Purpose: To address challenges inherent in single-cycle respiratory-correlated CT (4DCT), several surrogate-based motion models (SMMs) have been proposed. These models utilize the a-priori correspondence between the surrogate and internal anatomy observed during the 4D-simulation. Commonly occurring changes in this correlation and extrapolation beyond their training dataset can result in model breakdown. To address such issues, we present a hybrid finite-element (FE) model with transient boundary conditions (BCs) to estimate tumor position during radiotherapy delivery.
Methods: 4DCT data, VisionRT surfaces (VRT), and ~30-50s lateral and anterior-posterior fluoroscopic projections (FLs) were collected from five lung cancer patients. A previously validated volumetric SMM that integrated a few FLs to construct and update the correlation between VRT and internal motion was used to generate transient deformations for the lung-surface (as BCs). The exhale lung and tumor topology were imported into ABAQUS multi-physics software and a FE lung-tumor model was constructed. GTV position estimation was evaluated by comparing estimated centroid and Hausdorff distance with respect to the corresponding 4DCT phase. Transient BCs were used, and lung-tumor motion was simulated during the treatment session.
Results: 95th percentile absolute difference between FE-estimated and 4DCT-delineated GTV centroid in the SI/AP/ LR directions were 2.7/2.4/1 mm respectively. Submillimetre median errors were observed for all patients. For two patients, large errors (8mm and 5 mm) were present in peak-inspiration due to GTV-delineation inaccuracies from 4DCT binning artifacts. GTV excursions outside of the ITV were observed in transient simulations, but GTV motion was mainly within the PTV. For one FL-acquisition, baseline-shifts were present, with GTV outside of the PTV more than 50% of its time.
Conclusion: Our findings indicate that the tumour can move outside the irradiated volume. Clinical implementation and application of our model can lead to improvements in more precise RT delivery and target margin accuracy.
Lung, Image-guided Therapy, Finite Element Analysis