能谱单能量图像影像组学在胃癌分级中的应用

Application Value of Radiomics Features Based on Monochromatic Images of Spectrum CT in the Pathological Grading of Gastric Adenocarcinoma

目的:探讨基于单源双能CT单能量图像的影像组学特征评估胃腺癌病理分级的临床价值。方法:回顾性收集使用Discovery 750 CT机行上腹或全腹GSI扫描的胃腺癌患者196例,根据病理结果分为低分化组和中高分化组。在AW4.6工作站分析获得高质量、标准化的静脉期70keV单能量图像,提取基于该图像的影像组学特征,并利用降维后的特征建立Logistic回归模型。采用受试者工作特性曲线(ROC)评价Logistic回归模型的分级效能并验证其准确率。同时,对比分析常规能谱参数的分级结果及效能。结果:常规能谱参数标准化碘(水)浓度的分级效能是0.668;Logistic模型鉴别两组的ROC曲线下面积为0.872(敏感度为73.3%、特异度为83.3%),诊断准确率为78.3%。结论:能谱单能量图像的影像组学特征,可无创鉴别低分化与中高分化胃腺癌,且效能优于常规能谱参数,为临床治疗方案选择和术前评估提供更多定量、可重复的数据信息。

Purpose： To evaluate the clinical value of radiomics features based on monochromatic images of spectrum CT in the pathological grading of gastric adenoearcinoma. Methods： One hundred and ninety six patients with gastric adenocarcinoma who underwent upper or total abdominal GSI scanning using the Discovery 750 CT machine were collected. They were divided into poorly differentiated group and moderate-well differentiated group according to pathological results. High-quality, standardized venous phase single-energy images of 70 keV were obtained from the AW4.6 workstation, and then radiomics features based on the monochromatic images were extracted and a logistic regression model was established by the dimensionality reduced features. The efficacy of the logistic regression model was evaluated by the receiver operating characteristic curve （ROC） and its accuracy was verified. At the same time, the grading results and performance of conventional energy spectrum parameters were compared and analyzed. Results： The grading efficacy of the normalized iodine （water） concentration of conventional spectral parameters was 0.668. The area under the ROC curve （AUC） of the logistic regression model was 0.872 （sensitivity was 73.3% and specificity was 83.3%） and the diagnostic accuracy was 78.3%. Conclusion： Radiomics features based on the single-source dual-energy CT monochromatic images can be used to differentiate gastric adenocarcinoma from poorly differentiated and moderate-well differentiated tumors non-invasively. Its efficacy is better than that of conventional spectral CT, and it can provide more quantitative and repeatable information for clinical treatment options and preoperative evaluations.