Comparative analysis of pulmonary nodules segmentation using multiscale residual U-Net and fuzzy C-means clustering. (September 2021)

• Record Type: Journal Article
• Authors: Shi, Jianshe; Ye, Yuguang; Zhu, Daxin; Su, Lianta; Huang, Yifeng; Huang, Jianlong
• Is Part Of: Computer methods and programs in biomedicine. Volume 209(2021)
• Publication Date: 2021-09
• Subjects: Multiscale residual U-Net -- Convolutional neural network -- Fuzzy C-means clustering -- Pulmonary nodules; Medicine -- Computer programs -- Periodicals; Biology -- Computer programs -- Periodicals; Computers -- Periodicals; Medicine -- Periodicals; Médecine -- Logiciels -- Périodiques; Biologie -- Logiciels -- Périodiques; Biology -- Computer programs; Medicine -- Computer programs; Periodicals; Electronic journals; 610.28
• DOI: 10.1016/j.cmpb.2021.106332
• Languages: English
• ISSNs: 0169-2607

Abstract

Highlights: An automatic segmentation algorithm for pulmonary nodules based on multiscale residual U-Net is proposed. The accuracy of the multi-scale residual U-Net is 94.57% and the segmentation time is 3.15 s. CT scan images of 58 patients with different pulmonary nodules were selected. In the improved fuzzy C-means clustering segmentation, coordinate values of nodes and seed points in the image are combined with the spatial distance. Fuzzy C-means clustering and manual segmentation are used for comparison. Abstract: Background and Objective: Pulmonary nodules have different shapes and uneven density, and some nodules adhere to blood vessels, pleura and other anatomical structures, which increase the difficulty of nodule segmentation. The purpose of this paper is to use multiscale residual U-Net to accurately segment lung nodules with complex geometric shapes, while comparing it with fuzzy C-means clustering and manual segmentation. Method: We selected 58 computed tomography (CT) scan images of patients with different lung nodules for image segmentation. This paper proposes an automatic segmentation algorithm for lung nodules based on multiscale residual U-Net. In order to verify the accuracy of the method, we also conducted comparative experiments, while comparing it with fuzzy C-means clustering. Results: Compared with the other two methods, the segmentation of lung nodules based on multiscale residual U-Net has a higher accuracy, with an accuracy rate of 94.57%. This method not only maintains a high accuracy rate, but also shortens the recognition time significantly with a segmentation time of 3.15 s. Conclusions: The diagnosis method of lung nodules combined with deep learning has a good market prospect and can improve the efficiency of doctors in diagnosing benign and malignant lung nodules.