Background: By gating image acquisition in myocardial perfusion SPECT (MPS) to ECG, left ventricular (LV) volumes and function can be determined. Several previous studies have shown that existing MPS software packages underestimate LV volumes compared to cardiac magnetic resonance (CMR). The aim of this study was therefore to develop a new LV segmentation algorithm for gated MPS using CMR as reference standard.
Methods and results: A total of 126 patients with suspected coronary artery disease, who underwent both gated MPS and CMR were retrospectively included. The proposed LV segmentation algorithm (Segment) was trained in 26 patients, and tested in 100 patients in comparison to four commercially available MPS software packages (QGS, MyoMetrix, ECTb, and Exini) using CMR as reference standard. Mean bias ± SD between MPS and CMR was for EDV -5% ± 12%, -43% ± 8%, -40% ± 8%, -42% ± 9%, -32% ± 7%, for ESV 0% ± 17%, -41% ± 16%, -34% ± 15%, -54% ± 13%, -41% ± 10%, for EF -2% ± 13%, -1% ± 14%, -7% ± 15%, 17% ± 16%, 10% ± 17% for Segment, QGS, MyoMetrix, ECTb, and Exini, respectively, and for LVM 3% ± 18%, 33% ± 25%, 37% ± 24% for Segment, QGS, and ECTb, respectively. Correlation between MPS by Segment and CMR were for EDV R (2) = 0.89, for ESV R (2) = 0.92, for EF R (2) = 0.69, and for LVM R (2) = 0.72, with no difference compared to the correlation between the other MPS software packages and CMR (EDV R (2) = 0.86-0.92, ESV R (2) = 0.91-0.93, EF R (2) = 0.64-0.65, and LVM R (2) = 0.68-0.70).
Conclusion: The Segment software quantifies LV volumes and EF by MPS with similar correlation and a low bias compared to other MPS software packages, using CMR as reference standard. Hence, the Segment software shows potential to provide clinically relevant volumes and functional values from MPS.