Biomechanics-driven dose stress metrics for radiation-induced acute xerostomia prediction among head and neck radiation therapy.

Xerostomia is a condition commonly affecting patients subjected to radiation therapy (RT) for head and neck cancer (HNC) treatment. We propose dose metrics that consider the stress of parotid glands (PGs) during RT by using finite element analysis (FEA) of structural changes captured via computed tomography (CT) images acquired before and during RT to evaluate their effectiveness in predicting acute xerostomia. Thirty patients treated for HNC via in volumetric modulated arc therapy were enrolled. Patient complaints were considered by radiation oncologists based on the common terminology criteria for adverse events and scored as xerostomia grade 0 (XG-0), XG-1, or XG-2. All patients underwent CT both before and during RT (CT_ini and CT_bst, respectively). FE-based deformable image registration was performed from the CT_ini images to the CT_bst images, following which the stress of PGs was calculated and generate the dose-stress histograms (DSH). Four classical indices (volume change, mean dose, CT value change in PGs, and weight change), the mean stress, dose-volume histogram (DVH), and DSH metrics were used to evaluate the effectiveness of our approach. No significant differences among patients w/wo acute xerostomia groups were noted based on the four classical indices, mean stress, or DVH metrics; however, DSH metrics presented significant differences (p < 0.05) and demonstrated good predictive performance in distinguishing patients w/wo acute xerostomia (AUC > 0.70). The proposed metrics can analyze stress values without additional examinations and demonstrate significant differences between groups w/wo acute xerostomia and between different XGs.