Observer studies of image quality of denoising reduced-count cardiac single photon emission computed tomography myocardial perfusion imaging by three-dimensional Gaussian post-reconstruction filtering and deep learning

BACKGROUND

The aim of this research was to asses perfusion-defect detection-accuracy by human observers as a function of reduced-counts for 3D Gaussian post-reconstruction filtering vs deep learning (DL) denoising to determine if there was improved performance with DL.

METHODS

SPECT projection data of 156 normally interpreted patients were used for these studies. Half were altered to include hybrid perfusion defects with defect presence and location known. Ordered-subset expectation-maximization (OSEM) reconstruction was employed with the optional correction of attenuation (AC) and scatter (SC) in addition to distance-dependent resolution (RC). Count levels varied from full-counts (100%) to 6.25% of full-counts. The denoising strategies were previously optimized for defect detection using total perfusion deficit (TPD). Four medical physicist (PhD) and six physician (MD) observers rated the slices using a graphical user interface. Observer ratings were analyzed using the LABMRMC multi-reader, multi-case receiver-operating-characteristic (ROC) software to calculate and compare statistically the area-under-the-ROC-curves (AUCs).

RESULTS

For the same count-level no statistically significant increase in AUCs for DL over Gaussian denoising was determined when counts were reduced to either the 25% or 12.5% of full-counts. The average AUC for full-count OSEM with solely RC and Gaussian filtering was lower than for the strategies with AC and SC, except for a reduction to 6.25% of full-counts, thus verifying the utility of employing AC and SC with RC.

CONCLUSION

We did not find any indication that at the dose levels investigated and with the DL network employed, that DL denoising was superior in AUC to optimized 3D post-reconstruction Gaussian filtering.

Adherence of Tc-99 Sestamibi to Plastic Syringes Could Complicate Efforts in Dose Reduction in MPI SPECT

BACKGROUND

Multiple disposable syringe designs are available both with and without silicone lubricant. Tc99m sestamibi variably adheres to certain syringe designs. In practice, this increases pre-injection activity to ensure constant acceptable image quality in the setting of high adherence.

METHODS

The residual activity following 35 Tc99m MDP injections was used as a proxy for the dead-space residual activity. This was compared to Tc99m sestamibi injections in the same syringe in 107 injections, a syringe without lubricant in 35 injections, and a syringe without lubricant or rubber gasket in 16 injections. The syringe with lubricated barrel was disassembled and each part separately assayed to see which part demonstrated the most adherence.

RESULTS

The standard design demonstrated a mean adherence of 26% (SD 7.0%). This was significantly more than the 6% (SD 3.7%) demonstrated with MDP. The two designs without silicone lubricant demonstrated mean adherence of 7.2% (SD 2.3%) and 6.4% (SD 2.3%). Disassembly of the three part syringe showed the greatest adherence to the rubber gasket.

CONCLUSIONS

QC procedures determining adherence in the administration syringe are warranted for any change in syringe design or manufacturer. If adherence is present, conversion to non-lubricated plastic syringes with lower and more predictable retention of Tc99m sestamibi enables more precise titration of prescribed activities and reduction of overall patient absorbed doses.

Comparison of methods of acquiring attenuation maps for cardiac SPECT in the presence of respiratory motion

BACKGROUND

We investigated the effect of the respiratory motion on attenuation-corrected (AC) SPECT images for three different SPECT systems, each using a different approach in obtaining attenuation maps: scanning-line sources (SLS) acquired simultaneous with emission; slow cone-beam CT (CBCT) acquired sequentially to emission; and fast helical CT (HCT) acquired sequentially to emission.

METHODS

A torso phantom filled with (99m)Tc was used to model a cardiac perfusion study. Stationary baseline acquisitions were followed by the acquisitions with the phantom moving axially using a computer-controlled motion platform to simulate breathing. In addition, HCT acquisitions were made simulating breath-hold at different extents of misalignment between CT and emission. HCT images were also used to simulate the Average-CT method. Acquisitions were repeated with added breast attachments, and the heart insert in two different orientations. Visual comparison was made of AC maps, AC emission slices and polar maps. Quantitative comparisons were made of global uniformity based on the percent fractional standard deviation (%FSD) of the polar map segment values, and the ratio of the segment values in the Anterior and Inferior walls divided by that of the Lateral and Septal walls (AI/LS ratio).

RESULTS

The AC maps for the SLS were inferior to the CT's, and most impacted by added large breast attachment. Motion artifacts seen on CBCT slices were minimized in the derived attenuation maps. AC maps obtained from HCT showed inconsistent organ sizes depending on the direction of respiration at the time of acquisition. Both visually and quantitatively CBCT resulted in the best uniformity (up to 3.4 % lower in %FSD) for all the stationary acquisitions, and for the motion acquisition of the female phantom with large breast attachment (up to 4.0 % lower). For the motion acquisition of the male phantoms, HCT resulted in slightly better uniformity (<0.5 % lower) than CBCT. Breath-hold at end-expiration slightly improved (up to 1.1 %) the uniformity over the HCT acquired during regular breathing. Further improvement was achieved with the Average-CT method. For all the systems, phantom respiratory motion reduced the AI/LS ratio compared to when the phantoms were stationary.

CONCLUSIONS

The CBCT approach resulted in the best uniformity of the AC emission images. For the female phantom with larger breast attachment, HCT and SLS were truncated at some projection angles introducing artifacts into the AC emission images. The emission image artifacts observed with HCT could be mitigated by performing breath-hold acquisition at end-expiration or Average-CT type acquisitions.

Feasibility of stress only rubidium-82 PET myocardial perfusion imaging

BACKGROUND

Stress only SPECT myocardial perfusion imaging (MPI) is a validated strategy to streamline cardiac diagnostic imaging. The potential use of Rb82 PET stress only MPI has not been investigated.

METHODS AND RESULTS

Stress images from 200 Rb82 PET-MPI were reviewed by two blinded readers and categorized as not requiring additional rest images (normal) or requiring additional images (abnormal or equivocal). No additional images were deemed necessary for 95 (48%) and 99 (50%) by the two blinded readers. The stress only interpretation was compared to the previous read of the complete rest-stress study. The rate of detecting a normal result with stress only reading was 76%-79% with a negative predictive value of 94%-95%. Clinical predictors of a normal stress only PET-MPI included lower age, the absence of CAD, and female gender, but not body mass index. Blinded reads of 50 additional consecutive PET-MPI from patients with selected clinical predictors (age <65 years, no known CAD) were then performed. Of these, 40 (80%) were normal by previous rest-stress reading, and 34 (68%) were categorized as not requiring additional images after stress only reading. PET stress only imaging would have resulted in a mean reduction of radiation exposure of 2.4 mSv per study according to a published radiation estimate.

CONCLUSION

Stress only Rb82 PET-MPI is a feasible strategy to reduce resource utilization and radiation exposure associated with MPI. This strategy would be most applicable to patients with a lower pretest likelihood.