目的:探究超分辨率重建影像组学模型诊断原发性肝癌中医证型的效能。方法:回顾性选取原发性肝癌患者128例，按照8:2随机分为训练组和测试组，所有病例均进行MRI扫描;对原发性肝癌患者的临床资料和影像征象进行单因素分析;分别在超分辨率重建前后的磁共振弥散加权成像(DWI)图像上半自动勾画感兴趣区，然后采用最小冗余最大相关，K最佳(Select K Best)和LASSO回归降维并构建模型，绘制受试者工作特征(ROC)曲线和决策曲线评估模型的性能。结果:原发性肝癌不同证型间临床资料和影像征象比较差异均无统计学意义(P＞0。05);肝郁脾虚证模型NR在测试组中的曲线下面积(AUC)值、敏感度、特异度、准确度分别为0。585、58。3％、46。7％、53。0％，模型SR在测试组中的AUC值、敏感度、特异度、准确度分别为0。639、50。0％、69。4％、64。6％;气滞血瘀证模型NR在测试组中的AUC值、敏感度、特异度、准确度分别为0。608、66。7％、43。3％、56。1％，模型SR在测试组中的AUC值、敏感度、特异度、准确度分别为0。644、53。3％、63。6％、61。0％;肝肾阴虚证模型NR在测试组中的AUC值、敏感度、特异度、准确度分别为0。612、47。2％、63。3％、54。5％，模型SR在测试组中的AUC值、敏感度、特异度、准确度分别为0。644、60。0％、61。4％、61。0％;决策曲线显示模型SR有更高的净收益。结论:与模型NR相比，模型SR在判断原发性肝癌中医证型方面有良好的预测性能。
Study on the Diagnostic Efficacy of Super Resolution Reconstructed Image Omics Model in Diagnosis of Traditional Chinese Medicine Syndromes of Primary Liver Cancer
Objective:To explore the efficacy of super resolution reconstructed image omics model in the diagnosis of traditional Chinese medicine(TCM)syndromes of primary liver cancer.Methods:A total of 128 patients with primary liver cancer were retrospectively collected and randomly divided into the training group and the test group according to ratio of 8:2.All cases underwent MRI scanning.The clinical data and imaging signs of patients with primary liver cancer were analyzed by single factor analysis.The regions of interest were mapped semi-automatically on diffusion weighted imaging(DWI)images before and after super resolution reconstruction.Then,the minimum redundancy and maximum correlation were used to reduce dimension and construct the model by Select K Best and LASSO regression.The receiver operating characteristic(ROC)curve and decision curve were drawn to evaluate the performance of the model.Results:There were no significant differences in clinical data and imaging signs among different types of primary liver cancer(P＞0.05).The AUC value,sensitivity,specificity and accuracy of liver depression and spleen deficiency syndrome model NR in the test group were 0.585,58.3％,46.7％and 53.0％,respectively,and the AUC value,sensitivity,specificity and accuracy of model SR in the test group were 0.639,50.0％,69.4％and 64.6％,respectively.The AUC value,sensitivity,specificity and accuracy of qi-stagnation and blood-stasis syndrome model NR in the test group were 0.608,66.7％,43.3％and 56.1％,respectively,and the AUC value,sensitivity,specificity and accuracy of model SR in the test group were 0.644,53.3％,63.6％and 61.0％,respectively.The AUC value,sensitivity,specificity and accuracy of liver-kidney yin deficiency syndrome model NR in the test group were 0.612,47.2％,63.3％and 54.5％,respectively,and the AUC value,sensitivity,specificity and accuracy of model SR in the test group were 0.644,60.0％,61.4％and 61.0％,respectively.The decision curve showed that model SR had higher net income.Conclusion:Compared with model NR,model SR has good predictive performance in judging TCM syndrome type of primary liver cancer.
super resolution reconstructionimage omicsprimary liver cancertraditional Chinese medicine syndromesliver depression and spleen deficiency syndromeqi-stagnation and blood-stasis syndromeliver-kidney yin deficiency syndrome
万方数据
维普
