Real-time intrafraction 3D spine tracking using gantry-mounted kV imaging for single fraction high dose functional...

Spinal nerve root ablation of 70-90 Gy with single fraction stereotactic radiosurgery (SRS) delivered in short partial arcs (30-60 degrees) is an emerging treatment for managing chronic thoracoabdominal wall pain. The proximity of the target to spinal cord and the high per fraction dose demand intrafraction motion monitoring with sub-millimeter accuracy. Pre-treatment cone-beam CT (CBCT) and intermittent room-mounted triggered kV imaging do not provide continuous internal anatomical verification during beam delivery. Gantry-mounted real-time tracking solutions have been mostly validated for full arc stereotactic body radiation therapy (SBRT) at substantially lower doses.

We report comprehensive evaluations of a gantry-mounted, real-time 3D kV spine tracking system based on template matching of continuously acquired fluoroscopic images. Tracking performance was assessed under the demanding conditions of short-arc, high-dose functional SRS. We also present its first clinical use during spinal nerve root ablation for chronic thoracoabdominal wall pain.

We first performed pre-clinical evaluation with an anthropomorphic spine phantom placed on a programmable motion platform. Tracking performance was evaluated at incremental phantom offsets from -4 to 4 mm and varying arc lengths. The detected shifts were compared to the known offset, as well as estimated shifts from CBCT and room-mounted triggered kV imaging system. Moreover, the tracking accuracy was validated clinically in 12 spine SBRT patients using CBCT projections acquired as part of the routine care. The detected shifts were benchmarked against the CBCT-derived shifts, with offline re-analyses at reduced arc lengths. Finally, under an IRB-approved protocol, the real-time tracking system was deployed as a secondary monitor in two patients (three disease sites) receiving 70 Gy frameless single-fraction SRS to the spinal nerve dorsal roots. We evaluated overall tracking performance and compared the detected shifts in vertebral position to CBCT and room-mounted triggered kV system detected shifts.

The phantom study demonstrated an average accuracy of 0.2 ± 0.1 mm (maximum: 0.6 mm) in the detected spine position for offsets up to 4 mm. No substantial impact of the arc length was observed, and the agreement with CBCT and room-mounted triggered kV system was within 0.2 mm. The clinical validation with CBCT data showed an average agreement within 0.4 mm in all directions, with maximum deviations of 0.8 mm. During clinical nerve root ablation treatments, the vertebral body was successfully tracked throughout treatment. Intrafraction motion exceeding 1 mm was detected in 2 of 3 sites by our real-time tracking system, prompting corrective CBCTs and couch shifts. The median (interquartile range) difference in the detected vertebra position compared to room-mounted triggered kV system was -0.2(0.6) mm, 0.1(0.4) mm, and -0.1(0.4) mm in vertical, lateral, and longitudinal directions.

Our preliminary study demonstrated the feasibility of vertebral body tracking using a real-time 3D kV tracking system during nerve root ablation treatments. Real-time monitoring detected actionable intrafraction motion that prompted couch corrections in 2 of 3 sites, which room-mounted triggered imaging would not have identified until the subsequent arc. This capability might be particularly valuable for short-arc, high-dose treatments, where the window for intrafraction verification is narrow, and the consequence of undetected motion is high.