KEY POINTS
- This retrospective study included 22 patients previously treated with fiducial-based dynamic tumor tracking: 11 lung cancer and 11 locally advanced pancreatic cancer cases. Ten respiratory phases from treatment-planning 4D CT were used to compare tumor motion relative to fiducials and the diaphragm.
- For lung tumors, mean three-dimensional tumor–fiducial displacement was only 1.0 mm, compared with 4.6 mm for tumor–diaphragm displacement and 5.8 mm for uncorrected free-breathing tumor displacement (p<0.001 for fiducials versus both).
- In pancreatic cancer, corresponding mean displacements were 1.6 mm, 4.5 mm, and 3.2 mm, respectively. Tumor–diaphragm displacement was therefore greater than free-breathing centroid motion in this cohort (p=0.03), demonstrating that diaphragm correction can itself introduce error if synchrony is assumed.
- The 95th-percentile tumor–diaphragm displacement reached 14.1 mm in lung cancer and 13.3 mm in pancreatic cancer, compared with only 3.0 and 3.3 mm for tumor–fiducial displacement.
- The dominant axis of diaphragm–tumor asynchrony differed by disease: variation was largest in the superior-inferior direction for lung tumors but in the anterior-posterior direction for pancreatic tumors, with many deviations arising during inspiration.
- Tumor–diaphragm distance at end expiration modified the displacement pattern, with significant three-way interaction between distance, motion axis and tumor type (p=0.002). This argues against applying one fixed diaphragm-to-tumor offset throughout the breathing cycle.
- The analysis represents a single 4D CT snapshot and therefore does not quantify inter-fraction or full intra-fraction variability. Patients with motion-blurred diaphragm images were excluded, potentially selecting for more regular breathers and making the observed performance optimistic.
CLINICAL TAKEAWAY
The diaphragm is attractive as a non-invasive tracking surrogate, but it cannot simply replace an implanted fiducial using a fixed geometric offset. Markerless diaphragm-based tracking will probably require patient-specific, respiratory-phase-dependent correction and repeated recalibration before it can approach fiducial accuracy.