基于全偏振成像和机器学习的肝细胞癌与肝内胆管癌鉴别诊断研究

Differential diagnosis of hepatocellular carcinoma and cholangiocarcinoma based on Mueller Matrix Polarimetry and machine learning

目的探讨全偏振成像和机器学习在肝细胞癌与肝内胆管癌鉴别诊断中的作用。方法利用全偏振显微镜对术后经病理确诊为低分化肝细胞癌和低分化肝内胆管癌病例各8例进行偏振成像;病理医生根据HE切片选取3块感兴趣区域,测量每块区域的穆勒矩阵,并根据穆勒矩阵参数提取分析方法计算出一系列的偏振基准参数;将偏振特征参数输入到人工神经网络模型中,采用了8折交叉验证方法对模型进行三分类的训练和验证。结果模型结果显示,基于图像偏振特征区分肝细胞癌、肝内胆管癌与除了癌变细胞以外的其他组织的准确率为0.8463,灵敏度为0.8107。结论基于全偏振成像和机器学习构建的肝细胞癌和肝内胆管癌诊断模型具有重要的病理辅助诊断价值。

Objective To investigate the role of total polarization imaging and machine learning in the differential diagnosis of hepatocellular carcinoma and cholangiocarcinoma.Methods Polarization imaging was performed on 8 cases of poorly differentiated hepatocellular carcinoma and 8 cases of poorly differentiated cholangiocarcinoma.Pathologists selected three Regions of interest(ROIs)according to HE slices,measured the mueller matrix polarimetry of each ROI,and calculated a series of polarization basic parameters according to the mueller matrix extraction analysis method.The polarization basic parameters were input into the artificial neural Network(ANN)model,and the 8-fold cross-validation method was used to train and verify the model in three categories.Results ANN model showed that the The precision and sensitive of differentiating hepatocellular carcinoma cells,Cholangiocarcinoma cells and other tissues based on mueller matrix polarimetry was 0.8463 and 0.8107.Conclusion The diagnostic model of hepatocellular carcinoma and intrahepatic cholangiocarcinoma based on polarized imaging and machine learning is of great value in pathological auxiliary diagnosis,which can provide help for clinical accurate diagnosis and treatment.