针对CT图像肺肿瘤分割中复杂大肿瘤分割的准确性和自适应问题,提出了一种基于随机游走算法的分割方法.首先,根据图像灰度信息提取肺实质;针对大肿瘤与周围肺组织粘连的复杂情况,先提取有凹陷的肺实质,再根据肺实质先验轮廓,用曲线段形变模型修补肺实质的凹陷边界.然后,用区域生长法自动确定目标种子点和背景种子点;对于大血管与肿瘤粘连的情况,需要少量交互修改个别背景种子点.最后,用随机游走算法完成大肿瘤的分割.实验结果表明,该方法的准确性高,分割结果能够满足临床治疗效果分析和病理学研究的要求.
To the segmentation accuracy and adaption of large lung tumors with complex structures in CT images, a segmentation method based on random walk algorithm is proposed. First, the lung parenchyma was extracted according to the gray level information. If the large tumors adhered to normal tissues surrounding lungs, the lung parenchymas with concave edges are first extracted, and then the lung parenchyma with concave edges is refined by the curve section deformation model according to the prior shape of lung. Then a random walk algorithm is implemented, where the object and background seeds are automatically found by region growing method,and interaction are needed to modify a few background seed points if vasa adhere to tumors. The experiment results show that this method is accurate, which can meet the requirements of clinical treatment and pathological analysis.
[1] ALLEMANI C, MATSUDA T, CARLO V D, et al. Global surveillance of trends in cancer survival 2000-14(CONCORD-3):analysis of individual records for 37513025 patients diagnosed with one of 18 cancers from 322 population-based registries in 71 countries[J]. Lancet, 2018, 391(10125): 1023-1075. DOI: 10.1016/S0140-6736(17)33326-3.
[2] HOSSEINIASL E, ZURADA J, GIMEL F G, et al. 3D Lung segmentation by incremental constrained nonnegative matrix factorization[J]. IEEE Transactions on Biomedical Engineering, 2016, 63(5):952-963. DOI: 10.1109/TBME.2015.2482387.
[3] JOHNSON P B, YOUNG L A, LAMICHHANE N, et al. Quantitative imaging: correlating image features with the segmentation accuracy of PET based tumor contours in the lung[J]. Radiotherapy & Oncology, 2017, 123(2):257-262. DOI: 10.1016/j.radonc.2017.03.008.
[4] PULAGAM A R, KANDE G B, EDE V K R, et al. Automated lung segmentation from HRCT scans with diffuse parenchymal lung diseases[J]. Journal of Digital Imaging, 2016, 29(4):507-519. DOI: 10.1007/s10278-016-9875-z.
[5] CHAN T F, VESE L A. Active contours without edges[J]. IEEE Transactions on Image Processing, 2001, 10(2):266-277. DOI: 10.1109/83.902291.
[6] NIAN Y, LI M, CUI H, et al. Graph-based unsupervised segmentation for lung tumor CT images[Z]. IEEE International Conference on Computer & Communications, Chengdu,China,2018. DOI: 10.1109/CompComm.2017.8322865.
[7] GRADY L. Multilabel random walker image segmentation using prior models[Z]. IEEE Computer Society Conference on Computer Vision & Pattern Recognition, San Diego,USA, 2005. DOI: 10.1109/CVPR.2005.239.
[8] 顾潇蒙. 基于随机游走算法的CT图像肺区域和肺肿瘤的分割研究[D]. 保定: 河北大学, 2016.
[9] TAN Y, SCHWARTZ L H, ZHAO B. Segmentation of lung lesions on CT scans using watershed, active contours, and markov random field[J]. Medical Physics, 2013, 40(4):043502. DOI: 10.1118/1.4793409.
[10] PERONA P, MALIK J. Scale-space and edge detection using anisotropic diffusion[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1990, 12( 7):629-639. DOI: 10.1109/34.56205.
[11] GU Y, KUMAR V, HALL L O, et al. Automated delineation of lung tumors from CT images using a single click ensemble segmentation approach[J]. Pattern Recognition, 2013, 46(3):692-702. DOI: 10.1016/j.patcog.2012.10.005.
