文献速递消化系统肿瘤

多模态深度学习模型用于胃癌术后并发症的术前预测

A multimodal deep learning model for preoperative prediction of post-operative complications in gastric cancer.
PubMed 文献 · Ann Oncol · 2026-07-16

本研究开发并验证了多模态深度学习框架DeepComp,整合临床变量与影像特征,用于胃癌根治术前同时预测Clavien-Dindo≥II级并发症和长期生存。模型在5237例中国多中心患者中表现优异,并在九个外部队列及六项临床试验数据中验证,辅助外科医生可显著提高预测敏感性并支持个体化围手术期管理。

要点

  • DeepComp预测CD≥II并发症的AUC为0.888,外部验证AUC 0.824-0.869,优于所有临床评分。
  • 模型辅助使10名外科医生的平均敏感性从47.1%提升至87.9%。
  • DeepComp引导的预防性ICU监测、营养支持延迟手术和微创分诊可降低并发症绝对风险。

核心结论

  • DeepComp在验证队列中对胃癌术后并发症和长期生存表现一致,可支持个体化围手术期管理。

研究设计与样本

  • 多中心研究分析11个中国中心的5237例胃腺癌患者,并用六项注册临床试验的前瞻性数据额外验证。

主要发现

  • DeepComp预测CD≥II并发症的AUC在合并内部验证集为0.888(95%CI 0.854-0.921),九个外部队列为0.824-0.869。
  • DeepComp比最佳临床基线高15.3个百分点(P<0.001),并优于所有九种既定临床评分(均P<0.001)。
  • DeepComp辅助下10名外科医生的平均敏感性从47.1%升至87.9%(P<0.001)。
  • 通过目标试验模拟,DeepComp引导的预防性ICU监测、术前营养支持延迟手术和微创分诊分别带来CD≥II绝对风险降低5.9%、20.6%和11.7%(均P<0.01;E值1.90-5.73)。
  • DeepComp独立预测OS(每标准差调整HR 3.08,95%CI 2.91-3.25;合并C-index 0.766),5年生存率从五分位1的97.5%到五分位5的2.4%。

临床怎么用

  • 可用于胃癌根治术前识别CD≥II并发症高风险患者,辅助制定预防性ICU监测、营养支持延迟手术或微创分诊决策。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

BACKGROUND: Clavien-Dindo (CD) grade ≥II complications occur in approximately one in five patients after curative gastrectomy for gastric cancer and independently shorten overall survival (OS) by compromising adjuvant therapy delivery. Current nutritional, inflammatory, and global surgical-risk scores identify only a minority of affected patients. We developed and validated a multimodal deep learning framework for simultaneous preoperative prediction of CD grade ≥II complications and long-term survival. PATIENTS AND METHODS: This multicenter study analyzed 5237 patients with gastric adenocarcinoma from 11 Chinese centers and additionally validated the model using prospectively collected data from six registered clinical trials covering three neoadjuvant treatment settings (chemotherapy, chemoradiation, and immunochemotherapy). DeepComp integrates clinical variables with foundation-model image features from the target lesion region, 5-mm peritumoral region, and L3 body composition compartments through a tabular dual-task architecture to jointly predict CD grade ≥II complications and OS. RESULTS: DeepComp achieved an area under the receiver operating characteristic curve of 0.888 [95% confidence interval (CI) 0.854-0.921] in the merged internal validation set and 0.824-0.869 across nine external cohorts, outperforming the best clinical baseline by 15.3 percentage points (P < 0.001) and all nine established clinical scores (all P < 0.001). With DeepComp assistance, the mean sensitivity of 10 surgeons increased from 47.1% to 87.9% (P < 0.001). Through target-trial emulation, DeepComp-guided prophylactic intensive care unit monitoring, preoperative nutritional support with delayed surgery, and minimally invasive triage yielded absolute CD grade ≥II risk reductions of 5.9%, 20.6%, and 11.7%, respectively (all P < 0.01; E-values 1.90-5.73). DeepComp independently predicted OS (adjusted hazard ratio 3.08 per standard deviation, 95% CI 2.91-3.25; pooled C-index 0.766), with 5-year survival ranging from 97.5% in quintile 1 to 2.4% in quintile 5. CONCLUSIONS: DeepComp showed consistent performance across validation cohorts for preoperative risk stratification of post-operative complications and long-term survival and may support individualized perioperative management.

来源:Ann Oncol,查看原文。版权归原作者所有。

引用

  1. Ding PA, Yang S, Meng LJ, et al. A multimodal deep learning model for preoperative prediction of post-operative complications in gastric cancer. Ann Oncol. 2026. PMID: 42463045. doi:10.1016/j.annonc.2026.07.004

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。