SU-E-J-70: A Study of the Correlation of 3D Surface Matching and KV Imaging for Chestwall IMRT

We investigated the accuracy of surface matching compared to kV positioning of 10 post-mastectomy chestwall patients receiving inverse-planned, non-gated IMRT.

During 130 treatment sessions, non-gated 3D surfaces were captured using AlignRT (v4.5) before and after table translations as indicated by kV imaging. Surfaces were compared to a reference generated from CT data. Pearson's correlations between 'indicated' AlignRT and kV shifts were studied for surface registration of user-defined regions-of-interest (ROIs): the entire surface ('all') and the chestwall ('cw'). In 21% of sessions, two consecutive surfaces of patients in identical positions were used to estimate variability. Finally, the 'implemented' shifts detected by AlignRT were calculated and compared to the absolute table translations.

Correlations between 'indicated' AlignRT and kV shifts were higher for 'cw' than for 'all': r=0.65 (Anterior-Posterior), r=0.65 (Superior-Inferier), r=0.44 (Left-Right). Correlations exhibited large inter-patient variability; 60% had r<0.6 in at least 1 direction thereby resulting in ow correlation between 3D Euclidian shift distances (r=0.48). Correlations between 'cw' and kV increased (r>0.8) for 'implemented' table shifts; only 2 patients demonstrated r<0.6 in any single direction. Comparison of consecutively-acquired 'cw' surfaces controlled for patient movement demonstrated standard deviations (STDs) of: 1.2mm (A-P), 2.0mm (S-I), 1.6mm (L-R). While STDs between 'implemented' and kV shifts were on the same order, STDs between 'indicated' and kV shifts were twice as large: 2.9mm (A-P), 3.5mm (S-I), 4.1mm (L-R). Rotational differences >1degree were calculated more often when registering 'cw' (22%) than 'all' (5%). At the treatment position, residual STDs remained high (3.0-3.6mm).

On average, 'cw' shifts correlated with kV shifts but exhibited significant inter-patient variability and larger rotations than 'all'. Differences between AlignRT and kV were ∼3mm for initial patient positioning. The lack of a one-to-one correspondence between surface and kV shifts in any single session must be further investigated before clinical implementation.