A template-based approach to semi-quantitative SPECT myocardial perfusion imaging: Independent of normal databases

Two methods to generate templates for template-based partial volume effect correction: SPECT phantom experiments

In this paper, we explore the applicability of template-based compensation for the partial volume effect (PVE) for situations where (i) the image has multiple uptake sites (tumors and organs) but only one of them is treated as a region of interest (ROI) with the boundaries available from a high-resolution modality and (ii) no information regarding activities inside or outside this ROI is a priori available. We modeled this situation by performing SPECT acquisitions of phantoms containing 21 containers, which had different shapes and sizes and were filled with different levels of activity. In our analysis, each of these containers was treated as an individual ROI. We compared the performance of two methods of template construction. In method 1, the ROI template value was obtained from a conventionally reconstructed (without PVEC) image. In method 2, the ROI template value was directly (bypassing the PVE-affected conventional image) calculated from projections using region-based reconstruction. Our processing shows that method 1 resulted in consistent (activities for all 21 ROIs were improved) but relatively weak PVE compensation (errors of recovered total activities were equal to or lower than 10% for 5 ROIs only). Application of method 2 resulted in a selective (activities for 19 ROIs were improved) but considerably better compensation when compared to method 1 (errors of recovered total activities were equal to or lower than 10% for 10 ROIs).

Toward a practical template-based approach to semiquantitative SPECT myocardial perfusion imaging

PURPOSE

Our template-based quantitative perfusion single photon emission computed tomography (SPECT) method (T-QPS) performs semiquantitative analysis for myocardial perfusion imaging (MPI) without the use of normal databases. However, in its current form, T-QPS requires extensive calculations, which limits its clinical application. In the interest of clinical feasibility, the authors examine the trade-off between accuracy and processing time as the method is simplified.

METHODS

The T-QPS method uses the reconstructed SPECT image of the patient to create a 3D digital template of his∕her healthy heart. This template is then projected, reconstructed, and sampled into the bulls-eye map domain. A ratio of the patient and template images produces a final corrected image in which a threshold is applied to identify perfusion defects. In principle, the template should be constructed with the heart and all extracardiac activity, and the projection step should include primary and scatter components; however, this leads to lengthy calculations. In an attempt to shorten the processing time, the authors analyzed the performance of four template (T) generation methods: T(P-HRT), T(PS-HRT), T(P-HRTBKG), and T(PS-HRTBKG), where P and S represent primary and scattered photons included in the projection step, respectively; and HRT and HRTBKG represent template constructed with the heart only and the heart with background activity, respectively. Forty-eight thorax phantoms and 21 randomly selected patient studies were analyzed using each approach. All studies used GE's Infinia Hawkeye SPECT∕CT system and followed a standard cardiac acquisition protocol.

RESULTS

Approximate processing times for the T(P-HRT), T(PS-HRT), T(P-HRTBKG), and T(PS-HRTBKG) methods were less than a minute, 2-3 h, less than a minute and 3-4 h, respectively. In both the simulation and patient studies, a significant reduction in the quality of perfusion defect definition was exhibited by the T(P-HRT) method relative to the other three methods. The optimal method with respect to perfusion defect definition and processing time was T(P-HRTBKG) with a sensitivity, specificity, and accuracy in spatially defining the perfusion defects (simulation study) of 80%, 84%, and 83%, respectively.

CONCLUSIONS

The T-QPS method using T(P-HRTBKG) leads to accurate and fast semiquantitative analysis of SPECT MPI, without the use of normal databases.