Modified Canny Detector-based Active Contour for Segmentation

Abstract

In the present work, an integrated modified canny detector and an active contour were proposed for automated medical image segmentation. Since the traditional canny detector (TCD) detects only the edge’s pixels, which are insufficient for labelling the image, a shape feature was extracted to select the initial region of interest ‘IROI’ as an initial mask for the active contour without edge (ACWE), using a proposed modified canny detector (MCD). This procedure overcomes the drawback of the manual initialization of the mask location and shape in the traditional ACWE, which is sensitive to the shape of region of region of interest (ROI). The proposed method solves this problem by selecting the initial location and shape of the IROI using the MCD. Also, a post-processing stage was applied for more cleaning and smoothing the ROI. A practical computational time is achieved as the proposed system requires less than 5 minutes, which is significantly less than the required time using the traditional ACWE. The results proved the ability of the proposed method for medical image segmentation with average dice 87.54%.

Keywords

• segmentation
• canny edge detector
• shape feature extraction
• region-based active contour
• active contour without edges