This paper improves the regional term of the region-based geometric active contour model initially proposed by J.S.Suri.Thanks to the new region based regularity term

the improved algorithm not only solves the underlying problem on the stability of the primary algorithm

but also effectively improves the speed of segmentation.Along with the more accurate segmentation performances

the algorithm is also able to segment various cerebral tissues such as the white matter

gray matter and cerebrospinal fluid.The random multi-seed initialization is used to further minimize the sensitivity of the algorithm to the initial condition while disusing the manual intervention.The experiments on simulative and real MR images demonstrate the feasibility and the effectiveness of the improvement on the regional term.The comparison and analysis of the segmentation results under the noisy conditions also indicates the robustness of the proposed algorithm.