基于U-net网络的肺部肿瘤图像分割算法研究

Lung Tumor Image Segmentation Algorithm Based on U-net Network

肺癌病灶大多具有体积小、形态多变、与胸腔内膜粘连等特点。随着CT扫描技术推广应用,肺部影像数据呈爆炸增长趋势,这给仅能依赖肉眼观察CT图像从而作出判断的影像科医生带来极大的挑战。针对这一问题,笔者提出了一种基于U-net的肺部肿瘤图像自动分割算法来辅助影像科医生作出判断。其具体实现流程为:先以原CT图像和由专业医生分割的图像为基础,将预处理后的图像输入U-net网络进行模型训练,得到训练模型,并用验证集中的图像验证。笔者所提出的U-net神经网络分割模型的JS准确率达到0.885 5。实验结果显示,该方法能有效分割CT图像中的肿瘤区域,且相比于传统算法更加精确高效。

Most of the lung cancer lesions are characterized by small size,variable morphology and adhesion to the intima of the thoracic cavity.With the popularization and application of CT scanning technology,the image data of the lungs show an explosive growth trend.This poses a great challenge to the image doctors who can only rely on the naked eye to observe the CT images and make judgement.In order to solve this problem,a U-net based automatic segmentation algorithm for lung tumor images is proposed to assist the image surgeon to make a judgment.The specific implementation process is as follows:first,the original CT image and the image segmented by professional doctors are used as the basis,and the pre processed images are input into the U-net network to train the model,and the training model is obtained,and verified by the image in the verification set.The JS accuracy rate of the U-net neural network segmentation model proposed by the author is 0.8855.The experimental results show that the method can effectively segment the tumor area in the CT image,and is more accurate and efficient than the traditional algorithm.