METHODS: A retrospective analysis of 408 GC patients from two institutions: 288 patients from Institution I were divided 7:3 into a training set (n = 203) and a testing set (n = 85); 120 patients from Institution II served as an external validation set. Radiomics features were extracted and screened from CT images. Independent radiomics, clinical, and combined models were constructed to predict PNI. Model discrimination, calibration, clinical utility, and prognostic significance were evaluated using area under the curve (AUC), calibration curves, decision curves analysis, and Kaplan-Meier curves, respectively.
RESULTS: 15 radiomics features and three clinical factors were included in the final analysis. The AUCs of the radiomics model in the training, testing, and external validation sets were 0.843 (95% CI: 0.788-0.897), 0.831 (95% CI: 0.741-0.920), and 0.802 (95% CI: 0.722-0.882), respectively. A nomogram was developed by integrating significant clinical factors with radiomics features. The AUCs of the nomogram in the training, testing, and external validation sets were 0.872 (95% CI: 0.823-0.921), 0.862 (95% CI: 0.780-0.944), and 0.837 (95% CI: 0.767-0.908), respectively. Survival analysis revealed that the nomogram could effectively stratify patients for recurrence-free survival (Hazard Ratio: 4.329; 95% CI: 3.159-5.934; P < 0.001).
CONCLUSIONS: The radiomics-derived nomogram presented a promising tool for predicting PNI in GC and held significant prognostic implications.
RESULTS: The nomogram functioned as a non-invasive biomarker for determining the PNI status. The predictive performance of the nomogram surpassed that of the clinical model (P < 0.05). Furthermore, patients in the high-risk group stratified by the nomogram had a significantly shorter RFS (P < 0.05).
方法:对来自两个机构的408名GC患者进行回顾性分析:来自I机构的288名患者被7:3分为训练集(n=203)和测试集(n=85);来自II机构的120名患者作为外部验证集。从CT图像中提取并筛选影像组学特征。独立的影像组学,临床,并构建了组合模型来预测PNI。模型歧视,校准,临床效用,使用曲线下面积(AUC)评估预后意义,校正曲线,决策曲线分析,和Kaplan-Meier曲线,分别。
结果:最终分析包括15个影像组学特征和3个临床因素。培训中的影像组学模型的AUC,测试,外部验证集为0.843(95%CI:0.788-0.897),0.831(95%CI:0.741-0.920),和0.802(95%CI:0.722-0.882),分别。通过将重要的临床因素与影像组学特征相结合来开发列线图。训练中的列线图的AUC,测试,外部验证集为0.872(95%CI:0.823-0.921),0.862(95%CI:0.780-0.944),和0.837(95%CI:0.767-0.908),分别。生存分析显示,列线图可以有效地对患者的无复发生存进行分层(危险比:4.329;95%CI:3.159-5.934;P<0.001)。
结论:放射组学衍生的列线图为预测GC中的PNI提供了一个有希望的工具,并具有重要的预后意义。
结果:列线图用作确定PNI状态的非侵入性生物标志物。列线图的预测性能优于临床模型(P<0.05)。此外,根据列线图分层的高危组患者的RFS明显较短(P<0.05).