Phantom Evaluation of C-SPECT Performance for Myocardial Perfusion Imaging

1384 Objectives: Myocardial Perfusion Imaging (MPI) remains a cornerstone of cardiovascular care with ~8 million scans per year in the US. Relative to current SPECT technology, cardiac PET has improved sensitivity and specificity and allows for dynamic imaging and thus myocardial blood flow (MBF) quantification. Given growing awareness of the role of the coronary mirovasculature, the latter is particularly useful. Recently, several SPECT scanners have been developed that can acquire all lines of response simultaneously, trading a large Field of View (FOV) for multiple simultaneous projections of the cardiac region, potentially enabling fast dynamic scanning and MBF measurements in SPECT. C-SPECT is such a system with the additional capability of adaptive imaging, which allows for different resolution-sensitivity-FOV operating points. This can be used, for example, to switch between high-sensitivity mode for dynamic imaging and high-resolution mode for static imaging. In this study, we evaluate the capabilities of this novel system for identifying cardiac lesions in the high-resolution mode. Methods: We acquired high-count scans of emission data using an anthropomorphic torso phantom filled with aqueous 99mTc-pertechnetate with concentration ratios 1.0:1.3:4.4 for the lung:background:liver, respectively, without the left-ventricular (LV) insert to mimic the distribution in males. For females, two 500-ml saline bags filled with matching background concentration, were positioned to mimic breast tissue. After decay, the cardiac insert was filled and placed in the torso phantom. The torso phantom, while cold, was filled with water for scatter and attenuation. High-count scans of the cardiac insert were acquired with and without the saline bags so that the data could be merged with the background scans. The cardiac insert was acquired in 25 configurations: no lesion and six lesion sizes in each of the four walls. Lesions were 3D-printed to be 24 mm in axial extent, 45o or 90o in circumferential extent, and to involve 50%, 75% or 100% of the LV wall thickness. The data were merged in post-processing to create ensemble data sets equivalent to 2-, 5-, and 10-minute scans for each combination, and were reconstructed using 25 iterations of a maximum-likelihood estimation-maximization (ML-EM) algorithm with scatter correction, but without attenuation correction. We used a numerical algorithm similar to that described in the literature to automatically estimate Receiver-Operator-Characteristic (ROC) curves. Results: The area under the ROC curve (AROC) was nearly 1 for all lesions imaged with 5-minute and 10-minute scans. The AROC for 2-minute scans varied by lesion location and decreased for females. Discussion: C-SPECT is very effective in identifying lesions, including focal subendocardial lesions for both male and female anatomy. The trends seen in the ROC curves for the 2-minute scans are consistent with expectations that lateral lesions were the easiest to identify and that additional attenuation and scatter in the female phantoms decreased image quality. Conclusions: C-SPECT is capable of reliably identifying cardiac lesions of varying severity in male and female phantoms using realistic background distributions for MPI down to scans of two minutes.