Scholarlink

<p>Small-cell lung cancer (SCLC) is an aggressive malignancy with poor survival and no validated prognostic biomarkers. We developed and validated a multimodal artificial intelligence–based radiopathomic framework to predict progression-free survival (PFS) by integrating radiologic and pathologic imaging across multi-institutional SCLC cohorts. The model jointly captures tumor-intrinsic characteristics, vascular architecture, immune organization, and stromal context from routine chest CT scans and H&amp;E-stained histopathology slides. Radiopathomic integration consistently outperformed unimodal radiologic or pathologic models in both extensive-stage (ES) and limited-stage (LS) disease. In ES-SCLC, the integrated model demonstrated superior prognostic performance compared with unimodal approaches (C-index up to 0.79), with similar improvements observed in LS-SCLC (C-index up to 0.78). Vascular tortuosity–derived risk scores were associated with distinct, stage-dependent patterns of collagen organization, highlighting differential tumor microenvironment remodeling across disease stages. Overall, this AI-driven radiopathomic approach provides a robust, non-invasive strategy for risk stratification in SCLC while offering biologically interpretable insights into vascular–stromal interactions underlying tumor aggressiveness.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multimodal radiopathological integration for prognosis of limited and extensive stage small cell lung cancer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月5日

查看全文
  • Mohammadhadi Khorrami,
  • Cristian Barrera,
  • Himanshu Maurya,
  • Pushkar Mutha,
  • Vidya S. Viswanathan,
  • Prantesh Jain,
  • Kristin A. Higgins,
  • Anant Madabhushi

摘要

Small-cell lung cancer (SCLC) is an aggressive malignancy with poor survival and no validated prognostic biomarkers. We developed and validated a multimodal artificial intelligence–based radiopathomic framework to predict progression-free survival (PFS) by integrating radiologic and pathologic imaging across multi-institutional SCLC cohorts. The model jointly captures tumor-intrinsic characteristics, vascular architecture, immune organization, and stromal context from routine chest CT scans and H&E-stained histopathology slides. Radiopathomic integration consistently outperformed unimodal radiologic or pathologic models in both extensive-stage (ES) and limited-stage (LS) disease. In ES-SCLC, the integrated model demonstrated superior prognostic performance compared with unimodal approaches (C-index up to 0.79), with similar improvements observed in LS-SCLC (C-index up to 0.78). Vascular tortuosity–derived risk scores were associated with distinct, stage-dependent patterns of collagen organization, highlighting differential tumor microenvironment remodeling across disease stages. Overall, this AI-driven radiopathomic approach provides a robust, non-invasive strategy for risk stratification in SCLC while offering biologically interpretable insights into vascular–stromal interactions underlying tumor aggressiveness.