between 15 and 24, and the third after heart diseases and cancers in other ages. Saving the lives of such patients completely depends on detecting the correct location and type of the hemorrhage in an early stage. In this paper, an automatic brain hemorrhage detection and classification algorithm on CT images is proposed. To achieve this purpose, after preprocessing, a modified version of Distance Regularized Level Set Evolution (MDRLSE) is used to detect and separate the hemorrhage regions. Then a perfect set of shape and texture features from each detected hemorrhage region are extracted. Moreover, we define a synthetic feature that is called weighted grayscale histogram feature. In this feature, valuable information from shape, position and area of the hemorrhage are integrated with the grayscale histogram of hemorrhage region. After that a synthetic feature selection algorithm is applied to select the most convenient features. Eventually, the segmented regions are classified into four types of the hemorrhages such as EDH, ICH, SDH and IVH by a hierarchical structure of classification. Our proposed algorithm is evaluated on a perfect set of CT-scan images and obtains the accuracy rate of 96.15%, 95.96% and 94.87% for the segmentation of the EDH, ICH, and SDH types, respectively. Also our proposed classification structure provides the accuracy rate of 92.46% and 94.13% for the first and second classifiers of the hierarchical classification structure for classifying the IVH from normal class and the EDH, ICH and SDH hemorrhage classes, respectively. (C) 2015 Nalecz Institute of Biocybernetics and Biomedical Engineering. Published by Elsevier Sp. z o.o. All rights reserved.