AI-Based Dose Compliance of Secondary Organs at Risk in Head and Neck Cancer Radiotherapy

Area of Science:

• Medical Physics
• Radiation Oncology
• Artificial Intelligence in Medicine

Background:

• Accurate delineation of organs at risk (OARs) is crucial for effective radiotherapy planning.
• Secondary OARs are often under-reported, yet significantly impact patient toxicity.
• AI-driven auto-segmentation offers potential for improved efficiency and accuracy in OAR delineation.

Purpose of the Study:

• To evaluate the geometric and dosimetric performance of an AI auto-segmentation algorithm on primary OARs.
• To assess the added value of AI auto-segmentation for dosimetric evaluation of secondary OARs.
• To investigate the impact of AI auto-segmentation on radiotherapy planning time and OAR dose constraints.

Main Methods:

• Retrospective analysis of 50 head and neck cancer patients' VMAT plans.
• Comparison of manual versus AI auto-segmentation for 10 primary OARs using geometric metrics (DSC, HD, sensitivity, precision) and dosimetric differences.
• Dosimetric assessment of 29 secondary OARs delineated by the AI tool.

Main Results:

• AI auto-segmentation significantly reduced planning time (21.20 ± 2.25 min to 9.25 ± 1.42 min).
• Geometric performance metrics (DSC, sensitivity, precision) showed high accuracy for primary OARs.
• Significant dosimetric deviations were observed for the brainstem and right submandibular gland; 8.5% of secondary OARs exceeded dose constraints, with constrictor muscles most affected (56%).

Conclusions:

• Validated AI auto-segmentation provides a more comprehensive dosimetric evaluation of secondary OARs.
• This approach can help minimize radiation dose to sensitive structures, potentially reducing overall toxicity.
• AI tools enhance radiotherapy planning by improving efficiency and enabling thorough assessment of both primary and secondary OARs.