Research On Image Segmentation Method And System Of Medical Interest Area

The regional segmentation of medical image is the key technical step of computer aided detection, CT segmentation of lung tissue segmentation and capsule endoscopic gastrointestinal lesion segmentation are common medical imaging assistive technique step, which is of great significance to the subsequent analysis of medical image.This paper investigated the segmentation of lung tissue and the segmentation of gastrointestinal lesions of collagen endoscopy, the specific work and main achievements are as follows:Firstly, a method of image segmentation of CT lung tissue combined with regional growth and random walk was proposed. The image was denoised by the guidance filter and binarized to avoid the influence of noise on subsequent processing steps. Then the region growing strategy was used to extract thorax regions, lung regions were segmented using a seed-based random walk algorithm, then the lung tissue was filled with voids and the curvature of the lung area was corrected to improve the segmentation accuracy, and on the basis of the previous, with the lung masks, the lung region was automatically segmented from a CT slice. Experimental results show that this method demonstrated a more powerful discriminating ability and included sufficient pleural nodule regions and diffuse areas.Secondly,a method of segmentation of gastrointestinal images based on visual significance activity profile model is proposed. The significance of the detection method to select the capsule endoscopic image of the initial lesion area based on the and then use the active contour model to further divide the initial gastrointestinal lesion area. Experimental results show that this method has better effect on the lesion area, better convergence of the lesion area of the convex and concave boundary area.Thirdly, on the basis of the above research work, the image segmentation system of medical interest area is designed and completed. The system includes two methods and other classical segmentation algorithms.