Purpose: To predict the marker positions using deformation vectors from the registration of 4D-CTs of patients who may have had multiple tumors treated over different dates and compare the predicted positions with the reference positions
Methods: Two sets of 4D-CTs were acquired from a patient receiving robotic SBRT for separate lung tumors in 2017 and 2019. Deformation vector fields (DVFs) were acquired by B-spline registration between the expiration phases in 2017 and 2019, and the expiration and the inspiration phase in 2017. Two hypothetical markers were inserted in the 4D-CTs in 2017 to the positions acquired from the registration of 4D-CTs in 2019 and the 4D-CTs of 2017 were deformed using the DVFs acquired above. A scaling factor was applied to compensate the difference between the breathing amplitudes in 2017 and 2019 and it was determined iteratively. The center of masses of the predicted markers from 2017 and the reference markers from 2019 were identified and the Euclidean distances between two points were calculated and compared.
Results: The 4D-CTs were reconstructed using the DVFs and the positions of hypothetical markers were acquired. The average distance between the center of masses of prediction markers and reference markers was 7.8 mm (Marker A: 7.5 mm, Marker B: 8.0 mm). After a scaling factor was applied, the distance was reduced to 2.0 mm (Marker A: 1.1 mm, Marker B: 2.9 mm).
Conclusion: The data would suggest that the position discrepancy can be no more than 3.0 mm. However, more research is needed to validate this approach with more cases. Several uncertainties can be associate with these results such as registration accuracy, image quality of 4D-CTs, and DVFs variations with and without tumors and markers. This data can be very helpful when planning the fiducial marker placement for a clinical or research application.