A Random Sampling Method for Radiobiological Modeling

Document Type

Article

Publication Title

Radiation Physics and Chemistry

Abstract

Various radiobiological models have been proposed to describe the clonogenic cell survival after radiation. Least square model fitting to match experimental survival data can be tedious and sometimes difficult for complex models with multiple parameters. In this work, a random sampling method is developed to optimize the model parameters automatically to achieve best fitting with the experimental data. A computer program is developed to read the experimental data and to calculate the cell survival fractions based on a given mathematical model. Model parameters are randomly sampled, and the model predictions are compared with the experimental data. The mean square error (MSE) is used as an objective function to drive the iterative process. The goodness-of-fit of the linear quadratic (LQ), multitarget (MT), universal survival curves (USC) and multimode (MM) model was compared between the new method and previous studies for melanoma and non–small-cell lung cancer (NSCLC) cell lines. Good fitting was obtained for the MM model with the NSCLC data (MSE = 0.0005), which was better than the fits to the LQ, MT and USC models (MSE = 0.0019, 0.0082, 0.0024, respectively). The parameters used for the LQ, MT and USC models were α = 0.33Gy, α/β = 10Gy, D0 = 1.25Gy, n = 4.22 and DT = 6.2Gy. The parameters for the MM model (n = 4) were D1 = 3.0Gy, D2 = 8.27Gy, and D3 = 2.93Gy and D4 = 10.6Gy. It is concluded that the random sampling method provides a useful tool for the parameterization of radiobiological models for radiation research and clinical applications.

DOI

10.1016/j.radphyschem.2024.111760

Publication Date

8-2024

ISSN

1879-0895

Link to Full Text

Share

COinS