摘要
目的探讨基于扩散峰度成像(DKI)的全肿瘤直方图分析术前预测肝细胞癌(HCC)病理分化程度的价值。资料与方法回顾性收集经手术病理证实的低分化HCC 19例,非低分化HCC 33例,经后处理生成各向异性分数(FA)、平均扩散系数(MD)及平均扩散峰度(MK)图,并导入ITK-SNAP软件。在包含肿瘤的层面逐一勾画感兴趣区(ROI),自动生成覆盖全肿瘤的直方图参数,包括均值、25、50、75、90百分位数、偏度和峰度,比较两组各图信号强度直方图各参数的差异,采用受试者工作特征(ROC)曲线分析其诊断效能。结果低分化组FA信号强度均值、25、50、75、90百分位数和MD信号强度均值、50、75、90百分位数小于非低分化组(t/Z=-2.632、-2.385、-2.632、-2.923、-3.024、-2.310、-2.314、-2.686、-2.467,P均<0.05),FA信号强度峰度和MK信号强度均值、50、75、90百分位数大于非低分化组(t/Z=-2.404、-2.252、-2.366、-2.347、-2.252,P均<0.05)。FA和MD信号强度75百分位数、MK信号强度50百分位数预测低分化HCC与非低分化HCC的效能最佳,曲线下面积分别为0.742、0.703及0.699。结论基于DKI的全肿瘤直方图分析在术前预测HCC病理分化程度中有一定的价值。
Purpose To explore the value of preoperative prediction of pathological differentiation of hepatocellular carcinoma(HCC)based on whole-tumor histogram analysis derived from diffusion kurtosis imaging(DKI).Materials and Methods All patients with HCC confirmed by pathology,including the poorly differentiated HCC group(19 cases)and the non-poorly differentiated HCC group(33 cases),were retrospectively enrolled.The preprocessed data of fractional anisotropy(FA),mean diffusion coefficient(MD)and mean kurtosis coefficient(MK)maps were transferred to ITK-SNAP software.All regions of interest(ROIs)covering the whole tumor were drawn slice by slice of FA,MD and MK intensity maps,generating automatically histogram parameters and obtaining the mean value,the 25 th,50 th,75 th and 90 th percentiles,skewness and kurtosis,etc.All parameters were compared to get significant differences between these two groups and receiver operating characteristic curves(ROCs)were performed to evaluate the diagnostic efficiency.Results FA signal intensity(mean value,the 25 th,50 th,75 th,90 th percentiles)and MD signal intensity(mean value,the 50 th,75 th,90 th percentiles)of the poorly differentiated HCC group were significantly lower than those of the non-poorly differentiated HCC group(t/Z=-2.632,-2.385,-2.632,-2.923,-3.024,-2.310,-2.314,-2.686 and-2.467,P<0.05).The kurtosis of FA signal intensity and MK signal intensity(mean value,the 50 th,75 th,90 th percentiles)of the poorly differentiated HCC group were significantly higher than those of the non-poorly differentiated HCC group(t/Z=-2.404,-2.252,-2.366,-2.347 and-2.252,P<0.05).The area under curves(AUC)of the 75 th of FA,75 th of MD and 50 th of MK signal intensity were presented the highest prediction of HCC pathological differentiation degrees,with 0.742,0.703 and 0.699,respectively.Conclusion DKI-based whole-tumor histogram analysis could assist in preoperative prediction of pathological differentiation degrees in patients with HCC.
作者
林涛
赵莹
田士峰
宋清伟
郭妍
刘爱连
LIN Tao;ZHAO Ying;TIAN Shifeng;SONG Qingwei;GUO Yan;LIU Ailian(Department of Radiology,the First Affiliated Hospital of Dalian Medical University,Dalian 116011,China;不详)
出处
《中国医学影像学杂志》
CSCD
北大核心
2021年第7期691-696,共6页
Chinese Journal of Medical Imaging
基金
国家自然科学基金面上项目(61971091)
首都科技领军人才培养工程(Z181100006318003)。
关键词
肝肿瘤
癌
肝细胞
磁共振成像
扩散峰度成像
直方图分析
病理学
外科
诊断
鉴别
Liver neoplasms
Carcinoma,hepatocellular
Magnetic resonance imaging
Diffusion kurtosis imaging
Histogram analysis
Pathology,surgical
Diagnosis,differential
