Volume calculation of the thyroid gland from SPECT images based on Monte Carlo simulations

Abstract: In this study the volume determination which is a part of the doseplanning for patients with thyrotoxicosis was investigated. The aim was to find an accurate method to determine the active volume with single photon emission tomography (SPECT) which in several studies have shown better results than with the currently used method planar scintigraphy (PS). This was implemented on the Xeleris 3.1 by General Electric (GE) at the University Hospital of Umeå (NUS). The examination time with SPECT is required not to be significantly longer than the examination time with PS. Despite the relatively short examination time, the accuracy of the volume determination should be more accurate. This was analyzed with a true value of the volume, conducted with Monte Carlo simulations of digital anthropomorphic phantoms. It is also important that the developed method is user friendly. The study included ten patients with thyrotoxicosis from which, relative activity uptakes were measured. These uptakes were specified in 22 digital XCAT phantoms which were random sampled with respect to phantom mass. The Monte Carlo program "simulating medical imaging nuclear detectors" (simind) was used to simulate the SPECT-system. The projection images were ordered subset expectation maximization (OSEM) reconstructed and Butterworth filtered. Unfiltered images were used to compare volume calculations with filtered ones. The volume of the thyroid was segmented using threshold values applied to all voxels in the image-sets and the optimization of the thresholds was conducted by numerical calculations. The results in this study shows that the best choice of intensity threshold value is 21.1(24)% of the maximum voxel value for all phantoms. The threshold is valid for OSEM iteration number five and unfiltered image-sets. Butterworth filtered images were less suitable to use than unfiltered images when the thyroid volume was calculated with data from SPECT-simulations of phantoms.