Uncertainty Estimation of the Dose Rate in Real-Time Applications Using Gaussian Process Regression

Area of Science:

• Medical Physics
• Radiation Dosimetry

Background:

• Standard organizations recognize the need for uncertainty reporting in dose rate estimations.
• Estimating uncertainties in complex radiation environments, particularly for real-time dosimetry, presents significant challenges.

Purpose of the Study:

• To implement Gaussian process regression for dose rate estimation using a spectrum-to-dose conversion operator.
• To address and quantify uncertainties in dose rate estimations across diverse irradiation geometries.

Main Methods:

• Utilized Gaussian process regression (GPR) with a spectrum-to-dose conversion operator (G(E) function).
• Evaluated the GPR approach across various irradiation geometries to assess dose rate estimation and uncertainty.

Main Results:

• The proposed GPR method provides dose rate estimations as probability distributions in a single measurement.
• Mean dose rate values from the GPR method more closely approximated true values compared to traditional G(E) functions.
• 95% confidence intervals from the GPR method encompassed both conservative estimates and true values for energies between 50-3000 keV.

Conclusions:

• The GPR-based approach enhances real-time dosimetry applications by providing reliable uncertainty quantification.
• The method aligns with operational quantity concepts, offering conservative estimates while improving result reliability.
• This technique offers a more robust solution for dose rate estimation under varying radiation conditions.