The signature of drifting sensitivity and centre-of-rotation in SPM analysis of multi-head gammacamera brain SPECT and a method of correction

In the course of an SPM brain SPECT analysis, significant differences were detected between 2 groups of normal controls acquired on the same 3-head gamma camera but 3 years apart. A distinctive asymmetry was evident in the distribution of focal differences in the SPM maps. The source of this asymmetry was investigated by simulating calibration errors in the projections from one camera head of 30 scans of a 90 scan normal database, reconstructing them and using SPM to detect differences between the modified 30 and the remaining 60 scans. Two different calibration errors were simulated: either an increase by 10% in count levels to simulate differential sensitivity, or an offset of 1 pixel (3.5 mm) in X or Y to simulate an incorrect centre of rotation (COR). Error simulations were performed for both 3-head and 2-head cameras. The affect of a simulated 10% increase in sensitivity in head 3 of 3 most closely resembled the observed SPM difference image and we concluded that a drift in sensitivity had caused the observed SPM differences. In general, simulated sensitivity changes in one head caused an asymmetry in the SPM map centred on the mid acquisition angle for that head for both 3-head and 2-head systems. The blurring caused by the simulation of a 1 pixel COR error in 1 head did not translate into significant SPM differences between the two populations. Although an asymmetric clinical SPM result may be valid, it should be regarded with suspicion and the beta difference image computed to check whether the asymmetry is aligned with a single camera head. After applying an inverse correction derived from the SPM beta images to the affected normal controls, a repeat SPM analysis showed no significant difference. This method may be suitable for correction and subsequent use of databases from different cameras and institutions.