影像组学在透明细胞型肝癌诊断中的价值  

The value of radiomics in the diagnosis of primary clear cell carcinoma of the liver

作　　者：刘斌 陈枫[2] 张岩岩[2] 朱海峰 李宏军[2] LIU Bin;CHEN Feng;ZHANG Yan-yan(Department of Radiology,Civil Aviation General Hospital,Beijing 100123,China)

机构地区：[1]民航总医院放射科,北京100123 [2]首都医科大学附属北京佑安医院放射科,北京100069

出　　处：《放射学实践》2025年第1期24-28,共5页Radiologic Practice

基　　金：国家自然科学基金项目(项目批准号:61936013)。

摘　　要：目的:通过比较不同模型的诊断效能评估影像组学诊断透明细胞型肝癌的价值。方法:42例透明细胞型肝癌和109例普通型肝细胞癌患者以2:1的比率随机分为训练集和验证集,使用训练集建立模型,使用验证集进行验证。结果:在训练集中,临床模型诊断透明细胞型肝癌的AUC值、敏感度、特异度和准确度分别为0.797、0.808、0.662和0.700,影像组学融合模型分别为0.878、1.000、0.703和0.780,组合模型分别为0.940、1.000、0.757和0.820;在验证集中,临床模型诊断透明细胞型肝癌的AUC值、敏感度、特异度和准确度分别为0.647、0.625、0.600和0.608,影像组学融合模型分别为0.812、0.688、0.857和0.804,组合模型分别为0.868、0.688、0.886和0.824。结论:影像组学对于诊断透明细胞型肝癌具有一定临床价值,影像组学评分是诊断透明细胞型肝癌的独立影响因素。Objective:The aim of this study was to assess the value of radiomics in diagnosing primary clear cell carcinoma of the liver by comparing the diagnostic efficacy of different models.Methods:We randomly assigned 42 cases of primary clear cell carcinoma of the liver(PCCCL)and 109 cases of common hepatocellular carcinoma(CHCC)into training and validation sets at a ratio of 2:1.The training set was used to establish different models while the validation set was used to evaluate the performance of these models.Results:In the training set,the clinical model demonstrated an AUC value of 0.797,sensitivity of 0.808,specificity of 0.662,and accuracy of 0.700 in diagnosing PCCCL.The imaging fusion model showed improved performance with an AUC value of 0.878,sensitivity of 1.000,specificity of 0.703,and accuracy of 0.780.Additionally,the combined model demonstrated the highest performance with an AUC value of 0.940,sensitivity of 1.000,specificity of 0.757,and accuracy of 0.820.In the validation set,the clinical model had an AUC value of 0.647,sensitivity of 0.625,specificity of 0.600,and accuracy of 0.608.The imaging fusion model had an AUC value of 0.812,sensitivity of 0.688,specificity of 0.857,and accuracy of 0.804.The combined model continued to perform well with an AUC value of 0.868,sensitivity of 0.688,specificity of 0.886,and accuracy of 0.824.Conclusion:Our findings suggest that radiomics has significant clinical value in accurately diagnosing PCCCL.Importantly,our analysis revealed that Radscore,a radiomics-based score,is an independent factor in distinguishing between PCCCL and CHCC,demonstrating its potential for use in clinical practice.

关 键 词：透明细胞型肝癌 普通型肝细胞癌 影像组学 体层摄影术 X线计算机 

分 类 号：R735.7[医药卫生—肿瘤] R814.42[医药卫生—临床医学]