ETRI-Knowledge Sharing Plaform



논문 검색
구분 SCI
연도 ~ 키워드


학술대회 Texture Weighted Fuzzy C-Means Algorithm for 3D Brain MRI Segmentation
Cited 0 time in scopus Download 0 time Share share facebook twitter linkedin kakaostory
이지영, 김동윤, 문진영, 강석, 손성호, 신성
Research in Adaptive and Convergent Systems (RACS) 2018, pp.291-295
18ZR1200, 지능형 전파센서 및 무선 에너지 전송 원천기술 개발, 이호진
The segmentation of human brain Magnetic Resonance Image (MRI) is an essential component in the computer-aided medical image processing research. Fuzzy C-Means (FCM) algorithm is one of the practical algorithms for brain MRI segmentation. However, Intensity Non-Uniformity (INU) problem in brain MRI is still challenging to existing FCM. In this paper, we propose the Texture weighted FCM (TFCM) algorithm performed with Local Binary Patterns on Three Orthogonal Planes (LBP-TOP). By incorporating texture constraints, TFCM could take into account more global image information. The proposed algorithm is divided into following stages: Volume of Interest (VOI) is extracted by 3D skull stripping in the pre-processing stage. The initial FCM clustering and LBP-TOP feature extraction are performed to extract and classify each cluster's features. At the last stage, FCM with texture constraints refines the result of initial FCM. The proposed algorithm has been implemented to evaluate the performance of segmentation result with Dice's coefficient and Tanimoto coefficient compared with the ground truth. The results show that the proposed algorithm has the better segmentation accuracy than existing FCM models for brain MRI.
3D Brain MRI segmentation, Clustering, Feature extraction, Fuzzy C-Means, Local Binary Patterns
KSP 제안 키워드
3D brain MRI, Brain magnetic resonance image, Dice's coefficient, FCM Clustering, Feature extractioN, Fuzzy C-means (FCM) algorithm, Fuzzy c-means algorithm, Global image information, Intensity non-uniformity, LBP-TOP, Last stage