| 基于3DV-Net的肺结节检测分割算法 |
| |
| 引用本文： | 刘方1,孙鹏2,陈真诚3. 基于3DV-Net的肺结节检测分割算法[J]. 中国医学物理学杂志, 2023, 0(1): 77-82. DOI: DOI:10.3969/j.issn.1005-202X.2023.01.013 |
| |
| 作者姓名： | 刘方1 孙鹏2 陈真诚3 |
| |
| 作者单位： | 1.桂林电子科技大学生命与环境科学学院， 广西 桂林 541000； 2.桂林电子科技大学电子工程与自动化学院， 广西 桂林 541000；3.桂林电子科技大学生命与环境科学学院广西高校生物传感与仪器重点实验室， 广西 桂林 541000 |
| |
| 基金项目： | 广西自然科学基金（2018GXNSFDA281044）； |
| |
| 摘 要： | 目的：提出基于深度学习的肺结节识别与分割算法，以辅助医生进行肺部疾病检测。方法：针对LUNA16数据集数据量大以及肺结节种类大小多样性等特征，采用基于改进的深度神经网络3DV-Net实现多种肺结节的检测分割，然后使用ResNet对结节图像和非结节图像进行分类。对LUNA16数据集中的肺部CT图像进行图像去噪、插值采样等预处理，然后生成粗分割图像和Mask图像，再使用改进的3DV-Net模型对数据进行多次训练预测。网络层级越深，出现梯度消散、梯度爆炸等问题的概率越大，改进的3DV-Net使用残差连接来改善这一问题。结果：改进的3DV-Net的Dice相似系数和IoU分别达到88.29%和88.25%。结论：本文方法有助于肺结节的检测分割，在肺结节的辅助诊断方面有重要意义。
|
| 关 键 词： | 肺部疾病 CT图像 肺结节分割 3DV-Net 深度学习 |
Detection and segmentation of pulmonary nodules using improved 3DV-Net |
| |
| LIU Fang1,SUN Peng2,CHEN Zhencheng3. Detection and segmentation of pulmonary nodules using improved 3DV-Net[J]. Chinese Journal of Medical Physics, 2023, 0(1): 77-82. DOI: DOI:10.3969/j.issn.1005-202X.2023.01.013 |
| |
| Authors: | LIU Fang1 SUN Peng2 CHEN Zhencheng3 |
| |
| Affiliation: | 1. School of Life and Environmental Sciences, Guilin University of Electronic Technology, Guilin 541000, China 2. School of Electronic Engineering and Automation, Guilin University of Electronic Technology, Guilin 541000, China 3. Key Laboratory of Guangxi Colleges and Universities for Biosensors and Instruments, School of Life and Environmental Sciences, Guilin University of Electronic Technology, Guilin 541000, China |
| |
| Abstract: | Abstract: Objective To propose a deep learning-based algorithm for the recognition and segmentation of pulmonary nodules, thereby assisting doctors in the diagnosis of pulmonary diseases. Methods In view of the large amount of data in LUNA16 data set and the diversity of types and sizes of pulmonary nodules, an improved deep neural network 3DV-Net was adopted to complete the detection and segmentation of various pulmonary nodules, and then ResNet was used to classify the nodule and non-nodule images. The lung CT images in LUNA16 data set were preprocessed by image denoising and interpolation sampling. After coarse segmentation images and mask images were generated, the improved 3DV-Net model was used to carry out multiple training and prediction. The improved 3DV-Net network adopted skip block to solve the problem that with the deeper network level, the probability of gradient dissipation, gradient explosion and other issues was greater. Results The Dice similarity coefficient and IoU of the improved 3DV-Net reached 88.29% and 88.25%, respectively. Conclusion The proposed method is helpful to the detection and segmentation of pulmonary nodules and is of great significance in the auxiliary diagnosis of pulmonary nodules. |
| |
| Keywords: | Keywords: pulmonary disease CT image pulmonary nodule segmentation 3DV-Net deep learning |
|
| 点击此处可从《中国医学物理学杂志》浏览原始摘要信息 |
|
点击此处可从《中国医学物理学杂志》下载免费的PDF全文 |
