Abstract:Purpose: Single-photon emission computed tomography (SPECT) is widely used to detect myocardial ischemia and myocardial infarction. It is important to assess and compare different SPECT system designs in order to achieve the highest detectability of cardiac defects. Methods: Whitaker et al.'s study ["Estimating random signal parameters from noisy images with nuisance parameters: linear and scanning-linear methods," Opt. Express 16(11), 8150-8173 (2008)] on the scanning linear observer (SLO) shows that the SLO can be used to estimate the location and size of signals. One major advantage of the SLO is that it can be used with projection data rather than with reconstruction data. Thus, this observer model assesses the overall hardware performance independent of any reconstruction algorithm. In addition, the computation time of image quality studies is significantly reduced. In this study, three systems based on the design of the GE cadmium zinc telluride-based dedicated cardiac SPECT camera Discovery 530c were assessed. This design, which is officially named the Alcyone Technology: Discovery NM 530c, was commercialized in August, 2009. The three systems, GE27, GE19, and GE13, contain 27, 19, and 13 detectors, respectively. Clinically, a human heart can be virtually segmented into three coronary artery territories: the left-anterior descending artery, left-circumflex artery, and right coronary artery. One of the most important functions of a cardiac SPECT system is to produce images from which a radiologist can accurately predict in which territory the defect exists [http://www.asnc.org/media/PDFs/PPReporting081511.pdf, Guideline from American Society of Nuclear Cardiology]. A good estimation of the extent of the defect from the projection images is also very helpful for determining the seriousness of the myocardial ischemia. In this study, both the location and extent of defects were estimated by the SLO, and the system performance was assessed by localization receiver operating characteristic (LROC) [P. Khurd and G. Gindi, "Decision strategies maximizing the area under the LROC curve," Proc. SPIE 5749, 150-161 (2005)] or estimation receiver operating characteristic (EROC) [E. Clarkson, "Estimation receiver operating characteristic curve and ideal observers for combined detection/estimation tasks," J. Opt. Soc. Am. A 24, B91-B98 (2007)] curves. Results: The area under the LROC/EROC curve (AULC/AUEC) and the true positive fraction (TPF) at a specific false positive fraction (FPF) can be treated as the figures of merit. For radii estimation with a 1 mm tolerance, the AUEC values of the GE27, GE19, and GE13 systems are 0.8545, 0.8488, and 0.8329, and the TPF at FPF = 5% are 77.1%, 76.46%, and 73.55%, respectively. The assessment of all three systems revealed that the GE19 system yields estimated information and cardiac defect detectability very close to those of the GE27 system while using eight fewer detectors. Thus, 30% of the expensive detector units can be removed with confidence. Conclusions: As the results show, a combination of the SLO and LROC/EROC curves can determine the configuration that yields the most relevant estimation/detection information. Thus, this is a useful method for assessing cardiac SPECT systems.

Human observer efficiency for signal detection and localization in emission tomographic images

The efficiency of the human observer for lesion detection and localization in emission tomography.

Comparison of Human and Hotelling Observer Performance for a Fan-Beam CT Signal Detection Task.

Modeling human observer detection for varying data acquisition in undersampled MRI for two-alternative forced choice (2-AFC) and forced localization tasks

Accurate Computation of the Hotelling Observer for the Evaluation of Image Reconstruction Algorithms in Helical, Cone-Beam CT

Analysis of Fourier-domain task-based detectability index in tomosynthesis and cone-beam CT in relation to human observer performance

Evaluation of the Channelized Hotelling Observer for Signal Detection in 2D Tomographic Imaging

Region of interest based Hotelling observer for computed tomography with comparison to alternative methods.

Supervised Learning Of Model Observers For Assessment Of Ct Image Reconstruction Algorithms

Modeling human observer detection in undersampled magnetic resonance imaging (MRI) reconstruction with total variation and wavelet sparsity regularization

Evaluation of deep learning-based CT reconstruction with a signal-Laplacian model observer

Collimator optimization in SPECT based on a joint detection and localization task

Assessment of cardiac single-photon emission computed tomography performance using a scanning linear observer

Ideal-observer computation with anthropomorphic phantoms using Markov Chain Monte Carlo

Rapid calculation of detectability in Bayesian single photon emission computed tomography.

Foveated Model Observers for Visual Search in 3D Medical Images

Task-based assessment for neural networks: evaluating undersampled MRI reconstructions based on human observer signal detection

Channelized Hotelling and Human Observer Study of Optimal Smoothing in Spect Map Reconstruction

Unsupervised Evaluation and Optimization for Electrical Impedance Tomography

Strategies to Jointly Optimize Spect Collimator and Reconstruction Parameters for a Detection Task

On anthropomorphic decision making in a model observer