基于均值漂移和图割的肺结节检测方法
首发时间:2014-09-12
摘要:肺结节的自动检测是肺癌计算机辅助诊断的关键和难点。提出一种基于均值漂移(Meanshift)和图割的肺结节自动检测方法。首先,利用Meanshift对图像进行平滑滤波,滤除复杂背景的噪声干扰;其次,利用基于区域的图割方法实现网络图分割,完成肺部分割。然后,基于规则对肺实质区域定位,选择肺实质内感兴趣区域作为疑似肺结节,并利用轮廓跟踪和凸包运算检测胸腔粘连结节。最后,依照几何形态特征量化并检测肺结节。实验结果表明,所提出算法速度较快、检出率满足计算机辅助诊断要求,能够较好地检测出孤立性肺结节、低对比度结节和胸腔粘连结节。
关键词: 图像分割;肺结节;均值漂移
For information in English, please click here
Lung Nodule Detection based on Meanshift and Graph Cut
Abstract:Automatic lung nodule detection is the key and difficult point in computer aided diagnosis. A method for automatic lung nodule detection is proposed based on meanshift and graph cut. Firstly, use meanshift to smoothly filter the image, filter out noise in complicated background. Secondly, use region-based graph cut to achieve network segmentation, and achieve lung segmentation. Afterward, locate lung parenchyma based on regulation, select region of interesting in lung parenchyma as suspected pulmonary nodule, detect attached nodules by contour tracking and convex hull. Lastly, quantize and detect lung nodules according to geometric morphology. Experimental results show that, the proposed algorithm is faster, the detection rate meets the requirement of computer aided diagnosis, this method can better detect isolation nodules, low contrast nodules and nodules which adhere to lung wall
Keywords: Image segmentation lung nodule meanshift
基金:
论文图表:
引用
No.4608774100062314****
同行评议
共计0人参与
勘误表
基于均值漂移和图割的肺结节检测方法
评论
全部评论0/1000