Assessment of sestamibi CZT-SPECT reconstructed using deep-learning-based virtual attenuation correction maps...

Attenuation artifacts affect the analysis of single-photon emission computed tomography (SPECT) myocardial perfusion imaging, especially in the right coronary artery (RCA) territory, although computed tomography (CT)-based attenuation corrections minimize this issue. This study aims to assess stress-rest [^99mTc]sestamibi cadmium zinc telluride (CZT)-SPECT reconstructed using deep-learning-based attenuation correction (DLAC) maps according to coronary artery territory and with comparison to [⁸²Rb]rubidium-positron emission tomography (Rb-PET) conventionally reconstructed with CT-based attenuation correction.

We compared stress-rest [^99mTc]sestamibi CZT-SPECT reconstructed without (i.e., standard (SD-SPECT)) and with DLAC (DLAC-SPECT) in 84 patients at risk of attenuation artifacts (overweight individuals and women) with no previous history of coronary artery disease (CAD), and who also had Rb-PET and, when indicated, coronary angiography.

In 74 very low CAD-risk patients: (i) the stress segmental activity values were homogeneous between the three coronary artery territories with DLAC-SPECT and Rb-PET but exhibited a marked reduction in the RCA territory with SD-SPECT (69.1% ± 8.2% vs 81.2% ± 7.1% and 77.7% ± 8.5% for DLAC-SPECT and Rb-PET, respectively, both P < .001), and (ii) the absolute value of the difference between the rest and stress segmental activities was higher in the RCA territory with SD-SPECT (5.13% ± 4.04%) than with DLAC-SPECT (4.21% ± 3.24%) and Rb-PET (3.92% ± 3.08%, both P < .001).

When utilizing DLAC maps for reconstruction of [^99mTc]sestamibi CZT-SPECT, there is no more drop in the RCA territory uptake with lower fluctuations in stress-to-rest activity, aligning closely with the results observed using Rb-PET in the same patients.

NCT01679886 on clinicaltrials.gov.