基于术前脾脏CT影像组学列线图预测胃癌的浆膜浸润

目的：联合临床检验指标及影像学特征构建一种能够术前识别胃癌浆膜浸润的模型。方法：选取2015年1月至2019年12月温州医科大学附属第一医院经病理证实的656例胃癌患者，采用随机数字表法分为建模组（394例）和验证组（262例）。收集建模组患者的脾脏影像学资料，对收集的数据进行套索回归并选取差异有统计学意义的特征来构建浆膜浸润预测模型。在最大约登指数下取肿瘤浸润风险评分截断值将患者分为高危组（238例）和低危组（418例），然后与其他浸润相关因素如BMI、年龄、性别、高血压、糖尿病等进行单变量和多变量Logistic回归分析，结合显著的独立影响因素共同建立可视化的浆膜浸润预测列线图。结果：将患者以肿瘤浸润评分≤-0.335分为低危组，＞-0.335为高危组，经验证组验证，建模组和验证组的诊断准确性较为一致（P＜0.001）。经浸润影响因素的单变量和多变量Logistic回归分析发现，影像组学肿瘤浸润评分（OR=2.9，95%CI=2.1～4.2，P＜0.001）、术前低白蛋白（OR=1.3，95%CI=1.2～3.1，P=0.003）、血小板与淋巴细胞比值（OR=1.8，95%CI=1.2～2.7，P=0.004）、肿瘤分化程度（OR=2.6，95%CI= 1.8～3.7，P＜0.001）是浆膜浸润的独立影响因素。基于这4个指标建立的预测模型能够较为准确地预测浆膜浸润风险，其AUC值为0.733。结论：基于脾脏影像的肿瘤浸润评分联合其他临床因素可准确预测胃癌浆膜浸润与否，提高诊断精度。

Establishment of the radiologic tumor invasion index based on radiomics splenic features and clinical factors to predict serous invasion of gastric cancer

Objective: To construct a model that can identify serosa infiltration in gastric cancer before surgery by combining clinical data and radiological features. Methods: A total 656 patients with pathologically confirmed gastric cancer were selected from the First Affiliated Hospital of Wenzhou Medical University between January 2015 and December 2019, who were randomly divided into the validation group (262 cases) and the modeling group (394 cases). The spleen imaging data of the patients in the modeling group were collected and analyzed by lasso regression, and the serosa infiltration prediction model was constructed on the selected significant features. Patients were divided into high-risk group (238 cases) and low-risk group (418 cases) by tumor infiltration risk score according to the cut-off value under the largest Yoden index, and then univariate and multivariate Logistic regression analysis was made with other infiltration-related factors to establish a visual prediction nomogram. Results: Patients with tumor infiltration score ≤-0.335 and > -0.335 were divided respectively into low-risk group and high-risk group. The diagnostic accuracy of the modeling and validation groups was consistent (P<0.001) when verified by the validation group (P<0.001). Univariate and multivariate Logistic analysis of infiltration risk factors showed that tumor radiomic infiltration score (OR=2.9, 95%CI=2.1-4.2, P<0.001), preoperative albumin (OR=1.3, 95%CI=1.2-3.1, P=0.003), platelet-lymphocyte ratio (OR=1.8, 95%CI=1.2-2.7, P=0.004) and tumor differentiation (OR=2.6, 95%CI=1.8-3.7, P<0.001) were independent influential factors for serosa invasion. The prediction model based on these four indicators accurately predicted the risk of serosa invasion, and its AUC value was 0.733. Conclusion: Tumor infiltration score based on spleen imaging combined with other clinical factors can accurately predict the serosal invasion of gastric cancer and improve the diagnostic precision.