A COMPARISON OF BONE SPECT IMAGE QUALITIES OBTAINED FROM OSEM AND FBP ALGORITHMS

Abstract

The aim of the study was to compare the image qualities of bone SPECT obtained from ordered subset expectation maximization (OSEM) and filtered-backprojection (FBP) image reconstruction algorithms for bone SPECT imaging. This study was divided into two parts; the optimal numbers of subset and iteration of OSEM using a Zubal brain phantom were firstly determined. The projection datasets of the phantom were generated with 128x128 matrix sizes and 120 views over 360 degrees. Poisson noise and collimator-detector blur were added to the projection data for simulating the clinical condition. Then OSEM image reconstruction was performed with various numbers of subset and iteration. The minimum mean square error (MSE) was used as an index of the optimal parameter. The second part was to evaluate the image qualities of bone SPECT images obtained from OSEM compared to that from FBP. Thirty-two patients with lower back pain requested for bone SPECT imaging were used. A pilot study was conducted to find the preferable algorithm using a preference study. Two nuclear medicine physicians with experiences in bone SPECT images were participated. The frequencies of the scores and the agreement between readers were investigated. The results showed that the combination of the 6" subset and 2" iteration gave the minimum MSE and less time-consuming. For preference study, two readers preferred OSEM to FBP with the agreement of 65.6%. In conclusion, the optimal numbers of subset and iteration of OSEM were 6 and 2 respectively. The OSEM algorithm gave better image quality of bone SPECT than FBP.