Efficient uncertainty quantification for Monte Carlo dose calculations using importance (re-)weighting.

Area of Science:

• Medical Physics
• Computational Physics
• Radiotherapy Physics

Background:

• Proton intensity-modulated particle therapy (IMPT) offers high precision but is sensitive to range and set-up uncertainties.
• Accurate dose calculation using Monte Carlo (MC) simulations is computationally intensive, hindering robust uncertainty quantification.
• Existing methods for dose uncertainty quantification often rely on sampling error scenarios, which is challenging with complex MC simulations.

Purpose of the Study:

• To develop an efficient uncertainty quantification method for range and set-up errors in MC dose calculations.
• To approximate uncertainties from dynamic influences like interplay effects using error correlation models.
• To reduce the computational burden of dose uncertainty quantification in IMPT.

Main Methods:

• An importance (re-)weighting method was implemented within MC history scoring.
• This method concurrently estimates error scenarios, expected dose, and variance from a single simulation set.
• A multivariate Gaussian input and uncertainty model was utilized, allowing for various correlation models and adjustable accuracy trade-offs.

Main Results:

• The method was validated using the TOPAS MC code for proton IMPT.
• High accuracy was achieved for set-up uncertainties (e.g., 99.01% for expected dose, 98.04% for variance).
• Sufficient agreement was obtained for range uncertainties, with pass rates up to 99.39% (expected dose) and 93.70% (variance), and CPU time decreased by over an order of magnitude.

Conclusions:

• The proposed method provides an efficient and accurate approach for quantifying dose uncertainties in IMPT.
• It significantly reduces computational time compared to traditional methods.
• This facilitates the use of high-dimensional uncertainty models, enhancing treatment safety and reliability in proton therapy.