Quantifying the Effects of Acquisition Parameters in Cardiac SPECT Imaging and Comparison with Visual Observers

Abstract

Abstract-Significant debate remains whether the established imaging protocols for myocardial perfusion SPECT delivers optimal image quality. This debate has been intensified with the introduction of advanced iterative reconstruction techniques which differ considerably from the FBP reconstruction technology commonly available when the current protocols were established. In order to characterize and classify image quality affecting factors, simulations combined with time consuming observer studies are often used.In this work we seek to simplify these image assessment tasks by performing numerical characterizations based on real data, and try to unite quantitative performance with visual impression derived from human observer studies.For this purpose, we assess the response of cardiac image quality to variations in acquisition and reconstruction protocol. We analyze anthropomorphic phantom data derived from large dimensioned acquisitions via subsampling in terms of counts and spatial and angular resolution. Data is reconstructed using FBP and OSEM-3D ("Flash3D") with parameter variations and the reconstructed volumes are quantitatively analyzed for cardiac relevant features like myocardial uniformity and resolution. Receiver Operating Characteristic (ROC) studies are performed for selected imaging parameters. Results obtained characterize the sensitivity of the cardiac image quality to some changes in the imaging protocol like angular step, orbit radius and collimation and show correlations of key quantitative measures and related ROC behavior.

Estimation accuracy of ejection fraction in gated cardiac SPECT/CT imaging using iterative reconstruction with 3D resolution recovery in rapid acquisition Protocols

“…Gated timeslots, matrix size, and ejection fraction were varied according to Table II. From this measured data ("reference"), we created "whatif" datasets. For this we developed a DICOM compatible tool that extracts angular views from a projection data set, rebins the matrix and pixel size and reduces counts in a pixel by a selected fraction using binomial subsampling while maintaining the Poisson character of the data [5]. In this way, we create gated projection data sets with different levels of counts and angular sampling for the assessment of clinical and rapid protocols.…”

Development of Rapid SPECT Acquisition Protocol for Myocardial Perfusion Imaging

“…From this measured data ("reference"), we create "what-if" datasets. For this we developed a DICOM compatible tool which extracts angular views from a projection data set, rebins the matrix and pixel size and reduces counts in a pixel to some fraction f using binomial subsampling, and maintaining the Poisson character of the data [15]. On purpose we don't allow for projection view interpolation as such data would not correspond to data acquired with our systems.…”

Evaluation of image reconstruction methods for 123I-MIBG-SPECT: a rank-order study

“…Various studies have shown improvements in image quality and lesion detectability by optimizing the number of EI and choosing an appropriate reconstruction method by doing phantom measurements, as well as human observer studies (11)(12)(13)(14). However, the optimal number of EI and reconstruction method is dependent on the specific situation and needs to be optimized for the specific clinical task.…”