The impact of motion correction on quantitatively measured perfusion indices using cardiac-dedicated dynamic CZT-SPECT myocardial perfusion imaging.

backgroundDynamic myocardial perfusion imaging (MPI) with cadmium-zinc-telluride (CZT) cardiac-dedicated single-photon emission computed tomography (SPECT) enables the quantification of myocardial blood flow (MBF), myocardial flow reserve (MFR), and non-invasive fractional flow reserve (niFFR). However, motion was a prominent challenge during dynamic acquisition that could distort these indices. This research aims to characterize the motion vector, evaluate the impact of motion correction (MC) on quantitative perfusion indices, compare MC and non-motion correction (NMC) datasets, and assess their interchangeability.

methodsWe retrospectively analyzed 149 patients who underwent clinically indicated rest/adenosine-stress dynamic MPI. All listmode SPECT and CT attenuation data were transferred to the MyoFlowQ workstation, where the listmode data were rebinned into 18 frames, yielding paired MC and NMC datasets for both phases. Frame-by-frame post-MC vectors (X, Y, Z) were extracted from the MC output to characterize motion magnitude and direction. Paired difference, association, functional bias, agreement, accuracy, and reclassification with association (MFR/niFFR) were assessed between MC and NMC. Moreover, the impact of MC on subgroups was also examined.

resultsStatistically significant frame-wise motion was present in all axes, with predominantly leftward (X: 57.42%/70.86%), posterior (Y: 61.14%/80.47%), and inferior (Z: 83.84%/79.61%) motion in rest/stress phases affecting indices. Overall, the rest MBF (0.01–0.02 ml/min/g) was statistically underestimated (except LAD), while stress MBF (0.05–0.08 ml/min/g), MFR (0.09–0.12), and niFFR (0.02–0.03) were statistically overestimated (except stress MBF-RCA). However, bidirectional estimations were observed for all indices with both increasing and decreasing values depending on the underlying error pattern. The association between MC and NMC values was high (> 0.9), but concordance ranged from 0.8 to 0.93, indicating they are not interchangeable. After reclassification, the proportion of normal scan rates increased in MFR (up to 8%) and niFFR (up to 11.4%). Body mass index (BMI) was the only subgroup factor significantly associated with MC-NMC differences in stress MBF and MFR.

conclusionsThe frame-wise directional motion pattern leads to under- and overestimation. MC influences the systematic bidirectional changes and alters MFR/niFFR classification, making MC and NMC datasets non-interchangeable, which suggests a reduction in false-positive/negative results.