一种全新的两步自动化医学图像分割方案

A New Two-step Automatic Segmentation Scheme for Medical Images

针对现有图像分割方法存在需要手动分割,以及精确度较低的问题。采用一种全新的两步图像分割方案。该方案。以基于人工神经网络的模式识别技术,即人工神经网络的大规模培训的方法,通过对肺区不同子区域内结构进行分割处理,利用训练好的大规模人工神经网络对标准胸片中的肋骨、锁骨等骨质结构进行抑制,结合以基于区域的活动轮廓模型,即Snake模型,正确分割亮度不均匀的图像。文中选择与医护人员人工分割的图像进行对比,通过放射科医生采用等级法打分,原图的平均分为2.0分,而通过文中改进的分割方法平均分高达3.4分。

This paper proposed a new image segmentation scheme to solve the manual segmentation by and the poor accuracy of existing image segmentation methods. Firstly,the massive- training artificial neural networks（ MTANNs） are employed to suppress bones in the lungs,using artificial neural network trained on a large scale of standard chest radiograph of the ribs,clavicle,such as inhibition of bone structure. Then active contour model（ ACM） based on region,or the Snake Mode,is adopted to segment correctly images with non- uniform brightness.A comparison with images manually segmented by medical personnel and rated by radiologists shows that the average score of the original images is 2. 0 points while that by our improved segmentation method is 3. 4 points.