| 摘要 |
A computationally efficient and accurate procedure for calculating radiation dose distributions within a volume of interest in a patient incorporating variations in patient density and source particle energy, position, direction, and type. The procedure iteratively simulates energy depositions from source particles using the Monte Carlo radiation transport method and then applies methods for reducing noise inherent in Monte Carlo results to obtain an accurate and computationally efficient representation of an actual radiation dose distribution compared to that produced by Monte Carlo methods alone. The invention makes an analogy between real measured data and transformed Monte Carlo generated computer data and then applies data denoising techniques to reduce the noise in the Monte Carlo dose images. Conversely, the present invention can produce a dose distribution having a predetermined noise level in a reduced amount of computation time. Denoising techniques can include digital filtering, wavelet denoising, kernel smoothing and non-parametric regression smoothing. |
