Use of Conventional Late Imaging Protocol for Dynamic SPECT Imaging

Introduction: Using dynamic SPECT reconstruction algorithms, one can represent tracer dynamics as linear expansion over few temporal curves or factors, and corresponding coefficients or factor-maps that segment the reconstructed volume. We have shown this previously by developing the Spline Initialized Factor Analysis of Dynamic Structures (SIFADS), which performed the factorization directly from a dynamic sinogram acquired with rotating camera gantry. Earlier studies focused on early post-injection tracer dynamics. In this paper, we show that it is possible to obtain comparable results from late time frames as well. Late scans are typical in conventional static imaging and are acquired after some time of tracer administration to let the tracer stabilize in targeted tissues.

Methods: We used dynamic sinograms from a dual-headed SPECT camera acquired at UCSF with ^99mTc -tetrofosmin. The acquisition ran in a continuous rotation and continuous acquisition mode with each rotation taking a minute over total of 20 minutes for 20 rotations. Two sets of sinograms were used, those over (1) the 3^rd and 4^th rotations to represent early imaging data, and (2) the 9^th and 10^th rotations to represent late time data. The SIFADS ran for ten iterations to generate three factor-curves and the corresponding factor-maps, which corresponded myocardium, blood ventricles, and the background. We compared the signal-to-noise ratio (SNR) of the myocardium factor-maps from these two sets of data to measure the quality of the reconstruction. Also, static maximum likelihood expectation maximization reconstructions were performed for each of these two data sets, by aggregating the same sinogram from the third and fourth rotations, and that from the ninth and tenth rotations. The SNRs for manually segmented myocardium from the two reconstructed static images were also computed and compared.

Results: The factor-maps produced by both the early and late rotations clearly matched our expectation. The signal-to-noise ratio of the myocardium factor-map from the dynamic reconstructions with attenuation correction was 6.23 for the early rotations and that for the late rotations was 5.12. The SNR of the corresponding static reconstructions with attenuation correction were 4.07 for the early rotations and 2.48 for the late rotations.

Conclusions: This pilot study demonstrates that (1) the factor-maps produced by dynamic image reconstruction had better qualities than those from the static image reconstruction; and (2) that dynamic SPECT imaging of late rotations, a similar protocol for conventional static cardiac SPECT imaging, contains sufficient information about tracer dynamics relative to that from early rotations. Ability to use only late scan datasets will open the avenue for dynamic image analyses even for clinically used SPECT imaging protocols, instead of a modified dynamic imaging protocols that require long scan times starting at the time of tracer administration. To the extent we know, these two hypotheses have never been tested before. Our future work will attempt to prove these hypotheses with a larger number of datasets.

• Download figure
• Open in new tab
• Download powerpoint