Dose-shaping using targeted sparse optimization.

Area of Science:

• Radiation Oncology
• Medical Physics
• Computational Biology

Background:

• Dose volume histograms (DVHs) are standard for radiation therapy plan quality assessment.
• DVHs lack spatial information, necessitating manual review of dose maps for phenomena like hot spots.
• Controlling spatial dose distribution, crucial for clinical outcomes, has been underexplored.

Purpose of the Study:

• To introduce a novel objective function for radiation therapy planning.
• To enable direct tradeoff between target coverage, organ-at-risk (OAR) sparing, and planning target volume (PTV) homogeneity.
• To illustrate the method's advantages and applicability in various cancer cases.

Main Methods:

• Developed an energy minimization objective function (E tot (sparse)) using robust cost functions.
• Incorporated an asymmetric linear well for PTV dose penalties and a two-piece linear function for OAR dose penalties.
• Utilized total variation energy (L1 norm of dose gradient) for PTV homogeneity and compared with conventional quadratic objectives.

Main Results:

• The E tot (sparse) method demonstrated flexibility in balancing OAR sparing and PTV homogeneity.
• Plans optimized with E tot (sparse) showed improved PTV homogeneity and OAR sparing (e.g., spinal cord, parotid glands) in pancreas and head-and-neck cases.
• E tot (sparse)-optimized plans exhibited superior deliverability compared to clinical plans, assessed by the 2D modulation index.

Conclusions:

• The proposed formulation enhances dose-shaping and OAR sparing, particularly for complex sites like the pancreas and head and neck.
• The linear programming nature of the objective function ensures fast convergence to the global minimum and ease of clinical implementation.
• This method is expected to positively impact radiation therapy plan quality for challenging disease sites.