Patient-specific biophysical AI-powered modeling for predicting RFA volumes in HCC
- 1Siemens Healthineers, Swiss Innovation Hub, Lausanne, Switzerland
- 2Department of Radiology and Interventional Radiology, Lausanne University Hospital, Lausanne, Switzerland
- 3Siemens Healthineers, Digital Technologies and Innovation, Erlangen, Germany
Abstract
Objectives: To clinically evaluate a fully automated, artificial intelligence-powered, patient-specific biophysical model for predicting radiofrequency ablation (RFA) volumes in hepatocellular carcinoma (HCC) and to compare its performance with manufacturer-provided ablation estimates.
Materials and methods: This retrospective study included 96 RFA procedures in 65 HCC patients treated using a single ablation with a cluster needle. Tumors, needle position, and postablation zone were manually segmented on periprocedural computed tomography. Liver parenchyma and vascular structures were automatically extracted using deep learning. A graphics processing unit (GPU)-accelerated biophysical model simulated heat transfer, vascular heat-sink effects, and tissue necrosis using patient-specific anatomy and procedural data. Correlations between ablation volume manually segmented and clinical or anatomical parameters (ablation time, ablation tip temperature, tumor–capsule distance, vessel proximity, type, and diameter) were assessed. Model-predicted volumes were compared with ground truth and manufacturer estimates using relative volume difference (RVD) and volume similarity.
Results: Final ablation volumes showed wide variability (7.9–78.8 cm³) and were not correlated with any individual clinical or anatomical parameter (Pearson r – 0.02 to 0.29). Across 94 evaluable cases, simulated and ground-truth volumes were similar (p = 0.165), while manufacturer estimates significantly overestimated ablation volume (p < 0.001). The model achieved a median RVD of 0.39 (interquartile range 0.24–0.57), significantly lower than the manufacturer's algorithm (1.04; 0.64–1.61). Incorporating patient-specific ablation time further improved accuracy. Simulation was fast (0.45 s per second of ablation on GPU).
Conclusion: A fully automated, patient-specific biophysical model substantially improves RFA ablation volume prediction in HCC compared with manufacturer approximations, supporting its potential role in personalized treatment planning and outcome standardization.
Key Points: Question In patients with hepatocellular carcinoma, can a fully automated, AI-powered, patient-specific biophysical model predict radiofrequency ablation volume more accurately than manufacturer-provided estimates?
Findings A fully automated, AI-powered biophysical model integrating patient-specific anatomy and vasculature predicted RFA volume significantly better than manufacturer estimates (median relative volume difference (RVD) 0.39 versus 1.04).
Relevance statement Accurate ablation volume prediction could support personalized radiofrequency ablation (RFA) planning, reduce reliance on operator experience, and standardize treatment margins in hepatocellular carcinoma (HCC), potentially improving local tumor control and reducing recurrence in cirrhotic patients undergoing thermal ablation.
Graphical Abstract:
Related
- Patient-specific finite element modeling of image-derived shell morphoelastic growth: A general framework and application to stented aortic dissection
- Editorial: AI-powered insights: predicting treatment response and prognosis in breast cancer
- Airflow Characteristics of Empty Nose Syndrome: Insights From Patient‐Specific Computational Modeling of Airflow Distribution Patterns and Nasal Air‐Conditioning
- Intelligence-enabled smart inhalable biomaterials for chronic inflammatory airway diseases: integrating responsive carrier engineering, lung-on-chip validation, and patient-specific pulmonary modeling
- Predicting pandemics, predicting people: the ethical limits of AI-powered public health surveillance
BibTeX
@article{audigier2026patientspecific,
title = {{Patient-specific biophysical AI-powered modeling for predicting RFA volumes in HCC}},
author = {Chloé Audigier and Fouad Georges Akkari and Felix Meister and Andrea Tonglet and Bénédicte Maréchal and Oliver Frings and Rafael Duran},
journal = {European Radiology Experimental},
year = {2026},
month = nov,
doi = {10.1186/s41747-026-00839-3},
url = {https://doi.org/10.1186/s41747-026-00839-3},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Patient-specific biophysical AI-powered modeling for predicting RFA volumes in HCC
AU - Chloé Audigier
AU - Fouad Georges Akkari
AU - Felix Meister
AU - Andrea Tonglet
AU - Bénédicte Maréchal
AU - Oliver Frings
AU - Rafael Duran
PY - 2026
DA - 2026/11/03
JO - European Radiology Experimental
DO - 10.1186/s41747-026-00839-3
UR - https://doi.org/10.1186/s41747-026-00839-3
AB - Objectives: To clinically evaluate a fully automated, artificial intelligence-powered, patient-specific biophysical model for predicting radiofrequency ablation (RFA) volumes in hepatocellular carcinoma (HCC) and to compare its performance with manufacturer-provided ablation estimates. Materials and methods: This retrospective study included 96 RFA procedures in 65 HCC patients treated using a single ablation with a cluster needle. Tumors, needle position, and postablation zone were manually segmented on periprocedural computed tomography. Liver parenchyma and vascular structures were automatically extracted using deep learning. A graphics processing unit (GPU)-accelerated biophysical model simulated heat transfer, vascular heat-sink effects, and tissue necrosis using patient-specific anatomy and procedural data. Correlations between ablation volume manually segmented and clinical or anatomical parameters (ablation time, ablation tip temperature, tumor–capsule distance, vessel proximity, type, and diameter) were assessed. Model-predicted volumes were compared with ground truth and manufacturer estimates using relative volume difference (RVD) and volume similarity. Results: Final ablation volumes showed wide variability (7.9–78.8 cm³) and were not correlated with any individual clinical or anatomical parameter (Pearson r – 0.02 to 0.29). Across 94 evaluable cases, simulated and ground-truth volumes were similar (p = 0.165), while manufacturer estimates significantly overestimated ablation volume (p < 0.001). The model achieved a median RVD of 0.39 (interquartile range 0.24–0.57), significantly lower than the manufacturer's algorithm (1.04; 0.64–1.61). Incorporating patient-specific ablation time further improved accuracy. Simulation was fast (0.45 s per second of ablation on GPU). Conclusion: A fully automated, patient-specific biophysical model substantially improves RFA ablation volume prediction in HCC compared with manufacturer approximations, supporting its potential role in personalized treatment planning and outcome standardization. Key Points: Question In patients with hepatocellular carcinoma, can a fully automated, AI-powered, patient-specific biophysical model predict radiofrequency ablation volume more accurately than manufacturer-provided estimates? Findings A fully automated, AI-powered biophysical model integrating patient-specific anatomy and vasculature predicted RFA volume significantly better than manufacturer estimates (median relative volume difference (RVD) 0.39 versus 1.04). Relevance statement Accurate ablation volume prediction could support personalized radiofrequency ablation (RFA) planning, reduce reliance on operator experience, and standardize treatment margins in hepatocellular carcinoma (HCC), potentially improving local tumor control and reducing recurrence in cirrhotic patients undergoing thermal ablation. Graphical Abstract:
SN - 2509-9280
ER -