摘要
肺结节检测技术目前是医学图像处理领域中的一个热门研究课题。该文旨在探讨如何快速有效地完成打通整个肺结节检测的流程,实现对输入的胸部CT图像进行肺结节检测。该文利用patch与patch之间的空间关系,针对数据集进行合理的数据增强和模型调优,提高模型泛化能力。使用图像分割检测图像中所有可能是肺结节的区域,生成候选集,使用一种基于改进后的3D-Unet医学图像处理模型对上一步骤生成的结果进行分类,剔除假阳性的候选,保留真正的结节,实现对输入的胸部CT图像进行肺结节检测。该技术不仅可用于肺结节CT智能诊断,也能应用于甲状腺癌、乳腺癌、脑瘤、肾癌、肝癌等重大疾病,具有重要的医学价值和社会价值。
Lung nodule detection technology is a hot topic in the f ield of medical image processing.This paper aims to explore how to quickly and ef fectively complete the whole process of pulmonar y nodule detection,so as to realize the detection of pulmonary nodules on the input chest CT images.In this paper,the spatial relationship between patch and patch is used to enhance the data set and optimize the model to improve the generalization ability of the model.Image segmentation is used to detect all possible areas of lung nodules in the image,and candidate sets are generated.An improved 3D UNET medical image processing model is used to classify the results generated in the previous step.False positive candidates are eliminated and t he real nodules are retained to realize t he detection of pulmonary nodules in t he input chest CT image.The technology can not only be used in CT intelligent diagnosis of pulmonary nodules,but also be applied to thyroid cancer,breast cancer,brain tumor,renal cancer,liver cancer and other major diseases,which have important medical and social value.
作者
陈星宇
CHEN Xingyu(School of Mathematics and Statistics,Southwest University,Chongqing,401120 China)
出处
《科技资讯》
2020年第24期217-219,共3页
Science & Technology Information
基金
2019年西南大学大学生创新创业训练项目《基于深度学习的肺结节医学图像辅助诊断研究》(项目编号:X20190635030)支持。