Knowledge-Based Planning for Intensity-Modulated Proton Therapy for Patients with Locally Advanced Esophageal Cancer

Purpose: To introduce a knowledge-based planning (KBP) framework for intensity-modulated proton therapy (IMPT) that generates patient-specific dose-volume histogram (DVH) objectives using a database of geometric and dosimetric data from previously treated esophageal cancer patients. Methods: The framework employs overlap-volume histogram (OVH) to quantify spatial relationship between organs at risk (OARs) and targets. For a new patient, OVH is calculated from DICOM CT and RT-structure and compared with those in a reference database to derive individualized DVH objectives for IMPT optimization. The reference database contains pre-calculated OVHs and DVH data for targets and OARs from previously clinically delivered plans. Eighteen patients with locally advanced esophageal cancer treated with IMPT were retrospectively analyzed. KBPs were created using a leave-one-out approach: DVH objectives for each KBP were generated from the remaining 17 patients and applied in the RayStation TPS for IMPT optimization. KBPs were compared with clinical plans (CPs: clinically delivered plans, manually created by planners) in target coverage and OAR sparing. Results: KBPs achieved comparable dosimetric quality to CPs in target coverage. OAR sparing was generally similar, with a significantly reduced spinal canal dose in KBPs. Conclusions: This study demonstrates the feasibility of an OVH-driven KBP framework for IMPT in esophageal cancer. By leveraging geometric-dosimetric correlations from prior patients, the method enables the automatic generation of individualized planning objectives, achieving plan quality comparable to clinical standards.