This work focuses on brain stereotactic radiosurgery for patients with multiple metastases to identify a single "isocenter" location, the point the machine rotates around as it delivers radiation from many angles. Clinicians often pick this location manually based on experience and anatomy, and then rely on the planning system to optimize the beam shapes and intensities. That workflow can be slow, inconsistent, and sensitive to small setup errors, because the isocenter choice influences the final dose distribution in ways that are hard to predict by geometry alone. This work develops an automated approach that links these steps together, using Bayesian optimization to propose a "next-best" candidate isocenter location until one of the stopping criteria is met. By iterating this loop, the method searches for an isocenter that produces the best treatment plan quality.
