超声图像偏倚式干涉技术的建立及其在肝内占位病变中的应用  

Establishment of Bias Interference and Application in Space Occupying Lesions of Liver

作　　者：董振宇 刘镭 刘力 张玲利 叶然 谭莹 DONG Zhenyu;LIU Lei;LIU Li;ZHANG Lingli;YE Ran;TAN Ying(Department of Ultrasound,the University of Hong Kong-Shenzhen Hospital,Shenzhen 518000,China)

机构地区：[1]香港大学深圳医院超声科

出　　处：《中国医学影像学杂志》2019年第10期788-791,共4页Chinese Journal of Medical Imaging

基　　金：深圳市卫生计生系统科研项目（SZFZ2017090）

摘　　要：目的探讨基于病理大体标本的超声图像偏倚式干涉(BI)程序的建立及其在肝内占位病变性质中的鉴别诊断价值。资料与方法回顾性分析112例肝内发现占位病变患者的临床资料,并将病变随机分为建模组(用于建立干涉技术)和验证组(用于验证对占位性质的鉴别)。利用超声图像及对应病理大体标本图像获取中点、病灶中心至边缘长度25%、50%、75%病灶及边缘图像的灰阶数据进行相关性分析。利用分析结果对其他占位行自动化分析,并对病理与BI结果、病理与人工检查结果行一致性分析。结果建模组超声及病理多点取样良恶性图像差异有统计学意义(P<0.05)。超声及对应病理大体标本图像中点以及病灶中心至边缘长度25%、50%、75%病灶和边缘图像的灰阶值具有相关性(r=-0.600、0.287、-0.650、-0.785、-0.785,P<0.05)。BI与病理及人工检查结果一致性较好(Kappa=0.789、0.596,P<0.05),识别准确率为91.7%,与人工检查准确率比较,差异无统计学意义(χ^2=3.214,P=0.073)。结论病理结合超声图像定量分析可鉴别肝内结节良恶性。自动化分析与人工检查结果具有较好的一致性。Purpose To explore the establishment of bias interference(BI)procedure based on pathological gross specimens and its application value in the identification of space occupying nature of liver.Materials and Methods The clinical data of 112 patients with space occupying lesions of liver was analyzed retrospectively,and the lesions were randomly divided into a modeling group(for establishing an interference technique)and a verification group(for verification of the identification of space occupying nature).Correlation analysis was performed using the grayscale data of ultrasound images and corresponding pathological gross specimen image acquisition midpoint,and lesion center to edge length 25%,50%,75% lesion and edge images.The analysis results were used for automatic analysis of the remaining space occupying,and the consistency analysis of pathology and BI results,pathology and manual examination results was performed.Results There were significant differences between benign and malignant images on ultrasound and pathological multi-point sampling in the modeling group(P<0.05).The grayscale values of ultrasound images and corresponding pathological gross specimen image acquisition midpoint,and lesion center to edge length 25%,50%,75% lesion and edge images showed correlation(r=-0.600,0.287,-0.650,-0.785,-0.785,P<0.05).There was good consistencies between BI results and pathological results as well as manual examination results(Kappa=0.789,0.596,P=0.000).With the accuracy rate of 91.7%,and the difference in accuracy rate between BI and manual examination was not statistically significant(χ^2=3.214,P=0.073).Conclusion Quantification analysis combining the pathology and ultrasound images can distinguish the benign and malignant lesions of the intrahepatic nodules.The accuracy of automatic analysis results is similar to that of manual examination results.

关 键 词：肝疾病 肝肿瘤 超声检查 图像处理 计算机辅助 算法 病理学 外科 诊断 鉴别 

分 类 号：R735.7[医药卫生—肿瘤] R445.1[医药卫生—临床医学]