Room: ePoster Forums
Purpose: Using Organ dose prediction models to quantitatively assess treatment plan quality for Head and Neck Cancer.
Methods: Forty treatment plans for patients treated definitively with VMAT for oropharyngeal squamous cell carcinoma were retrospectively reviewed: 20 plans were randomly selected from June 2014 through July 2015 (2014) and 20 plans were from January 2016 through August 2016 (2016). For all patients, 70Gy was prescribed to the PTVHD and 56Gy was prescribed to the PTVLD. The dosimetric endpoints for Organs-at-Risk (OARs) for each patient were collected. The models from a commercial product, PlanIQ, were used to predict OAR dose according to the 3D anatomical relationship of PTVHD, PTVLD, and OARs while taking into account radiation beam characteristics. Actual plan dose-volume endpoints were compared to PlanIQ predicted dose-volume endpoints using Spearmanâ€™s rank-order correlation.
Results: Comparing the two patient groups (2014 vs. 2016), Wilcoxon rank sum test shows that the 2016 group has significantly (p<0.05) lower maximum doses to the brainstem and spinal cord, lower mean doses to the larynx, and borderline (p< 0.07) lower mean doses to the parotid glands. Other OARs were not significantly different between the groups. The 2016 group shows a higher correlation for all OARs analyzed except for the spinal cord. The highest correlation was: 0.5457 for the brainstem, 0.7273 for the larynx, 0.9053 for the oral cavity, 0.6534 for the left parotid, and 0.8898 for the right parotid.
Conclusion: OAR dose prediction models such as PlanIQ can be used to quantitatively assess plan quality in order to achieve the maximum OAR sparing. Our VMAT planning objectives have improved between 2014 and 2016, translating into significantly improved OAR sparing of the larynx, parotid glands, brainstem and spinal cord with higher correlation to PlanIQ dose prediction models.