Click here to


Are you sure ?

Yes, do it No, cancel

Predicting Radiation-Induced Xerostomia Using Parotid Dose Cluster Models

M Chao1*, J Wei2, Y Lo1, R Bakst1, J Penagaricano3, (1) The Mount Sinai Medical Center, New York, NY, (2) City College of New York, New York, NY, (3) H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL


(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: To develop and validate a predictive tool for clinical decision making in head-and-neck-cancer (HNC) irradiation with the parotid dose cluster model based on retrospective xerostomia data.

Methods: hundred HNC patients treated with intensity modulated radiation therapy from three institutions were retrospectively enrolled in the study. Post-treatment follow-up data including xerostomia grades were included to divide patients into two groups: X0 – that never developed (grade 0) and X1 – that developed xerostomia (grade 1-3). Mean dose (MD) was first examined for its relevance to xerostomia. Clustering pattern from the parotid 3D dose distribution was investigated using the percolation models to decipher its correlation with the radiation induced xerostomia. The average and largest sizes of the parotid dose clusters normalized by the gland volume were computed and fitted to the Lyman-Kutcher-Burman (LKB) model with the maximum likelihood technique. The scatterplots between the fitted parameters TD50 and m were employed to differentiate the morphological forms of dose aggregation in the parotid gland. The number of clusters and the weighted cluster sizes were also evaluated against the xerostomia data to further explore the model efficacy. Three connectivity choices were specified to reach the optimal choice of the cluster model.

Results: Average MD was significantly lower in X0 (18.4Gy) than that in X1 (31.9Gy) as anticipated from the current clinical recommendation. Scatterplots of the LKB fitted parameters showed significant differentiation between these two populations, strongly suggesting that the cluster sizes act as a better predicator of the xerostomia. Similar pattern was observed in the weighted cluster size distribution. The number of clusters in X0 displayed a distribution more spread out than that in X1, again revealing the model validity.

Conclusion: The novel cluster model could serve as a superior predictor for parotid radiation induced xerostomia as opposed to the oversimplified dose-volume-histogram based models.


Not Applicable / None Entered.


TH- Radiobiology(RBio)/Biology(Bio): RBio- LQ/TCP/NTCP/outcome modeling

Contact Email