[Correction of movements in cardiac SPECT studies]

Patient movement during SPECT acquisition of images is a frequent cause of artefacts in myocardial perfusion SPECT studies. In this paper we propose an algorithm for the correction of longitudinal movements.

The method is based on the calculation of the correlation between unidimensional functions obtained from successive projections with a window obtained automatically to eliminate unwanted data points before calculating the correlation. The algorithm was assessed on studies obtained from a cardiac phantom and from those corresponding to ten patients. Seven of these patients had significant movement during SPECT acquisition and those corresponding to the other three studies, in which no movement was observed, were used as a control.

In the phantom studies obtained, the mean error of the calculated displacements was less than 0.5 pixels when the window proposed was employed. The error reached 2.8 pixels when the length of the window increased. In patient studies, the method succeeded in reducing the patient motion in all cases and the three control studies remained unchanged. When no window was used to limit the correlation, three out of the seven studies that originally presented movements did not improve and one control study got worse.

The results obtained indicate the utility of the algorithm and demonstrate the need to window the data before calculating the correlation.