AI-driven methods for optimizing and accelerating radiotherapy treatment planning.Radiotherapy AI planning involves using artificial intelligence to automate and optimize the creation of radiation treatment plans for cancer patients. Traditionally, medical physicists manually adjust parameters like beam angles, intensities, and doses to target tumors while sparing healthy organs—a process that can take hours or days. AI-driven planning aims to reduce this to minutes while maintaining or improving plan quality. This is critical because radiotherapy is a balancing act: too little radiation fails to kill the tumor, too much damages surrounding tissue.
The application of AI to radiotherapy planning began in earnest in the late 2010s, with early work on dose prediction using convolutional neural networks. The field has since grown rapidly, with commercial systems now available.
The AI is trained on historical treatment plans that were manually created by experts. Given a new patient's CT scan and the contours of the tumor and organs at risk, the model predicts the optimal dose distribution or directly generates the machine parameters (like multi-leaf collimator positions). Some approaches use reinforcement learning to iteratively refine a plan based on a reward function that penalizes overdosing healthy tissue. The result is a plan that can be reviewed and approved by a human physicist, but generated in a fraction of the time.
AI planning is used in clinics to create intensity-modulated radiation therapy (IMRT) and volumetric modulated arc therapy (VMAT) plans for brain, prostate, and head-and-neck cancers, significantly reducing planning time from hours to minutes.
Some AI planners can generate a treatment plan that is indistinguishable in quality from a human expert's plan, but in less than one minute, allowing for same-day adaptive radiotherapy where the plan is adjusted to daily anatomical changes.


























