An integrated MRI-pathology model for predicting response to TACE combined with targeted therapy and immunotherapy in hepatocellular carcinoma.
Authors
Affiliations (9)
Affiliations (9)
- Department of Interventional Radiology, The Affiliated Wuxi People's Hospital of Nanjing Medical University, Wuxi People's Hospital, Wuxi Medical Center, Nanjing Medical University, Wuxi, China.
- Department of Pathology, The First Affiliated Hospital with Nanjing Medical University, Nanjing, China.
- Department of Interventional Radiology, The First Affiliated Hospital with Nanjing Medical University, Nanjing, China.
- Department of Interventional Radiology, The Affiliated Suqian First People's Hospital of Nanjing Medical University, Suqian, China.
- Department of Radiology, Sir Run Run Shaw Hospital, School of Medicine, Zhejiang University, Hangzhou, China.
- Department of Computer Science and Technology, Zhejiang University, Hangzhou, China.
- Department of Interventional Radiology, The Affiliated Wuxi People's Hospital of Nanjing Medical University, Wuxi People's Hospital, Wuxi Medical Center, Nanjing Medical University, Wuxi, China. [email protected].
- Department of Interventional Radiology, The First Affiliated Hospital with Nanjing Medical University, Nanjing, China. [email protected].
- Department of Interventional Radiology, The First Affiliated Hospital with Nanjing Medical University, Nanjing, China. [email protected].
Abstract
The efficacy of transarterial chemoembolization (TACE) plus targeted therapy and immunotherapy in unresectable hepatocellular carcinoma (HCC) is limited by variable responses. We aimed to develop and validate an integrated magnetic resonance imaging (MRI)-pathology model for response prediction and survival stratification (IMPRESS), which combines pretreatment MRI habitat radiomics and paired biopsy whole-slide images (WSIs) pathomics. This multicenter, retrospective study enrolled patients with unresectable HCC receiving TACE plus targeted therapy and immunotherapy from January 2019 to December 2024. Eligibility required pretreatment MRI and paired biopsy WSIs within 1 month before therapy, Child-Pugh class A/B, Eastern Cooperative Oncology Group (ECOG) performance status ≤ 1, and at least one measurable intrahepatic lesion ≥ 1 cm. Patients with missing or poor-quality MRI or WSIs, prior locoregional or systemic HCC therapy, second malignancy, ECOG ≥ 2, or incomplete data were excluded. Habitat MRI radiomics, WSI pathomics, and clinical features were integrated into IMPRESS. Model performance was evaluated using the area under the curve (AUC) with 95% confidence intervals (CIs). Single-cell analysis was performed to explore the biological basis of the IMPRESS predictions. Among 231 patients, 94 (40.7%) achieved a treatment response. IMPRESS achieved AUCs of 0.957 (95% CI: 0.931-1.0), 0.878 (95% CI: 0.789-0.966), and 0.875 (95% CI: 0.782-0.966) in the training, validation, and test cohorts, respectively. Predicted responders and non-responders showed significantly different overall survival in all cohorts (all p < 0.001). Single-cell RNA sequencing revealed distinct immune and proliferative patterns consistent with model predictions. IMPRESS model may provide a promising tool for pretreatment response prediction and survival risk stratification in unresectable HCC undergoing TACE-based combination therapy. Question Reliable pretreatment tools are needed to identify unresectable HCC patients likely to benefit from TACE plus targeted therapy and immunotherapy. Findings An integrated MRI-pathology model combining habitat radiomics and deep-learning pathomics predicted treatment response, with validation and testing areas under the curve of 0.878 and 0.875. Clinical relevance The integrated model may support pretreatment response stratification for TACE-based triple therapy and help inform surveillance intensity or alternative treatment considerations, pending prospective validation.