Integration of Adaptive Proton Therapy in Pediatric Solid Tumors and Hodgkin's Lymphoma

Pediatric patients receiving proton therapy for solid tumors or Hodgkin's lymphoma may experience anatomical changes during treatment that can affect proton therapy accuracy. This prospective single-arm study uses regular low-dose imaging to monitor these changes and adjust treatment plans as needed. Participants will receive weekly or every-other-week CT scans, with MRI when appropriate, to assess whether the original plan remains accurate. Treatment plans will be updated if tumor coverage decreases by more than 5% or if radiation dose to normal tissues increases by more than 10%; otherwise, the original plan will continue. The study aims to determine how often plan adjustments are needed and to identify which disease sites are most likely to experience significant anatomical changes during treatment. Primary Objective: * Define the frequency of replanning necessary to ensure tumor coverage never falls below 95% (or 5% drop) of the prescribed daily dose in participants with intact (gross) tumors to keep the tumor control optimal throughout the multi-week treatment regimen. * Define the frequency of replanning necessary to ensure organs-at-risk (critical organs) do not deviate by more than 10% of the initially approved dose constraints to keep the normal tissue complication minimal throughout the multi-week treatment regimen. Secondary Objectives * Establish a cone beam CT (CBCT)-based framework for quantifying body surface changes throughout the treatment course. This goal will be achieved by developing a novel algorithm that detects and tracks external anatomical variations longitudinally, without requiring CBCT image enhancement, enabling precise assessment of daily participant setup consistency and anatomical stability. * Overcome daily CBCT quality limitations by generating synthetic CT images that accurately represent daily anatomy and support proton dose recalculation or verification planning. This goal will be achieved by developing a hybrid pipeline that integrates deep learning models with the deformable image registration algorithm, trained and validated on disease site-specific data. This will enable precise dose mapping and tissue density estimation, directly supporting adaptive planning decisions without the need of diagnostic- quality CT images.