Skull Stripping in MR Images Using an Adaptive Deformable Model with Dynamic Brain Intensity Detection
Yu-Sheng Chen and Herng-Hua Chang
Computational Biomedical Engineering Laboratory (CBEL)/Department of Engineering Science and Ocean
Engineering/National Taiwan University, Daan 10617 Taipei, Taiwan
Abstract—Skull stripping is an important preprocessing step in many medical image applications. Deformable models are powerful as they provide robust abilities to deform contours under the guidance of geometric properties. In particular, the charged fluid model has been shown its superiority over many existing deformable models. This paper is in an attempt to propose a new skull stripping algorithm based on the charged fluid model. To improve the segmentation accuracy, a new image balancing coefficient of using the local intensity difference along the normal line of the evolving curve is introduced. Stimulated by the concept of the Mumford-Shah model, the other balancing coefficient obtained from a global intensity difference between the interior and exterior of the evolving contour is also introduced to automate the segmentation process. We have adopted the BrainWeb and internet brain segmentation repository (IBSR) image datasets to evaluate this new algorithm. Experimental results indicated that our method produced high segmentation accuracy across a wide variety of brain magnetic resonance (MR) images, which is promising in many MR image processing applications.
Index Terms—segmentation, MRI, charged fluid model, deformable model, skull stripping
Cite: Yu-Sheng Chen and Herng-Hua Chang, "Skull Stripping in MR Images Using an Adaptive Deformable Model with Dynamic Brain Intensity Detection," International Journal of Pharma Medicine and Biological Sciences, Vol. 4, No. 1, pp. 56-60, January 2015.