Probabilistic definition of the clinical target volume-implications for tumor control probability modeling and...

Area of Science:

• Radiation Oncology
• Medical Physics
• Computational Biology

Background:

• Traditional clinical target volume (CTV) definitions are binary.
• Probabilistic CTVs account for the likelihood of disease spread beyond the gross tumor.
• Optimizing tumor control probability (TCP) for probabilistic CTVs remains an open challenge.

Purpose of the Study:

• To derive and optimize tumor control probability (TCP) expressions for probabilistic clinical target volumes (CTVs).
• To investigate TCP optimization under both voxel independence and dependence assumptions.
• To develop computational strategies for determining optimal radiation dose distributions.

Main Methods:

• Derived TCP expressions for independent and dependent voxel models.
• Employed exhaustive search and Lagrange multiplier theory for small-scale optimization.
• Developed multi-start linear programming and greedy heuristic strategies for larger-scale optimization.
• Maximized non-convex TCP under convex dose constraints.

Main Results:

• Optimal dose distributions differ between independent and dependent models.
• Observed phase transitions in subvolume dosing (high dose or no dose).
• Greedy strategies closely approximated multi-start solutions, with minor discrepancies in complex scenarios.
• Demonstrated a tractable heuristic for both independent and dependent models.

Conclusions:

• Probabilistic CTVs offer advantages for treatment planning.
• Developed computational methods effectively optimize TCP for probabilistic CTVs.
• The proposed heuristic provides a practical approach for optimizing radiation dose distributions, even with unknown voxel correlation functions.