Odd vs even number of projections in 360 degree single head SPECT

Full text: It has been reported from simulation studies that the quality of reconstructed tomographic images can be improved by acquiring an odd number of projections over 360 degree rather than the (conventional) even number. This study examines whether using an odd number of projections yields the anticipated improvements in SPECT image quality. Methods: Two phantom studies were performed using a Picker Prism 1000XP, the first using a Jazczack phantom to examine reconstructed image uniformity, the second using a three line source phantom to examine reconstructed image resolution. Ten different angular sampling rates (128, 65, 64, 63, 33, 32, 31, 17,16,15 steps over 360 degree) with a matrix size of 64 x 64 and magnification 1.333 were used. Images were reconstructed with a ramp filter to a slice thickness of 7.0 mm and transferred to a Siemens ICON for analysis. Uniformity was measured from the root mean square (rms) of a 15 x 15 pixel ROI in the centre of a uniform slice. Resolution (FWHM) was measured from line profiles through the cross section of each line source. Results: There was no significant difference in the uniformity of neighbouring odd and even numbered studies (mean 6.15 + 0.51%overall). Radial resolution was almost constant down to 15 steps. Central resolution was significantly degraded in both 15 and 17 step images (+9% compared to the 16 step image). Tangential resolution was significantly degraded for most odd step counts (15, +87%; 17, +63%; 31, 26%; 33, 23%; 63, +3%; 65, 13%). Conclusion: There is no obvious advantage in performing 360 degree single head SPECT with an odd number of projections. On the contrary, there is actually a degradation in image quality