Zhao Xiaoyi, Zhao Heng, Dai Kongxu, Zeng Xiangyu, Li Yun, Yang Feng, Jiang Guanchao
Department of Thoracic Surgery, Peking University People's Hospital, No. 11 Xizhimen South Street, Xicheng District, Beijing 100044, China.
Thoracic Oncology Institute, Peking University People's Hospital, Beijing 100044, China.
Curr Oncol. 2025 Apr 11;32(4):223. doi: 10.3390/curroncol32040223.
The preoperative identification of occult pleural metastasis (OPM) in lung cancer remains a crucial clinical challenge. This study aimed to develop and validate a predictive model that integrates clinical information with chest CT radiomic features to preoperatively identify patients at risk of OPM.
This study included 50 patients diagnosed with OPM during surgery as the positive training cohort and an equal number of nonmetastatic patients as the negative control cohort. Using least absolute shrinkage and selection operator (LASSO) logistic regression, we identified key radiomic features and calculated radiomic scores. A predictive nomogram was developed by combining clinical characteristics and radiomic scores, which was subsequently validated with data from an additional 545 patients across three medical centers.
Univariate and multivariate logistic regression analyses revealed that carcinoembryonic antigen (CEA), the neutrophil-to-lymphocyte ratio (NLR), the clinical T stage, and the tumor-pleural relationship were significant clinical predictors. The clinical model alone achieved an area under the curve (AUC) of 0.761. The optimal integrated model, which combined radiomic scores from the volume of interest (VOI) with the CEA and NLR, demonstrated an improved predictive performance, with AUCs of 0.890 in the training cohort and 0.855 in the validation cohort.
Radiomic features derived from CT scans show significant promise in identifying patients with lung cancer at risk of OPM. The nomogram developed in this study, which integrates CEA, the NLR, and radiomic tumor area scores, enhances the precision of preoperative OPM prediction and provides a valuable tool for clinical decision-making.
肺癌隐匿性胸膜转移（OPM）的术前识别仍然是一项关键的临床挑战。本研究旨在开发并验证一种预测模型，该模型将临床信息与胸部CT影像组学特征相结合，以术前识别有OPM风险的患者。
本研究纳入50例在手术中诊断为OPM的患者作为阳性训练队列，以及数量相等的非转移患者作为阴性对照队列。使用最小绝对收缩和选择算子（LASSO）逻辑回归，我们识别出关键的影像组学特征并计算影像组学评分。通过结合临床特征和影像组学评分开发了一种预测列线图，随后使用来自三个医疗中心的另外545例患者的数据对其进行验证。
单因素和多因素逻辑回归分析显示，癌胚抗原（CEA）、中性粒细胞与淋巴细胞比值（NLR）、临床T分期和肿瘤与胸膜的关系是重要的临床预测指标。单独的临床模型的曲线下面积（AUC）为0.761。最佳的综合模型将感兴趣体积（VOI）的影像组学评分与CEA和NLR相结合，显示出更好的预测性能，训练队列中的AUC为0.890，验证队列中的AUC为0.855。
CT扫描得出的影像组学特征在识别有OPM风险的肺癌患者方面显示出巨大潜力。本研究开发的列线图整合了CEA、NLR和影像组学肿瘤面积评分，提高了术前OPM预测的准确性，并为临床决策提供了有价值的工具。
