Author: James E. Bates, Benjamin Hopkins, Kirk Luca, Shadab Momin, Justin R. Roper, Soumon Rudra, Eduard Schreibmann, Bill Stokes, Tu Thi, Xiaofeng Yang 👨🔬
Affiliation: Emory University, Department of Radiation Oncology and Winship Cancer Institute, Emory University 🌍
Purpose: Swallowing dysfunctions after radiotherapy are caused by multiple factors yet are strongly associated with the irradiation of pharyngeal musculature due to its role in the initiation and completion of swallowing. This work aimed to build a dysphagia optimized knowledge-based(DO-KBP) model by considering individual constrictor muscles and evaluate DO-KBP performance against a clinical KBP model(C-KBP).
Methods: The C-KBP model was trained with 175 previously-treated cases where the entire pharynx structure was penalized during optimization. The DO-KBP model included 213 training cases where individual pharynx constrictors (upper/middle and inferior constrictors) were incorporated for cases with bilateral oropharynx cancer. We evaluated both models on 25 additional cases. Two sets of treatment plans were compared: (1) C-KBP model-generated, and (2) DO-KBP model-generated. Plans were normalized so the prescribed dose covered 95% of a PTV. A two-tailed paired-t-test was performed to evaluate the performance of DO-KBP against C-KBP model across several dosimetric endpoints. A blind review of both sets of plans was performed by a physician.
Results: Overall results show significant(p<0.05) pharynx constrictor sparing by DO-KBP. On average, mean dose to inferior pharynx constrictor was 36.52Gy±9.87Gy and 19.52Gy±6.23Gy, whereas the average mean dose to superior/middle pharynx constrictor was 51.89Gy±6.31Gy and 47.46Gy±6.12Gy by C-KBP and DO-KBP, respectively. No significant differences were seen in the mean dose to left parotid gland, larynx and submandibular glands, amongst both sets of plans (p≥0.05). Statistically significant increases for the right parotid gland were observed with the DO-KBP model(p=0.02). Both sets of plans resulted in an average homogeneity index of 0.09. The DO-KBP plans were preferred over C-KBP plans during a blind review despite the small increase in dose to the other OARs.
Conclusion: The DO-KBP model significantly reduces doses to pharyngeal constrictors, structures critical for swallowing, with only minor dosimetric tradeoffs to other healthy tissues.