Brain tumor segmentation with radius contraction and expansion based initial contour detection for active contour model

Abstract

This paper proposes a novel brain tumor segmentation algorithm that uses Active Contour Model and Fuzzy-C-Means optimization. In Active Contour model, the initial Contour selection is a challenging task for MRI brain tumor segmentation because the accuracy of active contour segmentation depends on initial contour. This method uses the two level morphological reconstruction processes such as Dilation and Erosion along with thresholding process for minimizing the non-tumor region. The segmented region thus obtained is not accurate that also contains non-tumor region. Also there is a chance of missing the tumor region along with background while performing two level morphological reconstructions. In order to overcome these issues, active contour model is used to segment the complete tumor part. The initial Contour for Active Contour model is detected by forming a circular region around the tumor region. The radius of the circular region is contracted or expanded based on the shape of the tumor. This proposed Radius Contraction and Expansion (RCE) technique is used to select the initial contour of active Contour model. Further Fuzzy-C-Means algorithm is used to optimize the edge pixels because the boundary of active contour model output also contains the non tumor pixels. The performance of the proposed segmentation algorithm was evaluated using the metrics such as specificity, sensitivity, dice score, Probabilistic Rand Index (PRI) and Hausdorff Distance (HD) on T1- weighted contrast enhanced image dataset. The experimental result shows that the proposed segmentation algorithm provides a good performance when compared to the state-of-the-art segmentation methods.