Impact of image processing in the detection of ischaemia using CZT-SPECT/CT

The new multipinhole cardiac single photon emission computed tomography/computed tomography (SPECT/CT) cameras with cadmium-zinc-telluride (CZT) detectors are highly sensitive and produce images of high quality but rely on complex dedicated reconstruction algorithms. The aim of this study was to determine the impact of various processing steps on image formation and in the detection of ischaemia in CZT-SPECT/CT both with and without attenuation correction (AC).

Data on 20 consecutive patients who underwent a 1-day protocol stress-rest SPECT/CT using 99mTc-tetrofosmin were processed twice by three experienced operators, yielding 120 AC and 120 noncorrected (NC) data sets. Processing steps included selection and determination of myocardial axes, manual SPECT/CT coregistration for AC and myocardial masking. Using the 17-segment cardiac model, differences between stress and rest segmental uptake (%) were calculated for NC and AC image sets. Both interoperator and intraoperator variations were considered significant for the diagnosis of ischaemia when greater than 5%.

The mean interoperator variations were 2.4±1.4% (NC) and 3.8±1.9% (AC) (P<0.01). In 6% (NC) and 23% (AC) of the 120 processed cases, operator variation was larger than 5% and therefore potentially clinically interfering with the diagnosis of ischaemia. Differences between interoperator and intraoperator variations were nonsignificant.

Operator variations in the processing of myocardial perfusion image data using CZT-SPECT/CT are significant and may influence the diagnosis of ischaemia, especially when AC is applied. Clearer guidelines for image processing are necessary to improve the reproducibility of the studies and to obtain a more reliable diagnosis of ischaemia.