Using dose volume histogram (DVH) predictions to improve the plan quality of helical tomotherapy (HT)

This study aimed to investigate whether the RapidPlan (RP) model configured by volumetric modulated arc therapy (VMAT) plans of nasopharyngeal carcinoma (NPC) could be used to assist in the optimization of HT plans and improve their quality. An RP model was trained using 100 clinically accepted VMAT plans of NPC patients. The predicted dose constraints of the VMAT trained RP model were used to reoptimize 25 consecutive clinically accepted HT plans (HT_clinical) and perform new VMAT plans based on the same computed topography (CT). The dosimetric quality of the reoptimized HT plans (HT_reoptimized), HT_clinical, and VMAT group were compared. The minimum dose encompassing 2% target (D_2%), the minimum dose encompassing 98% target (D_98%), homogeneity index (HI) and conformity index (CI) were similar for most targets between the HT_clinical and HT_reoptimized plans, although certain targets in the HT_reoptimized plans had higher D_2% and HI and lower D_98%. The HT_reoptimized plans outperformed the HT_clinical plans in the Dmax and D1cc of the spinal cord, V_40Gy of the left temporal lobe, Dmean and V_30Gy of the oral cavity, Dmean of the larynx and thyroid, and the differences were statistically significant. HT plans had higher CI and HI than VMAT plans. HT plans outperformed VMAT plans in the Dmax of the spinal cord and lenses, V_30Gy of the oral cavity and parotids, and V_40Gy of the temporal lobes, but underperformed in the Dmax and D1cc of the brainstem, D1cc of the spinal cord and Dmean of the oral cavity. The VMAT-based RP model can be used to assist in the planning of HT plans and improve the dosimetry quality of HT plans.