文献速递头颈技术与安全CT

多模态深度学习用于喉鳞癌分期:CT与喉镜联合模型

Multimodal deep learning for laryngeal squamous cell carcinoma staging using CT and laryngoscopy.
PubMed 文献 · Eur Radiol · 2026-01-30

本研究基于两个中国医学中心450例经病理证实的喉鳞状细胞癌患者,整合临床资料、增强CT与白光喉镜图像,开发并验证区分早期(I-II期)与晚期(III-IV期)的多模态深度学习模型。该模型在内部和外部验证队列中AUC分别为0.902和0.888,优于所有单模态及双模态模型,并显示出一定的预后分层价值。

要点

  • 联合临床、增强CT和喉镜图像的多模态模型对喉鳞癌早期与晚期分期AUC达0.888–0.902。
  • 内部与外部验证均优于任何单模态或双模态模型的诊断性能。
  • 模型划分的不同风险组在无进展生存期上存在显著差异。

核心结论

  • 整合临床、增强CT和喉镜图像的多模态深度学习模型对喉鳞癌术前分期具有较高准确性,并在内部和外部验证中均优于各单模态及双模态模型。
  • 该模型可获得无创、可推广的分期信息,对个体化治疗规划具有辅助价值。

研究设计与样本

  • 回顾性多中心研究,纳入两个中国医学中心共450例经病理证实的LSCC患者。
  • 所有患者均有增强CT、白光喉镜图像和临床记录,分为训练集235例、内部验证集101例、外部验证集114例。
  • 构建并比较三种单模态模型(CT-DL、L-MIL、CL)及其组合,采用特征级融合策略,最终以SGD分类器建立综合多模态模型CL+CT+L。
  • 以AUC、准确率、敏感度、特异度、校准曲线、决策曲线分析、Kaplan-Meier和C指数进行性能与预后价值评估。

关键影像表现/征象

  • 综合多模态模型在内部验证队列AUC为0.902(95%CI:0.833–0.954),外部验证队列AUC为0.888(95%CI:0.826–0.944)。
  • 该综合多模态模型诊断效能显著高于所有单模态和双模态模型(p<0.05)。
  • 校准曲线与决策曲线分析验证了模型的一致性及临床实用性。
  • 模型划分的不同风险组在无进展生存期上存在差异(C指数=0.584,p=0.036)。

诊断标准与关键数值

  • 综合多模态模型内部验证AUC为0.902(95%CI:0.833–0.954),外部验证AUC为0.888(95%CI:0.826–0.944)。
  • 综合多模态模型对各单模态和双模态模型的性能优势具有统计学意义(p<0.05)。
  • 风险分层对无进展生存期的评估C指数为0.584(p=0.036)。

临床应用与阅片要点

  • 阅片和报告时,仅依据单一增强CT或喉镜判断早期与晚期喉鳞癌存在局限,综合临床信息可提高分期信心。
  • 该模型提示增强CT与喉镜征象互补,报告时应重视两者结合所反映的病变分期信息。
  • 模型输出的风险分组可为治疗决策提供参考,但当前无法替代病理分期。

局限性

  • 研究为回顾性设计,仅纳入两个中国医学中心数据。
  • 原文未报告模型在更广泛多中心、不同设备或不同人群中的外部泛化验证结果。
  • 模型对无进展生存期的判别能力有限(C指数=0.584)。

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)。

英文原文(摘要)· 点击展开

OBJECTIVES: To develop and validate a multimodal deep learning model integrating clinical data, contrast-enhanced CT, and laryngoscopic images for differentiating early-stage (I-II) from advanced-stage (III-IV) laryngeal squamous cell carcinoma (LSCC).

MATERIALS AND METHODS: This retrospective multicenter study included 450 patients with pathologically confirmed LSCC from two Chinese medical centers. All patients had contrast-enhanced CT, white-light laryngoscopy, and clinical records. They were divided into training (n = 235), internal validation (n = 101), and external validation (n = 114) cohorts. Three single-modality models (CT-based deep learning [CT-DL], laryngoscopy-based multiple instance learning [L-MIL], and a clinical logistic regression model [CL]) and their combinations were compared. A feature-level fusion strategy was applied, and the final integrated multimodal model (CL + CT + L) was built using a stochastic gradient descent (SGD) classifier. Performance was evaluated by AUC, accuracy, sensitivity, specificity, calibration, and decision curve analysis (DCA), with prognostic value assessed by Kaplan-Meier and concordance index (C-index).

RESULTS: A total of 450 patients were included (median age, 62 years [range, 31-88]; 365 men). The integrated multimodal model achieved AUCs of 0.902 (0.833-0.954) in the internal cohort and 0.888 (0.826-0.944) in the external cohort, outperforming all single- and dual-modality models (p < 0.05). Calibration and DCA confirmed strong consistency and clinical utility. The model categorized patients into distinct risk groups, which exhibited notable differences in progression-free survival (C-index = 0.584, p = 0.036).

CONCLUSION: The integrated multimodal model showed high accuracy and generalizability for preoperative LSCC staging and may aid individualized treatment planning.

KEY POINTS: Question Can a multimodal deep learning model combining clinical, CT, and laryngoscopic data improve preoperative staging accuracy of LSCC? Findings The integrated multimodal model achieved higher diagnostic accuracy and provided reliable prognostic stratification compared with conventional approaches. Clinical relevance This multimodal model offers a non-invasive, accurate, and generalizable tool for LSCC staging, supporting individualized treatment planning and enhancing patient management.

来源:Eur Radiol,查看原文。版权归原作者所有。

引用

  1. Liu, Zhou, Wang, et al. Multimodal deep learning for laryngeal squamous cell carcinoma staging using CT and laryngoscopy. Eur Radiol. 2026. PMID: 41617832. doi:10.1007/s00330-025-12315-4

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。