分享:
分享到微信朋友圈
X
临床研究
基于T2WI直方图纹理分析在胰腺实性病变中的诊断价值
彭林 查云飞 曾菲菲 柳柏玉 闫玉辰

Cite this article as: Peng L, Zha YF, Zeng FF, et al. The value-based T2 histogram analysis for differential diagnosis in solid pancreatic lesions. Chin J Magn Reson Imaging, 2020, 11(3): 201-206.本文引用格式:彭林,查云飞,曾菲菲,等.基于T2WI直方图纹理分析在胰腺实性病变中的诊断价值.磁共振成像, 2020, 11(3): 201-206. DOI:10.12015/issn.1674-8034.2020.03.008.


[摘要] 目的 探讨3.0 T磁共振T2WI灰度直方图纹理特征在胰腺实性病变中的诊断价值。材料与方法 回顾性分析117例胰腺实性占位患者的临床及MRI影像学资料,经手术病理证实导管腺癌(pancreatic ductal,PDAC)69例、实性假乳头状瘤(solid-pseudopapillary tumor of pancreas,SPT)12例、神经内分泌肿瘤(pancreatic neuroendocrine,pNET)15例及肿块型胰腺炎(mass-forming type chronic pancreatitis,MFCP)21例,将其分成恶性肿瘤组别(导管腺癌)、良性到低度恶性肿瘤组别(实性假乳头状瘤+神经内分泌肿瘤)、非肿瘤组别(肿块型胰腺炎)。采用GE Omni-Kinetics软件在T2WI序列病变最大层面勾画ROI并自动生成灰度直方图纹理参数,三组间比较采用单因素方差分析(ANOVA),组间两两比较采用LSD-t检验(方差齐)或多个独立样本Kruskal-Walls检验(方差不齐);筛选有统计学差异的参数绘制ROC曲线,评价其鉴别诊断胰腺实性病变的效能。结果 均值、变异度、能量、熵、第5百分位数、第10百分位数、第25百分位数、第50百分位数、第75百分位数、第90百分位数及第95百分位数在三组中差异有统计学意义(P值均<0.05);偏度、峰度在三组间无统计学差异(P值均>0.05);PDAC与MFCP之间变异度的敏感度为82.6%,特异度为85.7%,曲线下面积为0.899,最佳临界值5915.87;PDAC与SPT+pNET间均值的敏感度为64.4%,特异度为87.0%,曲线下面积为0.688,最佳临界值1113.55;MFCP与SPT+pNET之间第90百分位数的敏感度为88.9%,特异度为85.7%,曲线下面积为0.924,最佳临界值837.59,具有较高的鉴别效能。结论 T2WI直方图纹理参数在胰腺来源实性病变之间存在显著性差异,其中均值、百分位数对胰腺实性病变定性及良恶性鉴别具有重要的临床价值。
[Abstract] Objective: To investigate the diagnostic value of 3.0 T magnetic resonance imaging (T2WI) grayscale histogram texture features in solid pancreatic lesions.Materials and Methods: The clinical and MRI imaging data of 117 patients with solid pancreatic mass were retrospectively analyzed, 69 cases of ductal adenocarcinoma, 12 cases of solid pseudopapillary, 15 cases of neuroendocrine tumor and 21 cases of mass pancreatitis were confirmed by surgical pathology. They were divided into malignant tumor group (ductal adenocarcinoma), benign to low-grade malignant tumor group (solid pseudopapilloma and neuroendocrine tumor), and non-tumor group (mass pancreatitis). The region of interest (ROI) was delineated at the maximum lesion on T2WI sequence using GE Omni-kinetics software and automatically generate grayscale texture parameters of straight, comparison between the 3 groups using one-way analysis of variance (ANOVA), Pairwise comparisons between groups were made by LSD-t-test (homogeneity of variance) or by Kruskal-walls test of multiple independent samples (heterogeneity of variance). Parameters with a statistical difference were screened, and the ROC curve was drawn to evaluate the efficacy of differential diagnosis of solid pancreatic lesions.Results: Mean, Variance, Energy, Entropy, 5th percentile, 10th percentile, 25th percentile, 50th percentile, 75th percentile, 90th percentile, and 95th percentile were statistically significant (all P<0.05). There was no statistical difference in skewness and kurtosis between the three groups (all P>0.05). The sensitivity of variance to differentiate PDAC and MFCP was 82.6%, the specificity was 85.7%, the area under the curve was 0.899, and the best cut-off value was 5915.87. The sensitivity of mean to differentiate PDAC and SPT+pNET was 64.4%, the specificity was 87.0% and the area under the curve was 0.688, the cut-off was 1113.55. The sensitivity of 90th percentile to differentiate MFCP and SPT+pNET was 88.9%, the specificity was 85.7% and the area under the curve was 0.924, the cut-off was 837.59. They had a high identification efficiency.Conclusions: There are significant differences in the texture parameters of the T2WI histogram between solid pancreatic lesions. Mean and Percentile have significant clinical value in the qualitative and benign or malignant differentiation of solid pancreatic lesions.
[关键词] 实性病变;胰腺;磁共振成像;纹理分析;诊断价值
[Keywords] solid lesions;the pancreas;magnetic resonance imaging;texture analysis;diagnostic value

彭林 武汉大学人民医院放射科,武汉 430060

查云飞* 武汉大学人民医院放射科,武汉 430060

曾菲菲 武汉大学人民医院放射科,武汉 430060

柳柏玉 武汉大学人民医院放射科,武汉 430060

闫玉辰 武汉大学人民医院放射科,武汉 430060

通信作者:查云飞,E-mail:zhayunfei999@126.com

利益冲突:无。


基金项目: 国家自然科学基金面上项目 编号: 81871332
收稿日期:2019-09-17
接受日期:2019-11-21
中图分类号:R445.2; R735.9 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2020.03.008
本文引用格式:彭林,查云飞,曾菲菲,等.基于T2WI直方图纹理分析在胰腺实性病变中的诊断价值.磁共振成像, 2020, 11(3): 201-206. DOI:10.12015/issn.1674-8034.2020.03.008.

       胰腺来源实性病变是一组包含不同病理类型的疾病,主要包括恶性肿瘤如导管腺癌(pancreatic ductal adenocarcinoma,PDAC),良性到低度恶性肿瘤如实性假乳头状瘤(solid-pseudopapillary tumor of pancreas,SPT)、神经内分泌肿瘤(pancreatic neuroendocrine tumor,pNET)及非肿瘤性病变如肿块型慢性胰腺炎(mass-forming type chronic pancreatitis,MFCP)等,这些不同病理类型病变在一定情况下均可表现为胰腺实性占位,其临床特征及影像学表现相似,传统CT和MRI在胰腺病变的定性及良、恶性鉴别上仍存在一定难度,但不同病理类型治疗方式及患者的预后又截然不同[1]。纹理分析[2]作为一种新兴图像分析技术,对病灶图像像素灰度值变化规律及其分布模式进行研究,可以量化不被肉眼识别的变化,从而反映病变的异质性[3,4]。基于CT图像纹理分析量化胰腺囊腺瘤影像表型[5],不典型胰腺实性假乳头状肿瘤与胰腺导管腺癌[6]等,显示了纹理分析在胰腺疾病诊断中的可行性,但基于MRI图像的纹理分析对胰腺实性病变的鉴别诊断鲜有报道。本研究旨在探索基于MRI T2WI灰度直方图纹理特征在胰腺实性病变中的诊断价值。

1 材料与方法

1.1 临床资料

       回顾性分析2016年6月至2019年6月武汉大学人民医院经病理证实的117例胰腺肿块患者的临床及影像学资料。其中男66例,女51例,年龄26~ 77岁,平均(57.8±12.1)岁。病理类型为PDAC 69例、SPT 12例及pNET 15例、MFCP 21例。患者均以"腹痛、黄疸、消瘦"收入院。纳入标准:(1)手术病理明确诊断;(2)影像证实为胰腺实性占位,且影像资料完善,图像质量清晰。

1.2 检查方法

       采用3.0 T磁共振扫描仪(Discovery MR750 Plus,GE Healthcare,USA),8通道相控阵线圈进行检查。患者检查前禁食、禁水6 h以上,并接受自由呼吸及屏气训练。扫描时患者仰卧位,行胰腺常规MRI检查包括定位像、轴位压脂T2WI、轴位T1WI、冠状位T2WI序列,其中轴位T2WI序列采用带呼吸门控脂肪抑制(fat suppression,FS)技术,扫描参数为TE 79 ms,TR 12857 ms,FOV 40 cm×40 cm,层厚5 mm,层间距1 mm,回波链为28;动态增强扫描序列采用轴位肝脏容积加速采集(liver acquisition with volume acceleration,LAVA)序列,对比剂采用欧乃影(钆双胺,Gd-DTPA-BMA,GE Healthcare,Ireland)(0.2 mmol/kg),注射流率2 mL/s,注射完后用10~15 mL生理盐水以同流率注射冲洗。扫描图像推送至PACS工作站。

1.3 图像分析

       将纳入研究的患者所有MRI图像从PACS工作站以DICOM格式导出并储存。采用GE Omni-Kinetics软件进行图像分析后处理。由2名具有3年和7年影像诊断经验的医师评价图像质量,在T2WI图像上确定病灶最大层面,由1名医师沿病灶外缘勾画,注意避开血管影,ROI勾画结果如图1。由软件自动提取病变的纹理特征参数,生成基于灰度直方图纹理参数,包括均值(mean)、变异度(variance)、峰度(kurtosis)、偏度(skewness)、能量(energy)、熵(entropy)、第5百分位数、第10百分位数、第25百分位数、第50百分位数、第75百分位数、第90百分位数及第95百分位数这13个参数,并保存相应的数据文件。

图1  49岁,男性患者,皮肤巩膜黄染1个月余。A:示胰头区软组织肿块影,T2WI呈稍高信号,环绕胆总管生长,胆总管上段扩张,病理结果为胰腺导管腺癌;B:为病灶区ROI勾画
图2  A~I为发生在胰腺实性占位性病变轴位T2WI及病理图像。图A、D、G依次显示为导管腺癌、神经内分泌瘤及慢性肿块型胰腺炎;图B、C,图E、F及图H、I分别对应其病理结果(HE ×100)
Fig. 1  A 49-year-old male patient with yellow sclera skin for more than 1 month. A: Soft tissue mass shadow in the head of the pancreas. T2WI showed a slightly high signal, growing around the common bile duct and dilating in the upper part of the common bile duct. B: Outline the ROI of the focus area.
Fig. 2  A- I shows axial T2WI and pathological images of solid lesions in the pancreas. A, D and G are ductal adenocarcinoma, neuroendocrine tumor and chronic mass pancreatitis, respectively. B, C, E, F, H and I correspond to their pathological results, respectively (HE ×100).

1.4 统计学方法

       采用统计学软件(version 23.0,SPSS Inc.,Chicago,IL,USA)进行数据统计学分析。(1)首先对所有数据进行正态分布检验,符合正态分布的计量资料以(±s)表示,统计学处理PDAC组、SPT+pNET组、MFCP组定量纹理参数采用单因素方差分析(ANOVA),组间两两比较采用LSD-t检验(方差齐)或多个独立样本Kruskal-Walls检验(方差不齐);不符合正态分布的计量资料以M (1/4,3/4)表示,统计学处理采用秩和检验。计数资料以百分比表示,组间比较采用Fisher's确切概率检验;(2)分别构建PDAC组、SPT+pNET组、MFCP组定量纹理参数之间受试者工作特征曲线(receiver operation characteristic curves,ROC),计算曲线下面积(area under curve,AUC),比较纹理参数诊断效能。所有统计结果以P<0.05表示差异有统计学意义。

2 结果

2.1 临床资料及MRI影像表现

       不同病理组别患者的年龄、男性比例、胰头发生率、病灶最大直径、胰胆管扩张、胰尾萎缩及肾前筋膜增厚之间差异无统计学意义(P>0.05)。患者基本信息及胰腺实性肿块MRI表现见表1图2

表1  胰腺实性肿块患者基本信息及MRI表现
Tab. 1  Basic information and MRI findings of patients with solid pancreatic mass

2.2 纹理参数分析结果

       均值、第50百分数、第75百分数、第90百分数在三组之间差异均有统计学意义(P<0.05)。而偏度、峰度在三组之间差异无统计学意义。MFCP组与PDAC组之间均值、变异度、第5百分数、第10百分数、第25百分数、第50百分数、第75百分数、第90百分数及第95百分数差异均有统计学意义(P<0.05);MFCP组与SPT+pNET组之间均值、能量、熵、第5百分数、第10百分数、第25百分数、第50百分数、第75百分数、第90百分数及第95百分数差异均有统计学意义(P < 0.05);PDAC组与SPT+pNET组之间均值、能量、熵、第50百分数、第75百分数、第90百分数差异均有统计学意义(P<0.05);其中PDAC组均值最大、MFCP均值最小;而PDAC的第50百分数、第75百分数、第90百分数高于MFCP组、低于SPT+pNET组。三组胰腺实性病变灰度直方图纹理参数结果见表2

表2  三组胰腺实性病变灰度直方图纹理参数差异比较(±s)
Tab. 2  Comparison of the gray-scale histogram texture parameters of pancreatic solid lesions between the three groups (±s)

2.3  ROC判断诊断效能

       ROC曲线分析有统计学差异直方图参数在3组病变间两两比较的鉴别诊断效能见表3,表4,表5图3。在PDAC与MFCP之间变异度的鉴别诊断效果最好,AUC值为0.899;PDAC与SPT+pNET之间均值的鉴别诊断效果最好,AUC值为0.688;MFCP与SPT+pNET之间第90百分数的鉴别诊断效果最好,AUC值为0.924。

图3  A:PDAC与MFCP间的ROC曲线;B:PDAC与SPT+pNET间的ROC曲线;C:MFCP与SPT+pNET间的ROC曲线
Fig. 3  A: ROC curve between PDAC and MFCPB. B: ROC curve between PDAC and SPT+pNET. C: ROC curve between MFCP and SPT+pNET.
表3  PDAC与MFCP间鉴别效能
Tab. 3  Differential performance between PDAC and MFCP
表4  PDAC与SPT+pNET间鉴别效能
Tab. 4  Differential performance between PDAC and SPT+pNET
表5  MFCP与SPT+pNET间鉴别效能
Tab. 5  Differential performance between MFCP and SPT+pNET

3 讨论

       由于T2WI序列回波时间相对较长,组织间的对比度增加,图像中蕴含更多具有鉴别价值的差异性纹理特征,使得基于T2WI图像纹理分析成为可能。目前国内外胰腺病变纹理分析研究主要包括胰腺病变定性如鉴别胰腺神经内分泌肿瘤与胰腺肾细胞癌转移瘤[7]、胰腺淋巴瘤与胰腺癌[8]、非典型pNET与PDAC[9,10]、胰腺浆液性囊腺瘤与黏液性囊性肿瘤[11,12]以及病变病理分级如pNET病理分级[13,14,15](尤其G3期敏感)和评估病变风险及预测预后等[16,17]。本研究结果显示,均值、变异度、能量、熵、第5百分数、第10百分数、第25百分数、第50百分数、第75百分数、第90百分数及第95百分数在三组胰腺实性病变之间的差异有统计学意义。其中均值、百分位数鉴别三组病变的诊断效能良好。

       均值描述病变特征值的平均大小,PDAC组、SPT+pNET组与FMCP组间均值有统计学差异,PDAC组稍大,SPT+pNET组次之,说明PDAC组T2WI信号强度高于SPT+pNET组与FMCP组。这可能是因为PDAC主要由异型细胞形成不规则或不完整的管状或腺样结构,伴有丰富的纤维间质,细胞间排列更密集[18],所以PDAC组T2WI信号强度更高。

       变异度描述病变特征值的平均值分散程度,变异度越大,数据越偏离平均值,表示病变不均匀。由于本次研究SPT病例相对较少,笔者将SPT及pNET合并成一组,定义为良性到低度恶性组,这可能是该组的变异度最大的主要原因。FMCP为局部炎性肿大并形成肿块,成分相对PDAC更简单、均匀,因此FMCP的变异度较PDAC更小。

       能量描述图像像素值分布的均匀程度。熵描述图像直方图像素值分布的均匀性,熵越大代表肿瘤异质性越高。本研究中,能量、熵在SPT+pNET组分别与PDAC组、FMCP组间有统计学差异,且诊断效能良好,但在PDAC组、FMCP组之间无统计学差异。这可能是因为FMCP是PDAC的癌前病变表现,两者可共同存在,FMCP在修复过程中可能出现不典型增生及局限性癌变,使处于一定病程阶段的部分慢性胰腺炎患者合并胰腺癌。

       偏度描述直方图变量分布对称性的一类统计量。峰度描述变量分布形状陡缓度的统计量。朱晨迪等[19]发现T2WI全域灰度直方图分析可作为鉴别儿童常见的3种后颅窝肿瘤的重要手段,在室管膜瘤与星形细胞瘤之间偏度、星形细胞瘤与髓母细胞瘤之间变异度和峰度、室管膜瘤与髓母细胞瘤之间峰度具有较高诊断效能。而在本次研究只纳入胰腺来源实性病变,囊变、坏死、出血等病例剔除在外,这可能是造成三组间的偏度、峰度无统计学差异的重要原因。百分位数描述低于该百分位数的所观察对象的百分比,反映直方图的体素值[20]。本研究结果中百分位数的曲线下面积普遍偏高,诊断效能较好。

       但本研究仍存在一定局限性:(1)本研究为回顾性研究,纳入的SPT、pNET、MFCP病例数较少,尚需进一步扩大样本量深入分析;(2) ROI并未包含全部病变范围,因此纹理信息的提取可能不够全面;(3)未探讨同一肿瘤不同病理分级的差异性,这有待今后进一步研究。

       总之,不同胰腺实性病变在T2WI纹理特征存在明显差异,其中均值、百分位数诊断效能较高,对于胰腺实性占位性病变的鉴别诊断有一定临床应用价值。

[1]
Chantarojanasiri T, Kongkam P. Endoscopic ultrasound elastography for solid pancreatic lesions. World J Gastrointest Endosc, 2017, 9(10): 506-513.
[2]
Lambin P, Rios-Velazquez E, Leijenaar R, et al. Radiomics: extracting more information from medical images using advanced feature analysis. Eur J Cancer, 2012, 48(4): 441-446.
[3]
Yu Y, Ding YQ, Hua C, et al. Value of multi-phase enhanced MRI combined with histogram analysis of apparent diffusion coefficient in differential diagnosis of intrahepatic biliary cystadenoma and intrahepatic biliary cystadenocarcinoma. Chin J Radiol, 2018, 52(6): 442-446.
于泳,丁玉芹,花村,等.多期增强MRI和表观扩散系数直方图鉴别肝内胆管囊腺瘤及囊腺癌的价值.中华放射学杂志, 2018, 52(6): 442-446.
[4]
Zhang ZW, Hua T, Xu TT, et al. Differentiation of benign and malignant breast lesions using texture analysis of conventional MRI: a preliminary study. Chin J Radiol, 2017, 51(8): 588-591.
张竹伟,华婷,徐婷婷等.常规MRI纹理分析鉴别乳腺良、恶性病变的价值初探.中华放射学杂志, 2017, 51(8): 588-591.
[5]
Hu ZY, Shen QJ, Feng Z, et al. Diagnostic value of CT texture analysis in the quantification of radiological phenotype for pancreatic cystadenoma. Chin J Radiol, 2017, 17(5): 330-334.
胡征宇,沈起钧,冯湛,等. CT纹理分析在量化胰腺囊腺瘤影像表型中的诊断价值.中华胰腺病杂志, 2017, 17(5): 330-334.
[6]
Huang ZX, Li M, Yu HP, et al. Texture analysis based on CT images to differentiate atypical pancreatic solid pseudopapillary tumor and pancreatic adenocarcinoma: a preliminary study. Chin J Bases Clin General Surg, 2018, 25(10): 1249-1253.
黄子星,李谋,于浩鹏,等. CT图像纹理分析鉴别不典型胰腺实性假乳头状肿瘤与胰腺导管腺癌的初步研究.中国普外基础与临床杂志, 2018, 25(10): 1249-1253.
[7]
van der Pol CB, Lee S, Tsai S, et al. Differentiation of pancreatic neuroendocrine tumors from pancreas renal cell carcinoma metastases on CT using qualitative and quantitative features. Abdominal Radiology, 2019, 44(3): 992-999.
[8]
Huang Z, Li M, He D, et al. Two-dimensional texture analysis based on CT images to differentiate pancreatic lymphoma and pancreatic adenocarcinoma: a preliminary study. Acad Radiol, 2019, e189-e195.
[9]
Li J, Lu J, Liang P, et al. Differentiation of atypical pancreatic neuroendocrine tumors from pancreatic ductal adenocarcinomas: using whole-tumor CT texture analysis as quantitative biomarkers. Cancer Med, 2018, 7(10): 4924-4931.
[10]
Guo C, Zhuge X, Wang Q, et al. The differentiation of pancreatic neuroendocrine carcinoma from pancreatic ductal adenocarcinoma: the values of CT imaging features and texture analysis. Cancer Imaging, 2018, 18(1): 37.
[11]
Liang M, Zhao L, Ma XH, et al. CT texture analysis of cyst fluid in differential diagnosis between pancreatic serous cystadenoma and mucinous cystic neoplasms. J Pract Radiol, 2018, 34(9): 1381-1385.
梁萌,赵丽,马霄虹,等.囊液CT纹理分析在鉴别胰腺浆液性囊腺瘤与黏液性囊性肿瘤中的价值.实用放射学杂志, 2018, 34(9): 1381-1385.
[12]
Wang BT, He L, Liu G, et al. Value of magnetic resonance imaging texture feature analysis in the differential diagnosis between pancreatic serous cystadenoma and mucinous cystadenoma. Acta Acad Med Sinicae, 2018, 40(2): 187-193.
王波涛,何蕾,刘刚,等.磁共振成像纹理特征分析在胰腺浆液性囊腺瘤及黏液性囊腺瘤鉴别诊断中的价值.中国医学科学院学报, 2018, 40(2): 187-193.
[13]
Guo C, Zhuge X, Wang Z, et al. Textural analysis on contrast-enhanced CT in pancreatic neuroendocrine neoplasms: association with WHO grade. Abdominal Radiology, 2019, 44(2): 576-585.
[14]
D Onofrio M, Ciaravino V, Cardobi N, et al. CT enhancement and 3D texture analysis of pancreatic neuroendocrine neoplasms. Sci Rep, 2019, 9(1): 2176.
[15]
Choi TW, Kim JH, Yu MH, et al. Pancreatic neuroendocrine tumor: prediction of the tumor grade using CT findings and computerized texture analysis. Acta Radiologica, 2017, 59(4): 383-392.
[16]
Choi MH, Lee YJ, Yoon SB, et al. MRI of pancreatic ductal adenocarcinoma: texture analysis of T2-weighted images for predicting long-term outcome. Abdominal Radiology, 2019, 44(1): 122-130.
[17]
Attiyeh MA, Chakraborty J, Gazit L, et al. Preoperative risk prediction for intraductal papillary mucinous neoplasms by quantitative CT image analysis. HPB (Oxford), 2019, 21(2): 212-218.
[18]
Numata K, Ozawa Y, Kobayashi N, et al. Contrast-enhanced sonography of pancreatic carcinoma: correlations with pathological findings. J Gastroenterol, 2005, 40(6): 631-640.
[19]
Zhu CD, Zhang Y, Cheng JL, et al. Value of MRI histogram analysis in differential diagnosis of pediatric posteriorfossa tumor. Radiol Pract, 2018, 33(3): 285-289.
朱晨迪,张勇,程敬亮,等. MRI灰度直方图分析在儿童后颅窝常见肿瘤中的鉴别诊断价值.放射学实践, 2018, 33(3): 285-289.
[20]
Lu SS, Kim SJ, Kim N, et al. Histogram analysis of apparent diffusion coefficient maps for differentiating primary CNS lymphomas from tumefactive demyelinating lesions. AJR Am J Roentgenol, 2015, 204(4): 827-834.

上一篇 基于非增强MRI的影像组学术前预测肝细胞癌微血管浸润的研究
下一篇 磁共振对剖宫产瘢痕妊娠的诊断及临床价值
  
诚聘英才 | 广告合作 | 免责声明 | 版权声明
联系电话:010-67113815
京ICP备19028836号-2