Normal organ dose reduction by fluence map prediction plan adjustment.

Area of Science:

• Radiation Oncology
• Medical Physics
• Artificial Intelligence in Medicine

Background:

• Anatomical changes during radiation therapy (RT) necessitate plan adaptation for optimal outcomes.
• Current RT planning may lead to suboptimal dose delivery and increased risk to healthy tissues due to anatomical variations.
• Online adaptive RT (OART) aims to address these changes for improved treatment efficacy and safety.

Purpose of the Study:

• To investigate a fluence map prediction-based method (FM_PD) for rapid plan adaptation in OART.
• To assess the potential of FM_PD for improved normal tissue sparing by accounting for structural changes during treatment.
• To evaluate the dosimetric impact of FM_PD on organs-at-risk (OARs) compared to non-adapted plans.

Main Methods:

• A 2D Dense-U-Net model was trained using 93 intensity-modulated RT plans for hypopharyngeal carcinoma, incorporating PTV-specific loss.
• Planning target volumes (PTVs) and dose distributions were converted into 2D projection matrices for model training.
• Nine re-planning scenarios simulating OART were used for testing; predicted fluence maps were applied to daily CTs to calculate updated dose distributions.

Main Results:

• FM_PD significantly reduced radiation dose to normal tissues while preserving tumor coverage.
• Dose reductions were observed in critical structures: Dmax to the lens, optic nerves, and brainstem decreased by 18.67%, 19.17%, and 14.54% respectively.
• Mean total body dose decreased by 25.65%, with a notable 46.40% decrease in mean dose to adjacent lung tissue.

Conclusions:

• FM_PD provides a rapid and effective strategy for adapting RT plans in response to anatomical changes.
• This method significantly reduces radiation doses to healthy tissues, enhancing patient safety.
• FM_PD is recommended for cross-fraction adaptation in clinical OART practice compared to maintaining the initial plan.