Purpose: Orthogonal kilovoltage (KV) image pair is a standard image guidance approach for radiotherapy. Instead of matching treatment image to planned DRR in the traditional approach, “2D-3D” match as a newer technique reverses this patient alignment principle: In 2D-3D match, the patient dataset is rotated and translated to produce a live DRR that matches the treatment image. The treatment position is then achieved by reversing the rotation and translation performed on the dataset. 2D-3D match also allows 6 degrees-of-freedom (6DOF) alignment. There may theoretical and technical limitations of such an approach, especially when 6DOF is concerned. This study uses a single phantom to validate the accuracy and reproducibility of 2D-3D auto-match with 6DoF.
Methods: The imaging and match of a single phantom was performed on Varian TrueBeam (v2.7MR3) and ARIA (v15.6) with 6DOF couch top. Ground truth of best alignment was first established by cone-beam CT. KV pairs were then acquired to perform 2D-3D auto-match in repeated steps. Manual match was also performed after auto-match, to emulate the clinical process.
Results: It was found that 2D-3D auto-match with 6 DoF can aggravate setup errors. Repeated imaging does not produce converging results. Manual match after auto-match may not bring the image registration to the best alignment, as a user could be trapped in a local optimum set up by the preceding auto-match process.
Conclusion: 2D-3D auto-match with 6DoF on ARIA v15.6 with TrueBeam 2.7MR3 is not reliable for the single phantom used in this study; auto-match may set user to a local optimum trap that prevents better manual alignment. It is unknown that the same algorithm would perform better in other phantoms or in clinical practice. Clinical users should be aware of the theoretical and practical limitations when deploying 2D-3D auto-match with 6DoF in their clinical operations.