Academic radiology · 2026
A Fusion-based Radiomics Approach Combining Tumor, Peritumoral, and Habitat Features for Predicting Visceral Pleural Invasion in Stage IA Lung Adenocarcinoma.
СтатьяМногоцентровое исследованиеНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Chen Zhenjing, Lin Xueyao, Lou Kaihua, Zheng Jinjing, Chen Hang, Xu Guodong, Ren Feng, Hu Hongjie, Dong Haibo, Huang Shaoyu
Аннотация
RATIONALE AND OBJECTIVES: Visceral pleural invasion (VPI) is a critical prognostic factor in early-stage lung adenocarcinoma (LUAD). This study aimed to develop a multi-scale radiomics model integrating intratumoral, peritumoral, and habitat heterogeneity features to predict VPI status.
MATERIALS AND METHODS: This multicenter retrospective study enrolled 424 stage IA LUAD patients divided into the training (n = 247), internal validation (n = 107), and external validation (n = 70) cohorts. Gross-tumoral, peritumoral radiomics features, and intratumoral habitat clustering features were extracted from CT images. Independent risk factors were identified from clinical and radiological characteristics through univariate and multivariate analyses. Extreme Gradient Boosting (XGB) was used to build individual prediction models (PMs), which were then integrated via feature-based prefusion and decision-based postfusion strategies. Model performance was evaluated using the area under the curve (AUC), decision curve analysis (DCA) and calibration curve. Associations between features were assessed using Pearson correlation analysis. The SHapley Additive exPlanations (SHAP) method was utilized for model interpretability analysis.
RESULTS: The multi-scale postfusion model (PMPostfusion) achieved the highest predictive performance, with AUCs of 0.923, 0.793, and 0.784 in the training, internal validation, and external validation cohorts, respectively. DCA indicated its higher clinical benefit, while calibration curve verified its predictive reliability. Feature correlation analysis revealed low redundancy across different scale features (only 0.23% strongly correlated). SHAP interpretation identified the PMHabitat-based signature as the most influential predictor of VPI status.
CONCLUSION: PMPostfusion showed favorable diagnostic performance in preoperative prediction of VPI in stage IA LUAD, and may serve as a promising auxiliary tool for clinical decision-making, pending prospective validation and clinical workflow integration.
MATERIALS AND METHODS: This multicenter retrospective study enrolled 424 stage IA LUAD patients divided into the training (n = 247), internal validation (n = 107), and external validation (n = 70) cohorts. Gross-tumoral, peritumoral radiomics features, and intratumoral habitat clustering features were extracted from CT images. Independent risk factors were identified from clinical and radiological characteristics through univariate and multivariate analyses. Extreme Gradient Boosting (XGB) was used to build individual prediction models (PMs), which were then integrated via feature-based prefusion and decision-based postfusion strategies. Model performance was evaluated using the area under the curve (AUC), decision curve analysis (DCA) and calibration curve. Associations between features were assessed using Pearson correlation analysis. The SHapley Additive exPlanations (SHAP) method was utilized for model interpretability analysis.
RESULTS: The multi-scale postfusion model (PMPostfusion) achieved the highest predictive performance, with AUCs of 0.923, 0.793, and 0.784 in the training, internal validation, and external validation cohorts, respectively. DCA indicated its higher clinical benefit, while calibration curve verified its predictive reliability. Feature correlation analysis revealed low redundancy across different scale features (only 0.23% strongly correlated). SHAP interpretation identified the PMHabitat-based signature as the most influential predictor of VPI status.
CONCLUSION: PMPostfusion showed favorable diagnostic performance in preoperative prediction of VPI in stage IA LUAD, and may serve as a promising auxiliary tool for clinical decision-making, pending prospective validation and clinical workflow integration.