Charland P, el-Khatib E, Wolters J
Department of Physics and Astronomy, University of British Columbia, Canada.
Med Phys. 1998 Feb;25(2):152-60. doi: 10.1118/1.598176.
We present a method for obtaining the line spread function (LSF) of any radiation detector from measured data. The problem of finding a LSF is essentially a discrete deconvolution from known values of the input (Monte Carlo generated data) and the output (measured data) which can be put into matrix form. We applied the total least squares (TLS) method which is particularly useful when there are errors in both the input and output data. Results from computer simulation as well as from actual data are shown. In a practical application, however, our technique is currently limited by the ability of the Monte Carlo data to simulate correctly the inherent data from the head of the linear accelerator (linac). To overcome this difficulty we have solved by deconvolution and TLS for a more realistic inherent beam profile of our linac using the information from both profile data as measured with film and the film densitometer response function. The LSF of the densitometer was estimated with a simple method of direct measurement of a slit image and a full width at half maximum (FWHM) of 0.997 mm was recorded. Additionally, using the knowledge of this realistic inherent profile of the linac, a blurring function representing the finite source size effect missing in our current Monte Carlo profile simulation was determined. Finally, with the realistic inherent beam profile we have applied the deconvolution and TLS method to find a LSF for the Markus chamber and found a resulting FWHM of 5.39 mm. The TLS approach for deconvolving can find a useful application for both finding the LSF and correcting for the detector size effect once its LSF is known. This type of correction is required when a high spatial resolution is needed (e.g., in small field off-axis measurements). Convolved and measured profiles are also presented to illustrate the effect of the blurring due to different LSFs.
我们提出了一种从测量数据中获取任何辐射探测器的线扩展函数（LSF）的方法。寻找LSF的问题本质上是一个基于输入（蒙特卡罗生成数据）和输出（测量数据）已知值的离散去卷积问题，可将其转化为矩阵形式。我们应用了总体最小二乘法（TLS），当输入和输出数据都存在误差时，该方法特别有用。展示了计算机模拟结果以及实际数据结果。然而，在实际应用中，我们的技术目前受到蒙特卡罗数据正确模拟直线加速器（直线加速器）头部固有数据能力的限制。为克服这一困难，我们利用胶片测量的轮廓数据和胶片密度计响应函数的信息，通过去卷积和TLS求解了直线加速器更真实的固有束轮廓。用一种简单的直接测量狭缝图像的方法估计了密度计的LSF，记录的半高宽（FWHM）为0.997毫米。此外，利用直线加速器这种真实固有轮廓的知识，确定了一个表示我们当前蒙特卡罗轮廓模拟中缺失的有限源尺寸效应的模糊函数。最后，利用真实的固有束轮廓，我们应用去卷积和TLS方法找到了马克斯腔室的LSF，得到的FWHM为5.39毫米。用于去卷积的TLS方法在找到LSF以及一旦其LSF已知时校正探测器尺寸效应方面都能找到有用的应用。当需要高空间分辨率时（例如在小场离轴测量中），需要这种类型的校正。还给出了卷积和测量的轮廓，以说明不同LSF引起的模糊效应。
