Segmentation of solid subregion of high grade gliomas in MRI images based on active contour model (ACM)

13th South-East Asian Congress of Medical Physics 2015

物理学;医药卫生

Seow, P.^1,2 ; Win, M.T.^1,2 ; Wong, J.H.D.^1,2 ; Abdullah, N.A.^3 ; Ramli, N.^1,2

Department of Biomedical Imaging, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia^1

University of Malaya Research Imaging Centre, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia^2

Department of System and Computer Technology, Faculty of Computer Science and Information Technology, University of Malaya, Kuala Lumpur, Malaysia^3

关键词: Active contour model;    Automatic segmentations;    Gradient echo sequences;    High-grade gliomas;    Interstitial tissues;    Manual segmentation;    Quantitative evaluation;    Skull stripping;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/694/1/012043/pdf
DOI  :  10.1088/1742-6596/694/1/012043

学科分类：卫生学

来源: IOP

【 摘 要 】

Gliomas are tumours arising from the interstitial tissue of the brain which are heterogeneous, infiltrative and possess ill-defined borders. Tumour subregions (e.g. solid enhancing part, edema and necrosis) are often used for tumour characterisation. Tumour demarcation into substructures facilitates glioma staging and provides essential information. Manual segmentation had several drawbacks that include laborious, time consuming, subjected to intra and inter-rater variability and hindered by diversity in the appearance of tumour tissues. In this work, active contour model (ACM) was used to segment the solid enhancing subregion of the tumour. 2D brain image acquisition data using 3T MRI fast spoiled gradient echo sequence in post gadolinium of four histologically proven high-grade glioma patients were obtained. Preprocessing of the images which includes subtraction and skull stripping were performed and then followed by ACM segmentation. The results of the automatic segmentation method were compared against the manual delineation of the tumour by a trainee radiologist. Both results were further validated by an experienced neuroradiologist and a brief quantitative evaluations (pixel area and difference ratio) were performed. Preliminary results of the clinical data showed the potential of ACM model in the application of fast and large scale tumour segmentation in medical imaging.