Jiang Wei, Wang Huaiming, Dong Xiaoyu, Zhao Yandong, Long Chenyan, Chen Dexin, Yan Botao, Cheng Jiaxin, Lin Zexi, Zhuo Shuangmu, Wang Hui, Yan Jun
Department of General Surgery, Guangdong Provincial Key Laboratory of Precision Medicine for Gastrointestinal Tumor, Nanfang Hospital, The First School of Clinical Medicine, Southern Medical University, Guangzhou, Guangdong, 510515, People's Republic of China.
School of Science, Jimei University, Xiamen, Fujian, 361021, People's Republic of China.
J Transl Med. 2024 Jan 25;22(1):103. doi: 10.1186/s12967-024-04851-2.
Lymph node metastasis (LNM) is a prognostic biomarker and affects therapeutic selection in colorectal cancer (CRC). Current evaluation methods are not adequate for estimating LNM in CRC. H&E images contain much pathological information, and collagen also affects the biological behavior of tumor cells. Hence, the objective of the study is to investigate whether a fully quantitative pathomics-collagen signature (PCS) in the tumor microenvironment can be used to predict LNM.
Patients with histologically confirmed stage I-III CRC who underwent radical surgery were included in the training cohort (n = 329), the internal validation cohort (n = 329), and the external validation cohort (n = 315). Fully quantitative pathomics features and collagen features were extracted from digital H&E images and multiphoton images of specimens, respectively. LASSO regression was utilized to develop the PCS. Then, a PCS-nomogram was constructed incorporating the PCS and clinicopathological predictors for estimating LNM in the training cohort. The performance of the PCS-nomogram was evaluated via calibration, discrimination, and clinical usefulness. Furthermore, the PCS-nomogram was tested in internal and external validation cohorts.
By LASSO regression, the PCS was developed based on 11 pathomics and 9 collagen features. A significant association was found between the PCS and LNM in the three cohorts (P < 0.001). Then, the PCS-nomogram based on PCS, preoperative CEA level, lymphadenectasis on CT, venous emboli and/or lymphatic invasion and/or perineural invasion (VELIPI), and pT stage achieved AUROCs of 0.939, 0.895, and 0.893 in the three cohorts. The calibration curves identified good agreement between the nomogram-predicted and actual outcomes. Decision curve analysis indicated that the PCS-nomogram was clinically useful. Moreover, the PCS was still an independent predictor of LNM at station Nos. 1, 2, and 3. The PCS nomogram displayed AUROCs of 0.849-0.939 for the training cohort, 0.837-0.902 for the internal validation cohort, and 0.851-0.895 for the external validation cohorts in the three nodal stations.
This study proposed that PCS integrating pathomics and collagen features was significantly associated with LNM, and the PCS-nomogram has the potential to be a useful tool for predicting individual LNM in CRC patients.
淋巴结转移（LNM）是一种预后生物标志物，影响结直肠癌（CRC）的治疗选择。目前的评估方法不足以估计CRC中的LNM。苏木精-伊红（H&E）图像包含大量病理信息，胶原蛋白也会影响肿瘤细胞的生物学行为。因此，本研究的目的是探讨肿瘤微环境中完全定量的病理组学-胶原蛋白特征（PCS）是否可用于预测LNM。
将接受根治性手术且组织学确诊为I-III期CRC的患者纳入训练队列（n = 329）、内部验证队列（n = 329）和外部验证队列（n = 315）。分别从标本的数字H&E图像和多光子图像中提取完全定量的病理组学特征和胶原蛋白特征。利用套索回归开发PCS。然后，构建一个PCS列线图，纳入PCS和临床病理预测因子，用于估计训练队列中的LNM。通过校准、区分度和临床实用性评估PCS列线图的性能。此外，在内部和外部验证队列中对PCS列线图进行测试。
通过套索回归，基于11个病理组学特征和9个胶原蛋白特征开发了PCS。在三个队列中发现PCS与LNM之间存在显著关联（P < 0.001）。然后，基于PCS、术前癌胚抗原（CEA）水平、CT上的淋巴结清扫、静脉栓塞和/或淋巴侵犯和/或神经周围侵犯（VELIPI）以及pT分期的PCS列线图在三个队列中的曲线下面积（AUROC）分别为0.939、0.895和0.893。校准曲线表明列线图预测结果与实际结果之间具有良好的一致性。决策曲线分析表明PCS列线图具有临床实用性。此外，PCS仍然是第1、2和3站LNM的独立预测因子。在三个淋巴结站中，PCS列线图在训练队列中的AUROC为0.849 - 0.939，在内部验证队列中为0.837 - 0.902，在外部验证队列中为0.851 - 0.895。
本研究表明，整合病理组学和胶原蛋白特征的PCS与LNM显著相关，并且PCS列线图有可能成为预测CRC患者个体LNM的有用工具。
