摘要目的：探讨CT图像纹理分析对胰腺神经内分泌肿瘤（Pancreatic neuroendocrine neoplasm，PNEN）病理分级的诊断价值。方法：回顾性分析109例经手术或病理活检确诊为胰腺神经内分泌肿瘤患者的CT动脉期及门脉期图像，ITK-Snap软件手动勾画感兴趣区（ROI），A.K.软件提取ROI纹理特征，采用最小绝对收缩与选择算子（The least absolute shrinkage and selection operator，LASSO）降维运算获取纹理特征及其受试者工作特征曲线下面积（AUROC）。结果：筛选出动脉期纹理特征共5个，分别为熵（GLCMEntropy_AllDirection_offset7_SD）、峰度（kurtosis）、肿瘤在三维空间内的最大直径（Maximum 3D diameter）、变换后形成的正态曲线内0.025分位处的数值（Quantile0.025）及肿瘤表面积与体积的比值（Surface volume ratio），其AUROC分别为0.715、0.529、0.724、0.672及0.698，门静脉期特征2个，为肿瘤在三维空间内的最大直径及肿瘤表面积与体积的比值，其AUROC分别为0.722及0.703。结论：CT图像纹理分析可用于判断PNEN的病理分级。
Abstract：Objective: To evaluate the accuracy of the texture analysis to determinate the pathological grades of pancreatic neuroendocrine neoplasms(PNEN). Methods: 109 cases of pancreatic neuroendocrine neoplasms, confirmed by surgery or pathological biopsy, were retrospectively enrolled in our study. Both arterial phase and portal vein phase CT images of including patients were manual sketch the region of interest(ROI) by ITK Snap software. A.K. software was used for texture extraction, and R software with The least absolute shrinkage and selection operator(LASSO) was used for calculation. Results: In aterial phase, 5 texture features were selected, including maximum 3D diameter, kurtosis, GLCMEntropy_AllDirection_offset7_SD, quantile0.025 and Surface volume ratio, with AUROC of 0.715, 0.529, 0.724, 0.672 and 0.698, respectively. In portal vein phase, 2 texture features were selected including maximum 3D diameter and surface volume ratio, with AUROC of 0.722 and 0.703, respectively. Conclusions: Texture analysis of CT images can be used to evaluate the pathological classification of PNEN.
于浩鹏，李 谋，张 琳，杨成敏，张永嫦，宋 彬. CT图像纹理分析评估胰腺神经内分泌肿瘤的病理分级[J]. 中国临床医学影像杂志, 2018, 29(11): 788-791.
YU Hao-peng, LI Mou, ZHANG Lin, YANG Cheng-min, ZHANG Yong-chang, SONG Bin. Texture analysis of CT images used to assess pathological grades of pancreatic neuroendocrine neoplasms. JOURNAL OF CHINA MEDICAL IMAGING, 2018, 29(11): 788-791.
[1]Lewis RB, Lattin GE, Paal E. Pancreatic endocrine tumors: radiologic-clinicopathologic correlation[J]. Radiographics, 2010, 30(6): 1445-1464.
[2]Jeon SK. Nonhypervascular pancreatic neuroendocrine tumors: differential diagnosis from pancreatic ductal adenocarcinomas at MR imaging-retrospective cross-sectional study[J]. Radiology, 2017, 284(1): 77-87.
[3]Bosman FT. WHO classification of tumours of the digestive system[M]. 4th ed. Geneva, Switlzerland: World Health Organization, 2010.
[4]Tsutsumi K. Analysis of risk factors for recurrence after curative resection of well-differentiated pancreatic neuroendocrine tumors based on the new grading classification[J]. J Hepatobiliary Pancreat Sci, 2014, 21(6): 418-425.
[5]Ohtsuka T. Surgical treatment of gastroentero-pancreatic neuroendocrine tumor[J]. Gan To Kagaku Ryoho, 2013, 40(7): 843-846.
[6]王齐艳. 胰腺神经内分泌肿瘤的多层螺旋CT表现与其病理分级的关系[J]. 中国普外基础与临床杂志，2016，23（2）：243-247.
[7]Takumi K. Pancreatic neuroendocrine tumors: correlation between the contrast-enhanced computed tomography features and the pathological tumor grade[J]. Eur J Radiol, 2015, 84(8): 1436-1443.
[8]Kim DW. Neuroendocrine neoplasms of the pancreas at dynamic enhanced CT: comparison between grade 3 neuroendocrine carcinoma and grade 1/2 neuroendocrine tumor[J]. Eur Radiol, 2015, 25(5): 1375-1383.
[9]Canellas R. Characterization of portal vein thrombosis(neoplastic versus bland) on CT images using software-based texture analysis and thrombus density(hounsfield units)[J]. AJR, 2016, 207(5): 81-87.
[10]Ryu YJ. Glioma: application of whole-tumor texture analysis of diffusion-weighted imaging for the evaluation of tumor heterogeneity[J]. PLoS One, 2014, 9(9): 108335.
[11]Ng F. Assessment of primary colorectal cancer heterogeneity by using whole-tumor texture analysis: contrast-enhanced CT texture as a biomarker of 5-year survival[J]. Radiology, 2013, 266(1): 177-184.
[12]Partelli S. Systematic review of active surveillance versus surgical management of asymptomatic small non-functioning pancreatic neuroendocrine neoplasms[J]. Br J Surg, 2017, 104(1): 34-41.
