Automated population-based planning for whole brain radiation therapy

Area of Science:

• Radiation Oncology
• Medical Physics
• Image Processing

Background:

• Whole-brain radiation therapy (WBRT) planning is time-consuming due to repetitive tasks.
• Current methods can lead to suboptimal treatment plans, risking underdosing or excessive irradiation of normal tissues.

Purpose of the Study:

• To develop and validate an automated method for WBRT treatment planning.
• To improve the efficiency and quality of WBRT planning through automatic segmentation and database matching.

Main Methods:

• Automatic segmentation of target and critical structures using morphological operations.
• Validation of segmentation against manual segmentation using Dice coefficient and Hausdorff distance.
• Treatment plan generation via database searching of similar patient anatomies, ranked by a custom metric, and modification of best-match plans.

Main Results:

• High accuracy in automated segmentation (e.g., Dice coefficient of 0.97 for the brain).
• Significant reduction in maximum doses to critical structures like eyes and lenses compared to clinical plans.
• Automated plan generation time reduced to 3-4 minutes.

Conclusions:

• Automated database-based matching is a viable alternative to traditional WBRT planning.
• The proposed method improves treatment plan quality and offers a cost-effective solution.
• This approach enhances patient safety by better protecting critical normal tissues.