Cerebral Cortex Extraction Methods Based on a Priori Knowledge for T1-Weighted MRI Images

Cerebral Cortex Extraction (CCE) plays a significant role in clinical applications such as pre-surgical planning and tumor segmentation. However, designing an efficient CCE technique is still a challenging task. In this work, we propose two efficient methods for CCE from T1-weighted MRI images. The first method (named CCE-AK) is divided in two phases: Pretreatment phase and CCE phase. Indeed, the input image is firstly filtered by a Gaussian filter to smooth the image and reduce noise. Thereafter, we apply the anisotropic diffusion to improve the texture quality on the filtered image. Thus, a binary image is obtained after the integration of a priori knowledge and the thresholding steep using the Otsu's method to simplify treatment and eliminate non-brain portions. After that, we start the second phase by eroding the image via a structuring element to eliminate the outer brain parts. In order to extract the Cerebral Cortex (CC), we look for the Largest Connected Component (LCC) in the eroded image. Finally, we use the dilation operation to preserve the totality of the CC region. However, the LCC concept failed in few slices to identify the CC correctly. To address this issue, we introduce a second method (CCE2), which makes use of information in the adjacent slices. To assess the performance, experiments are conducted on different MRI datasets collected from the Surgical Planning Laboratory (SPL). The proposed methods achieve better results in both visual effects and objective criteria than three popular methods (SPM, BET and BSE).