Analytical and experimental FWHM of a gamma camera: theoretical and practical issues

Introduction. It is well known that resolution on a gamma camera varies as a function of distance, scatter and the camera's characteristics (collimator type, crystal thickness, intrinsic resolution etc). Manufacturers frequently provide only a few pre-calculated resolution values (using a line source in air, 10-15 cm from the collimator surface and without scattering). However, these are typically not obtained in situations resembling a clinical setting. From a diagnostic point of view, it is useful to know the expected resolution of a gamma camera at a given distance from the collimator surface for a particular setting in order to decide whether it is worth scanning patients with "small lesion" or not. When dealing with absolute quantification it is also mandatory to know precisely the expected resolution and its uncertainty in order to make appropriate corrections. Aim. Our aims are: to test a novel mathematical approach, the cubic spline interpolation, for the extraction of the full width at half maximum (FWHM) from the acquisition of a line source (experimental resolution) also considering measurement uncertainty; to compare it with the usually adopted methods such as the gaussian approach; to compare it with the theoretical resolution (analytical resolution) of a gamma camera at different distances; to create a web-based educational program with which to test these theories. Methods. Three mathematical methods (direct calculation, global interpolation using gaussian and local interpolation using splines) for calculating FWHM from a line source (planar scintigraphy) were tested and compared. A NEMA Triple Line Source Phantom was used to obtain static images both in air and with different scattering levels. An advanced, open-source software (MATLAB/Octave and PHP based) was created "ad hoc" to obtain and compare FWHM values and relative uncertainty. Results and Conclusion. Local interpolation using splines proved faster and more reliable than the usually-adopted Gaussian interpolation. The proposed freely available software proved effective in assessing both FWHM and its uncertainty.