Evaluation of deformable image registration vs offline adaptive replanning in post-op oral cavity cancer treated with volumetric modulated Arc therapy

Anatomical changes during adjuvant radiotherapy (RT) for post-operative oral cavity squamous cell carcinoma (OCSCC) have greater dosimetric effects with highly conformal techniques, thereby necessitating plan adaptation. We compared conventional adaptive RT (cART) with deformable image registration (DIR)-based ART (dART). Post-operative OCSCC patients receiving adjuvant RT with a ≥ 5 mm change in skin contour on cone beam computed tomography (CBCT) were enrolled. After re-simulation, planning CT (pCT) and extended CBCT were imported into Velocity^® to generate a synthetic CT (sCT) using DIR. Separate plans were created on the re-simulation CT (rCT) and sCT. The initial plan he was also projected onto the sCT anatomy and compared with the original plan on the pCT. Geometric (volume, Dice similarity coefficient [DSC], mean distance to agreement [MDA]) and dosimetric (mean dose, dose-volume histogram [DVH]) differences were evaluated between rCT- and sCT-based plans. Twenty-five patients were prospectively enrolled. Based on DSC and MDA, DIR showed acceptable geometric accuracy for all structures except the spinal cord (DSC = 0.75). Compared to pCT, most structures had significant volume reduction on sCT, except the low-risk planning target volume (PTV_LR) (p = 0.14) and the larynx (increased, p = 0.04). Projection of the initial plan onto sCT revealed significant loss in PTV_LR coverage (V_95% p = 0.001; D_98% p = 0.01) and a non-significant loss for the high-risk PTV (PTV_HR). Organs at risk (OARs) doses increased non-significantly, except for the mandible (p = 0.007). Comparison of rCT and sCT volumes showed a 3.7% increases in PTV_HR (p = 0.009) and 5.7% in PTV_LR (p = 0.049), with non-significant OAR volume increases (2.4%-6.3%) except for the larynx (decreased). DVH comparison showed non-significant dose reductions to the parotids (0.6%), the mandible (0.5%), and the larynx (5.8%), but slight increases for the spinal cord (2.6%) and its planning organ at risk volume (PRV) (3.7%). Target coverage was significantly lower with sCT-based plans (PTV_HR V_95% by 3.2%, D_98% by 3.8%, PTV_LR V_95% by 4.1% and D_98% by 4.3%). dART improved OAR sparing but was compromised in a few target coverage parameters. Further enhancement of DIR accuracy is needed. Currently, DIR is primarily used in ART for dose accumulation to assess the need for plan adaptation.