My institution has not been doing frameless SRS as long as yours, but we have had a very similar experience. We’ve reached out to an number of other clinics and the issue of real-time delta deviations for larger couch rotations is fairly common. There seems to be two interrelated issues:
- As the patient rotates, the view that the cameras have of the tracking surface changes. This results in growth in the real-time delta readouts at extreme couch rotations (near 90 and 270). Like you, our imaging typically shows that there is no real deviation. The deviations are usually in the longitudinal and lateral directions. For us, they are usually on the order of 1 mm or less, but we have seen 1.5 mm on a couple occasions.
- As the gantry rotates, the camera opposite of the couch rotation direction can become blocked. Depending upon the patient and couch angle, this can result in a change in real-time deltas. We often see changes on the order of 0.5 to 1 mm.
In speaking with other clinics, many have noted that the phenomenon has an isocenter depth dependence too: When isocenter is more posterior (deeper relative to the surface), the deviations are larger.
The clinics I have spoken to have responded differently to the issue. One clinic has chosen to limit their couch rotations to +/- 45 degrees to avoid problematic angles. Some repeat their CBCT or orthogonal imaging for all deviations that result in an out-of-tolerance level. Initially, we performed AP kV planar imaging at the problematic couch angle to verify patient position. If shifts were not required, we capture a VRT at that angle and continue treatment. Lately, we have transitioned to driving the couch back to zero degrees to verify the position rather than using imaging, as the real-time deltas at couch zero had corresponded well to what we see on imaging.
To my knowledge, no one has been able to eliminate the issue entirely. Vision RT is set to release an advanced camera optics calibration, I believe later this autumn. They have some beta test sites who already have the calibration, and these sites report significant improvement. From what I’ve heard, it sounds like the new calibration cuts the real time delta deviations by about half and significantly reduces the effect of one of the camera pods getting occluded.
We have also had Vision RT remote into our system to review our tracking ROIs and observe treatment. They have made some recommendations regarding the ROIs that have reduced the frequency of large deviations. One thing that is especially important is to review your ROI on the VRT capture used for treatment to make sure it transferred from the DICOM surface with high fidelity. If the ROI splashes onto the mask, you can get some unexpected deviations. For us, sometimes the top of the nose ends up looking very jagged or spiky, and we will remove the ROI here. I would recommend having the vendor remote in at least once to look for something easy to fix.