Automated deep learning segmentation and planning for left-sided breast radiotherapy with minimised adaptations based on dose, TCP and NTCP criteria

Key Points

• This research aims to evaluate automated deep learning methods for segmentation and planning in left-sided breast radiotherapy.
• Compared four scenarios with varying deep learning segmentations and planning adaptations.
• Evaluated setups based on DVH goals and radiobiological metrics.
• Assessed clinical goal achievement in a fully automated workflow.
• 43% of plans from the deep learning workflow met all clinical goals.
• Edits to target delineation did not affect organs-at-risk.
• The new workflow decreased active planning time by 50% and reduced time between steps by 30%.