Optimization of classical pulmonary SPECT on a revolutionary camera

Aim: Shorter acquisition times results in greater throughput and patient comfort. We investigated if a new 12-detector cadmium-zinc-telluride (CZT) SPECT could enable shorter acquisition times for pulmonary ventilation/perfusion (V/Q) SPECT while maintaining image quality. Further, we investigated the potential of utilizing a Block Sequential Regularized Expectation Maximization (BSREM) reconstruction algorithm, which recently became commercially available in SPECT.

Method: Eight patients were scanned both on the StarGuide 12-detector CZT SPECT (GE Healthcare) and a Discovery 670 2-detector NaI SPECT (GE Healthcare). All had Q-SPECT with an acquisition time of 17 minutes after injection of 200 MBq +/-10% Tc-99m macro-aggregated albumin. Four patients further underwent 12 minutes of V-SPECT after inhaling approximately 25 MBq of Tc-99m-technegas. Images from the 12-detector SPECT were reconstructed with list-mode files at reduced times of 33%, 42%, 50%, and 67% using Ordered Subset Expectation Maximization (OSEM) with 5 iterations (i), 8 subsets (s) and 3mm Gaussian post-filter. The BSREM algorithm Q.Clear (GE Healthcare) was investigated at 50% acquisition time with 20i10s using the relative differences for maximum regularization method with gamma of 1 and varying regularization strength factors (β) of 0.02, 0.05 and 0.08. Three nuclear medicine physicians evaluated image quality through absolute Visual Grading Analysis. The criteria diagnostic confidence (DQ), visibility of interlobar fissures (VIF), and visibility of lung boundaries (VLB) were assessed using a four-point forced-choice scale, except for VIF, which was done binarily. Visual grading characteristics curves (VGC) were generated and area under the curve (AUCVGC) were calculated. Fleiss' Kappa and Cohen's Weighted Kappa were used to assess interrater/intrarater reliability.

Results: AUCVGC curves on OSEM reconstructions were below 0.5 for all acquisition times for most parameters on the 12-detector SPECT compared to 2-detector SPECT, except for VIF, which had an AUCVGC of 0.507 (CI: 0.375-0.670) when using 100% acquisition time on OSEM. However, when BSREM was used on 50% acquisition time and β-value of 0.08, AUCVGC increased from 0.037, 0.070, 0.436 and 0.170 to 0.440, 0.473, 0.559 and 0.480 for IQ, DC, VIF and VLB, respectively.

Conclusion: Our study indicates the possibility of reducing acquisition time by 50% for V/Q SPECT without compromising image quality on the 12-detector SPECT when using BSREM 20i10s with β-value of 0.08. Image quality was lower across all OSEM reconstructions for the 12-detector SPECT compared to the 2-detector SPECT, possibly due to the use of smaller voxel volumes (2.46 mm), which impact needs to be investigated further.
References: none