A Mathematical Observer Study for the Evaluation and Optimization of Compensation Methods for Myocardial SPECT Using a Phantom Population That Realistically Models Patient Variability

Abstract

The goal of this study was to develop and apply a population of phantoms that realistically models patient variability and use it to optimize and evaluate different compensation methods used during reconstruction process with respect to defect detection in myocardial SPECT images. Various combinations of attenuation, detector response and scatter compensation were used in this study. A major difference between this and previous studies was that the level of realism was significantly increased by inclusion of variability in heart and organ uptakes, in the heart size and orientation, and in the defect size and contrast. In this study we used a population of 24 4-D NCAT phantoms [1] (half male, half female) recently developed with statistical models for organ uptake and organ size based on clinical data. Almost noise-free projection data of the torso, heart, liver, lungs, and other organs were simulated for each phantom using the SIMSET MC simulation code. They were then combined to form 72 sets of projections for each phantom using randomly sampled activity ratios from a clinically realistic distribution. Poisson noise was then added to the projection data. We applied the channelized hotelling observer (CHO) and receiver operating characteristic (ROC) analysis to optimize iteration number for OSEM and cutoff frequency of a 3-D post-reconstruction Butterworth filter. We found that the area under the ROC curve (AUC) values were reduced compared to a previous study that included significantly less phantom variability, even though the defect contrast was higher and noise level was lower. The resulting AUC values were similar to those obtained using patient data. We found, in agreement with the previous study, that including compensation for more effects resulted in improved defect detectability. However, the optimal filter cutoff frequency was increased compared to the previous study. These studies demonstrate the importance of including realistic levels of phantom variability in myocardial perfusion studies using simulated data.

An anthropomorphic phantom for quantitative evaluation of breast MRI

“…Anthropomorphic phantoms have been implemented in other imaging modalities and have been used to demonstrate that simple, nonanthropomorphic phantoms produce misleading results when used to optimize imaging systems. 24,[57][58][59] The phantom described in this work will provide a much-needed platform for better understanding the interaction of breast MRI acquisition parameters with lesion detection and estimation.…”

“…Breast anatomy is composed of interlacing fat-and water-containing tissues that produce imaging artifacts ͑e.g., chemical shift artifact͒ and can hinder lesion detectability. The importance of realistic phantom structure in the evaluation of imaging systems for the purpose of abnormality detection was demonstrated in a study by He et al, 24 in which both phantom and patient data were used to perform reader studies evaluating compensation methods for myocardial SPECT image reconstruction algorithms. They found that phantoms with more complicated, realistic image structures resulted in performance measures for defect detection that more closely matched results derived from patient data than studies with more simplistic phantoms.…”

Optimization of energy window for 90Y bremsstrahlung SPECT imaging for detection tasks using the ideal observer with model‐mismatch

“…The channelized hotelling observer (CHO) has been previously validated with human observer studies and showed good agreement with the human observers in a variety of applications. [17][18][19][20] However, using the CHO for optimizing acquisition parameters requires reconstruction of a large number of projection data and optimization of various reconstruction and regularization parameters (e.g., iteration number and cutoff frequency of smoothing filters). Thus, using an observer that performs directly on projection data is computationally less expensive than an observer that performs on reconstructed images (e.g., the CHO).…”

Myocardial perfusion SPECT using a rotating multi-segment slant-hole collimator

“…13,14 These channel sets have been used in several cardiac studies and shown good correlation with human observer performance. [13][14][15][16] The spatial domain templates were obtained by taking the inverse Fourier transform of the four channels in frequency domain and shifting the template centers to the defect centroid. For each reconstructed and resliced image, a four-element feature vector was obtained by taking the dot product of the extracted image and the spatial domain templates for each channel.…”