Michael Moche1,2, Harald Busse1, Jurgen J Futterer3, Camila A Hinestrosa1, Daniel Seider1, Philipp Brandmaier1, Marina Kolesnik4, Sjoerd Jenniskens3, Roberto Blanco Sequeiros5, Gaber Komar5, Mika Pollari6, Martin Eibisberger7, Horst Rupert Portugaller7, Philip Voglreiter8, Ronan Flanagan9, Panchatcharam Mariappan9,10, Martin Reinhardt11. 1. Department of Diagnostic and Interventional Radiology, University of Leipzig Medical Center, Liebigstraße 20, 04103, Leipzig, Germany. 2. Department of Interventional Radiology, Helios Park-Klinikum Leipzig, Leipzig, Germany. 3. Department of Radiology and Nuclear Medicine, Radboudumc, Nijmegen, Netherlands. 4. Fraunhofer Institute for Applied Information Technology FIT, Sankt Augustin, Germany. 5. Department of Radiology, Turku University Hospital, Turku, Finland. 6. Department of Computer Science, Aalto University School of Science and Technology, 02150, Espoo, Finland. 7. University Clinic of Radiology Graz, Graz, Austria. 8. Institute of Computer Graphics and Vision, Graz University of Technology, Graz, Austria. 9. NUMA Engineering Services Ltd., Louth, Ireland. 10. Indian Institute of Technology, Tirupati, India. 11. Department of Diagnostic and Interventional Radiology, University of Leipzig Medical Center, Liebigstraße 20, 04103, Leipzig, Germany. martin.reinhardt@medizin.uni-leipzig.de.
Abstract
OBJECTIVES: To evaluate the accuracy and clinical integrability of a comprehensive simulation tool to plan and predict radiofrequency ablation (RFA) zones in liver tumors. METHODS: Forty-five patients with 51 malignant hepatic lesions of different origins were included in a prospective multicenter trial. Prior to CT-guided RFA, all patients underwent multiphase CT which included acquisitions for the assessment of liver perfusion. These data were used to generate a 3D model of the liver. The intra-procedural position of the RFA probe was determined by CT and semi-automatically registered to the 3D model. Size and shape of the simulated ablation zones were compared with those of the thermal ablation zones segmented in contrast-enhanced CT images 1 month after RFA; procedure time was compared with a historical control group. RESULTS: Simulated and segmented ablation zone volumes showed a significant correlation (ρ = 0.59, p < 0.0001) and no significant bias (Wilcoxon's Z = 0.68, p = 0.25). Representative measures of ablation zone comparison were as follows: average surface deviation (absolute average error, AAE) with 3.4 ± 1.7 mm, Dice similarity coefficient 0.62 ± 0.14, sensitivity 0.70 ± 0.21, and positive predictive value 0.66 ± 0. There was a moderate positive correlation between AAE and duration of the ablation (∆t; r = 0.37, p = 0.008). After adjustments for inter-individual differences in ∆t, liver perfusion, and prior transarterial chemoembolization procedures, ∆t was an independent predictor of AAE (ß = 0.03 mm/min, p = 0.01). Compared with a historical control group, the simulation added 3.5 ± 1.9 min to the procedure. CONCLUSION: The validated simulation tool showed acceptable speed and accuracy in predicting the size and shape of hepatic RFA ablation zones. Further randomized controlled trials are needed to evaluate to what extent this tool might improve patient outcomes. KEY POINTS: • More reliable, patient-specific intra-procedural estimation of the induced RFA ablation zones in the liver may lead to better planning of the safety margins around tumors. • Dedicated real-time simulation software to predict RFA-induced ablation zones in patients with liver malignancies has shown acceptable agreement with the follow-up results in a first prospective multicenter trial suggesting a randomized controlled clinical trial to evaluate potential outcome benefit for patients.
OBJECTIVES: To evaluate the accuracy and clinical integrability of a comprehensive simulation tool to plan and predict radiofrequency ablation (RFA) zones in liver tumors. METHODS: Forty-five patients with 51 malignant hepatic lesions of different origins were included in a prospective multicenter trial. Prior to CT-guided RFA, all patients underwent multiphase CT which included acquisitions for the assessment of liver perfusion. These data were used to generate a 3D model of the liver. The intra-procedural position of the RFA probe was determined by CT and semi-automatically registered to the 3D model. Size and shape of the simulated ablation zones were compared with those of the thermal ablation zones segmented in contrast-enhanced CT images 1 month after RFA; procedure time was compared with a historical control group. RESULTS: Simulated and segmented ablation zone volumes showed a significant correlation (ρ = 0.59, p < 0.0001) and no significant bias (Wilcoxon's Z = 0.68, p = 0.25). Representative measures of ablation zone comparison were as follows: average surface deviation (absolute average error, AAE) with 3.4 ± 1.7 mm, Dice similarity coefficient 0.62 ± 0.14, sensitivity 0.70 ± 0.21, and positive predictive value 0.66 ± 0. There was a moderate positive correlation between AAE and duration of the ablation (∆t; r = 0.37, p = 0.008). After adjustments for inter-individual differences in ∆t, liver perfusion, and prior transarterial chemoembolization procedures, ∆t was an independent predictor of AAE (ß = 0.03 mm/min, p = 0.01). Compared with a historical control group, the simulation added 3.5 ± 1.9 min to the procedure. CONCLUSION: The validated simulation tool showed acceptable speed and accuracy in predicting the size and shape of hepatic RFA ablation zones. Further randomized controlled trials are needed to evaluate to what extent this tool might improve patient outcomes. KEY POINTS: • More reliable, patient-specific intra-procedural estimation of the induced RFA ablation zones in the liver may lead to better planning of the safety margins around tumors. • Dedicated real-time simulation software to predict RFA-induced ablation zones in patients with liver malignancies has shown acceptable agreement with the follow-up results in a first prospective multicenter trial suggesting a randomized controlled clinical trial to evaluate potential outcome benefit for patients.
Authors: M J van Amerongen; P Mariappan; P Voglreiter; R Flanagan; S F M Jenniskens; M Pollari; M Kolesnik; M Moche; J J Fütterer Journal: Int J Comput Assist Radiol Surg Date: 2021-05-11 Impact factor: 2.924