Cone Beam Computed Tomography (CBCT)-Based Online Adaptive Radiotherapy for Patients With Metal Implants: An Institutional Experience

. The impact of MA on AI-based auto-contouring was assessed by comparing automatically generated contours with expert-modified contours using the Dice similarity coefficient (DSC) and maximum Hausdorff distance (MHD). Contouring times were also compared between patients with and without implants. Results Dosimetric avoidance through the MI achieved clinically acceptable plan quality. No clinically meaningful differences in dose distribution were observed between plans with and without MA overrides, suggesting overrides may be unnecessary unless MA overlaps with target volumes. Although the presence of MI and CBCT artifacts affected auto-contouring performance, they did not compromise the feasibility of oART delivery on the Ethos system. Contouring times were comparable between implant and non-implant patients. Overall, DSC and MHD values demonstrated good agreement between auto-contoured and manually edited structures, with the exception of the bladder. Conclusions Our proposed workflow demonstrates the feasibility of CBCT-based oART for patients with MIs based on institutional experience. Additional validation may be required to meet institutional standards and policy requirements.