High-Precision Lung Radiotherapy by Intra-treatment Dynamic Cone-beam CT imaging and Dosimetry-guided Plan Adaptation - PROJECT SUMMARY The recent advancement of stereotactic body radiation therapy (SBRT) enables highly focused dose delivery to tumors while sparing surrounding normal tissues. Both past and emerging clinical evidence has strengthened the role of radiotherapy in lung cancer management, and SBRT is considered the standard of care for many diagnoses. However, radiation-induced toxicity, especially for cases with centrally located lung tumors, poses a lingering challenge to lung SBRT. The respiration-induced motion of lung tumors and surrounding organs-at-risk (OARs) introduces substantial uncertainties to the delivery accuracy of lung SBRT, causing under-dosing to tumors and over-irradiation to surrounding OARs. The current motion management techniques, including internal-target-volume (ITV)-based treatment, respiratory phase gating, breath hold, and motion tracking, all suffer from various sources of uncertainty and inaccuracy, potentially resulting in large dose deviations that miss the tumor and damage normal tissues. Obtaining intra-treatment tumor and OAR motion/deformation, and using such information to derive the actually delivered doses to optimize remaining radiotherapy treatments, will systematically address the intra-treatment motion challenge. However, reconstructing intra-treatment dynamic and volumetric images for motion/deformation tracking remains an unmet clinical need, mostly due to the challenging spatiotemporal inverse problem of reconstructing volumetric images from extremely under-sampled signals. In addition, currently, there are no existing techniques and workflows that use the actually delivered dose to optimize future SBRT plans and deliveries. The overarching goal of this project is to develop an intra-treatment dynamic imaging and plan adaptation (IDIPA) system, which is composed of two sub-systems, spatial and temporal implicit neural representation (STINR) and dosimetry-guided plan adaptation (DGPA). STINR solves dynamic cone-beam CT (CBCT) images and intra- treatment deformation vector fields (DVFs) from x-ray projections with each x-ray projection corresponding to a dynamic CBCT volume and a DVF. The solved dynamic CBCTs and DVFs will then be used by DGPA for treatment dose calculation, dose accumulation, and dosimetry-guided plan adaptation. The STINR sub-system uses an Artificial Intelligence-driven method to address the substantial challenge of dynamic CBCT reconstruction. The DGPA sub-system features the first closed-loop, dosimetry-guided optimization framework that uses delivered doses to adapt the following plans to ensure the treatment doses will be delivered to where it is intended. We have three Specific Aims for this project: 1) Develop and optimize the dynamic CBCT imaging sub-system (STINR), 2) Develop and optimize the dosimetry-guided plan adaptation sub-system (DGPA), and 3). Evaluate the overall IDIPA system via a clinical study. The success of the project will result in the first end-to-end system to improve the dose delivery accuracy of lung radiotherapy under intra-treatment motion, and to unleash the full potential of SBRT in advancing lung cancer care.
