Medical Evaluation of Improved Label Fusion Based Haematoma Segmentation in Traumatic Brain Injury Images
P. Manimegalai1, U. S. Ragupathy2
1P. Manimegalai, PG Scholar, Department of Electronics and Instrumentation Engineering, Kongu Engineering College, Erode (Tamil Nadu), India.
2Dr. U. S. Ragupathy, Professor & Head, Department of Electronics and Instrumentation Engineering, Kongu Engineering College, Erode (Tamil Nadu), India.
Manuscript received on December 21, 2018. | Revised Manuscript received on December 28, 2018. | Manuscript published on January 05, 2018. | PP: 9-13 | Volume-7 Issue-6, January 2018. | Retrieval Number: F3101017618/2018©BEIESP
Open Access | Ethics and Policies | Cite
©The Authors. Published By: Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/
Abstract: Atlas based segmentation is a well-known method of automatically computing the segmentation. When multiple atlases are available, then each atlas can be used to compute a ‘label’, which is an estimation of the ground truth segmentation of a target image. By combining these labels, a more accurate approximation of the ground truth segmentation can be made. In the proposed work, the axial view of brain CT image for target and prelabelled images are taken for haematoma segmentation. The canny edge detection is performed to detect the wide range of edges in the images. The edge detected images are registered by using the rigid transformation method to spatially align one image to fit into another. The atlas images are selected based on the fixed threshold value and all the selected atlases are combined by using Selective and Iterative Method of Performance Level Estimation (SIMPLE) algorithm in label fusion process for the accurate segmentation of haematoma. The label fusion process is performed for a set of 6 labelled images and 10 target images and from the results it is observed that the error is reduced by 3% and similarity coefficient is increased by 16%, which indicates that the proposed method performs better when compared to the existing method.
Keywords: Multi Atlas based segmentation, Registration, Edge Detection, label fusion, Brain Images, SIMPLE.