CT图像对食管鳞癌术后营养指标评估的模型构建

Model Construction of CT Image for Evaluation of Postoperative Nutritional Indicators of Esophageal Squamous Cell Carcinoma

摘要: 目的：探究基于CT图像的营养指标在评估食管鳞癌术后患者预后中的价值，并建立一个列线图模型。方法：纳入了67例2016年10月至2020年12月于青岛市市立医院胸外科接受食管切除术的患者的临床数据，并将队列以2:1分为训练集和验证集。构建基于CT图像的PPR指标，PPR为术后T12肌肉指数与术前T12肌肉指数的比值。通过单因素和多因素COX回归分析影响食管鳞癌术后患者生存的独立预后因素。构建新营养评估工具的列线图模型，比较新模型与传统营养评估工具的优劣，并对该模型进行区分度、校准度及临床适用度分析。结果：单因素及多因素COX回归分析提示，TNM分期(P = 0.0427)和PPR (P = 0.0078)是食管鳞癌患者的独立预后影响因素。构建新的TNM-PPR评估模型。KM生存曲线显示该模型能有效地预测食管癌术后患者的预后(Log-rank P < 0.01)。该评估工具的列线图的C指数为0.835 (95% CI: 0.752, 0.919)。内部、外部验证的校准曲线表明TNM-PPR模型与实际情况有较好的拟合效果。ROC曲线及DCA曲线显示，此模型相比传统营养评估工具有更好的区分度和临床适用性。结论：本研究构建的基于CT图像的TNM-PPR预后模型对食管鳞癌术后患者的预后有较好的预测能力。

Abstract: Objective: To explore the value of nutritional indicators based on CT images in evaluating the prog-nosis of postoperative patients with esophageal squamous cell carcinoma, and a nomogram model was established. Methods: The clinical data of 67 patients who underwent esophagectomy in the Department of Thoracic Surgery of Qingdao Municipal Hospital from October 2016 to December 2020 were included, and the queue is divided into training set and verification set at the proportion of 2:1. The PPR index was constructed based on CT images. PPR was the ratio of postoperative T12 muscle index to preoperative T12 muscle index. The independent prognostic factors affecting the survival of postoperative patients with esophageal squamous cell carcinoma were analyzed by uni-variate and multivariate COX regression. The nomogram model of the new nutrition assessment tool was constructed, and the advantages and disadvantages of the new model and the traditional nutri-tion assessment tool were compared, and then the differentiation, calibration and clinical applica-bility of the model were analyzed. Results: Univariate and multivariate COX regression analysis showed that TNM stage (P = 0.0427) and PPR (P = 0.0078) were independent prognostic factors in patients with esophageal squamous cell carcinoma. A new TNM-PPR scoring model was constructed. KM survival curve showed that the model could effectively predict the prognosis of postoperative patients with esophageal cancer (Log-rank P < 0.01). The C-index of the nomogram of the evaluation tool was 0.835 (95% CI: 0.752, 0.919). The calibration curves verified internally and externally showed that the TNM-PPR model has a good fitting effect with the actual situation. ROC curve and DCA curve showed that this model has better differentiation and clinical applicability than tradi-tional nutrition assessment tools. Conclusions: The TNM-PPR prognostic model based on CT image constructed in this study can predict the prognosis of postoperative patients with esophageal squamous cell carcinoma.