Automatic segmentation of intracranial hematoma and volume measurement

Abstract

In this paper, a two-step segmentation method is developed for segmenting the hematoma area from brain CT images. The volume of hematoma area is calculated after the segmentation. During the second segmentation process, the method of two-dimensional entropy is introduced to separate hematoma. In using the method of two-dimensional entropy, most important is to find the optional threshold which can be achieved by an improved genetic algorithm (GA) i.e. hierarchical genetic algorithm (HGA). HGA is more efficient than simple GA in overcoming the shortcoming of standard GA in local optimal solution and low precision convergence. An experiment is designed to test the effectiveness of automatic segmentation. The results prove that the precision of automatic segmentation is better than artificial segmentation, and the clinical needs are met.

Development and Validation of an Automatic Segmentation Algorithm for Quantification of Intracerebral Hemorrhage

“…11,[26][27][28] Nowinski and colleagues have compared agreement of different segmentation algorithms with respective manual segmentations. CCCs for algorithms based on clustering, graph theory, and modified thresholding were 0.87, 0.91, and 0.97, respectively, and they are comparable to the CCCs calculated for our segmentation algorithm.…”

Computer-aided diagnosis of intracranial hematoma with brain deformation on computed tomography

“…dx Imaging modality Current hematoma yes yes no yes CT CBMS98 [8] hematoma N/D yes no NO CT SPIE99 [9] hematoma yes yes no NO CT CIARP07 [10] hematoma N/D yes no NO CT CMIG07 [11] hematoma NO NO YES yes CT JNT08 [12] hematoma yes N/D N/D yes CT EMBS08 [13] hematoma yes yes no N/D CT CMIG09 [14] hematoma yes yes no NO CT AJNR01 [15] stroke NO no yes NO CT EMBS04 [16] stroke N/D yes NO NO MRI (DWI) MIAR04 [17] stroke yes no yes NO MRI qualitative diagnosis. Section 5 will focus on hematoma volume measurement at the original resolution using a multiresolution binary level set method.…”

“…There has been a couple of CAD systems for brain images in the literature, most of them are used to diagnose intracranial hematoma [8][9][10][11][12][13][14] or ischemic stroke [15][16][17]. A brief summery is given in Table 1.…”

“…CT is the study of choice in all systems. The intracranial hematoma regions are defined as those with large average CT number, determined using thresholds on the histogram derived manually or automatically by some algorithms such as fuzzy C-means [8,9], probabilistic neural network [8], fuzzy maximal likelihood estimation [9], and genetic algorithm [13]. Preprocessing procedures such as median filtering [10] and top-hat morphological transformation [11] have been applied in some systems.…”

“…Approaches used by CAD systems for diagnosing intracranial hematomas are generally similar [8][9][10][11][12][13][14]. CT is the study of choice in all systems.…”