Development and validation of a knowledge-based model for robotic radiosurgery planning for brain lesions

Area of Science:

• Medical Physics
• Radiation Oncology
• Radiosurgery

Background:

• Achieving steep dose gradients in stereotactic planning is crucial for sparing nearby healthy tissues.
• Manual planning for single brain lesions can be challenging in optimizing dose distribution.

Purpose of the Study:

• To develop and validate a knowledge-based (KB) model for automated stereotactic planning of single brain lesions.
• To achieve plan quality comparable or superior to manually generated plans using a robotic stereotactic system.

Main Methods:

• A knowledge-based model was trained using 60 retrospective stereotactic plans.
• Relationships between Planning Target Volume (PTV) radius and isodose volumes were established to create patient-specific dose shells and optimization templates.
• Model performance was validated internally (15 cases) and externally (13 cases), comparing Conformity Index (CI), Dose Gradient Index (DGI), and healthy brain dose with clinical plans.

Main Results:

• The KB model demonstrated high predictive accuracy (≥0.98) for all isodose levels.
• KB plans showed significantly steeper dose fall-off (higher DGI) and reduced irradiated healthy brain volume compared to manual plans.
• Fewer beams were required for KB plans, indicating increased efficiency.

Conclusions:

• The developed knowledge-based model enables automated stereotactic planning.
• This automated approach ensures high plan quality, efficiency, and standardization in radiosurgery.
• The KB model offers a promising solution for optimizing dose delivery in brain lesion treatment.