Li Peishen, Huang Shujie, Xu Haijie, Li Zijie, Wang Sichao, Gao Zhen, Dong Yuejiao, Chen Zhuofeng, Qiao Guibin, Wu Hansheng, Hong Liangli
Department of Thoracic Surgery, The First Affiliated Hospital of Shantou University Medical College, Shantou, China.
Department of Thoracic Surgery, Guangdong Provincial People's Hospital (Guangdong Academy of Medical Sciences), Southern Medical University, Guangzhou, China.
Cancer Med. 2025 Jul;14(14):e71054. doi: 10.1002/cam4.71054.
Tumor-infiltrating lymphocytes (TILs) have been proven to be important prognostic factors for various tumors. However, their prognostic significance within the context of esophageal squamous cell carcinoma (ESCC) remains inadequately explored. This study aims to assess the prognostic potential of TILs in ESCC using deep learning (DL) methods.
We retrospectively enrolled 626 pathologically confirmed ESCC patients from two research centers. Their digital whole-slide imaging (WSI) and corresponding clinical information were collected. Subsequently, the DL method was employed to identify the tumor margin and TILs within the WSI. Tissue was divided into intratumor, peritumoral, and stromal regions based on their distance from the tumor margin. TILs were counted in each region. Optimal cut-off values of TILs were determined using the X-tile software. To mitigate selection bias and intergroup heterogeneity, a propensity score matching (PSM) analysis was employed. Survival analysis was performed using Kaplan-Meier curves and the log-rank test. The Cox proportional hazards regression model was used to identify independent prognostic factors.
We classified patients based on the cell counts and cut-off values of intratumor-infiltrating lymphocytes (I-TILs) and peritumoral infiltrating lymphocytes (P-TILs). Patients with high I-TILs and P-TILs were defined as those whose counts of both I-TILs and P-TILs exceeded the determined cutoff value. Patients with high I-TILs and P-TILs showed significantly better overall survival (OS, p = 0.0092) and recurrence-free survival (RFS, p = 0.0088) than patients with low I-TILs and P-TILs after PSM. Multivariable Cox proportional hazards regression further supported this conclusion and recognized I-TILs and P-TILs as independent prognostic factors (p = 0.0136, hazard ratio = 0.63 for OS; p = 0.0098, hazard ratio = 0.63 for RFS).
In the present study, we identified the quantitative distribution of TILs in ESCC patients with the help of the DL method. We established that I-TILs and P-TILs serve as independent prognostic factors for these patients. Further studies should focus on the lymphocyte subgroups and make better use of the spatial information to improve the predictive efficacy of TILs.
肿瘤浸润淋巴细胞（TILs）已被证明是多种肿瘤的重要预后因素。然而，其在食管鳞状细胞癌（ESCC）中的预后意义仍未得到充分探索。本研究旨在使用深度学习（DL）方法评估TILs在ESCC中的预后潜力。
我们回顾性纳入了来自两个研究中心的626例经病理确诊的ESCC患者。收集了他们的数字全切片成像（WSI）及相应的临床信息。随后，采用DL方法识别WSI中的肿瘤边缘和TILs。根据组织与肿瘤边缘的距离，将组织分为肿瘤内、肿瘤周围和基质区域。对每个区域的TILs进行计数。使用X-tile软件确定TILs的最佳截断值。为减轻选择偏倚和组间异质性，采用倾向评分匹配（PSM）分析。使用Kaplan-Meier曲线和对数秩检验进行生存分析。采用Cox比例风险回归模型识别独立预后因素。
我们根据肿瘤内浸润淋巴细胞（I-TILs）和肿瘤周围浸润淋巴细胞（P-TILs）的细胞计数及截断值对患者进行分类。I-TILs和P-TILs计数均超过确定截断值的患者被定义为高I-TILs和P-TILs患者。PSM后，高I-TILs和P-TILs患者的总生存期（OS，p = 0.0092）和无复发生存期（RFS，p = 0.0088）明显优于低I-TILs和P-TILs患者。多变量Cox比例风险回归进一步支持了这一结论，并将I-TILs和P-TILs识别为独立预后因素（OS，p = 0.0136，风险比 = 0.63；RFS，p = 0.0098，风险比 = 0.63）。
在本研究中，我们借助DL方法确定了ESCC患者中TILs的定量分布。我们确定I-TILs和P-TILs是这些患者的独立预后因素。进一步的研究应聚焦于淋巴细胞亚群，并更好地利用空间信息以提高TILs的预测效能。
