Preoperative CT-based topologically distinct intratumoral heterogeneity scores for predicting intratumoral tertiary lymphoid structures and outcomes in hepatocellular carcinoma: A multicenter study.
Authors
Affiliations (10)
Affiliations (10)
- Department of Radiology, The Second Affiliated Hospital, Jiangxi Medical College, Nanchang University, Nanchang, Jiangxi, 330006, China; Jiangxi Provincial Key Laboratory of Intelligent Medical Imaging, Nanchang, Jiangxi, 330006, China; Nanchang Key Laboratory of Medical-Engineering Integration and Clinical Translation, Nanchang, Jiangxi, 330006, China.
- Department of Radiology, The Quzhou Affiliated Hospital of Wenzhou Medical University, Quzhou, People's Hospital, Quzhou, Zhejiang, 324000, China.
- Department of Radiology, The First Affiliated Hospital, Jiangxi Medical College, Nanchang University, Nanchang, Jiangxi, 330006, China.
- Department of Pathology, Jiangxi Cancer Hospital, Nanchang, Jiangxi, 330029, China.
- Department of Radiology, The Third Hospital of Quzhou, Quzhou, Zhejiang, 324000, China.
- Department of Ultrasound, Jiangxi Maternal and Child Health Hospital, Nanchang, Jiangxi, 330038, China.
- School of Mathematics and Computational Science, Xiangtan University, Xiangtan, Hunan, 411105, China.
- Department of Radiology, Jiangxi Cancer Hospital, Nanchang, Jiangxi, 330029, China. Electronic address: [email protected].
- Department of Radiology, Xiangtan Central Hospital, Xiangtan, Hunan, 411000, China. Electronic address: [email protected].
- Department of Radiology, The Second Affiliated Hospital, Jiangxi Medical College, Nanchang University, Nanchang, Jiangxi, 330006, China; Jiangxi Provincial Key Laboratory of Intelligent Medical Imaging, Nanchang, Jiangxi, 330006, China; Nanchang Key Laboratory of Medical-Engineering Integration and Clinical Translation, Nanchang, Jiangxi, 330006, China. Electronic address: [email protected].
Abstract
Intratumoral tertiary lymphoid structures (iTLSs) are prognostic biomarkers for hepatocellular carcinoma (HCC). This study aimed to develop a machine learning approach based on topologically distinct intratumoral heterogeneity (ITH) scores derived from CT images to predict iTLS status and patient outcomes. In this multicenter study, patients from Centers 1 and 2 were divided into training (n = 475) and internal validation (n = 204) cohorts, with an independent cohort (n = 208) used for external validation from Center 3. Two complementary ITH scores were developed: a 2D score integrating local radiomics with global pixel patterns on the largest cross-sectional slice, and a 3D score extending this quantification to the entire tumor volume. A stacking ensemble classifier (2D3D-TD-ITH-Ensemble) incorporating clinicoradiological features and ITH scores was constructed to predict iTLS status. Model performance was compared with clinical and traditional radiomics models. SHapley Additive exPlanations (SHAP) analysis was used for interpretability. Disease-free survival (DFS) was assessed using Kaplan-Meier analysis. The 2D3D-TD-ITH-Ensemble demonstrated superior diagnostic performance compared to reference models. In the internal validation cohort, the ensemble model achieved an AUC of 0.904, outperforming the radiomics (AUC 0.887) and clinical models (AUC 0.811). Consistent results were observed in the external validation cohort, where the ensemble model yielded an AUC of 0.890, versus 0.864 for the radiomics model and 0.817 for the clinical model. SHAP analysis identified the 3D ITH score as the most influential contributor to model output. Furthermore, HCC patients with the presence of iTLS and lower 3D ITH scores exhibited significantly better DFS (p < 0.05). The preoperative CT-based 3D ITH score serves as a robust non-invasive biomarker for predicting iTLS status and prognosis in HCC, potentially guiding stratified immunotherapy strategies.