Application of task-based measures of image quality to optimization and evaluation of three-dimensional reconstruction-based compensation methods in myocardial perfusion SPECT

Frey, Eric; Gilland, K.L.; Tsui, B.M.W.

doi:10.1109/tmi.2002.804437

Cited by 116 publications

(117 citation statements)

References 32 publications

(30 reference statements)

Supporting

Mentioning

113

Contrasting

Unclassified

Order By: Relevance

“…Because there are so many parameter values to test in determining the optimum parameter settings, numerical observers, such as the Channelized Hotelling observer (CHO) [2]- [5], provide a more practical method for parameter optimization than human observer studies, provided the numerical observer can predict human observer performance. Previous studies have compared the performance of the CHO to human observers for tumor detection in Ga SPECT imaging [6]- [8] and cardiac defect detection in Tc-sestamibi SPECT imaging [9]- [13]. The Ga SPECT studies showed good agreement between the CHO and human observer performance; however, for the Tc-sestmibi cardiac SPECT studies, the results were mixed.…”

Section: Comparison Of Channelized Hotelling and Humanmentioning

confidence: 99%

“…Narayanan [11] showed the good correlation between CHO and human observer performance for parameter optimization, but only if the images were grayscale instead of real-valued or if an observer internal noise model [14] was included in the CHO. Sankaran and Frey [12], [13] showed good correlation between the CHO and human observer performance for optimization of the filter cutoff frequency in post-reconstruction filtering, but did not compare the CHO and human performance in optimizing the number of iterations in their iterative reconstruction. The Sankaran and Frey studies did use grayscale images, but did not use internal noise in the CHO and used a very low lesion contrast (12.5%).…”

Section: Comparison Of Channelized Hotelling and Humanmentioning

confidence: 99%

“…In this study, we increase the lesion contrast to a more clinically relevant level of 20%, but use the internal observer noise model to adjust the AUC values to a level that allows meaningful statistical analysis and comparison with human observer results. In addition, in this study we will use a more anatomically realistic phantom and simulate SPECT projection data with a more realistic Monte Carlo simulation as opposed to the analytical projection data simulation used by Frey [13]. Specifically, in this study, the CHO with the internal noise model will be compared to human observer performance for the optimization of the (number of subsets)/iteration and the number of iterations in the iterative ordered subset-expectation maximization (OS-EM) image reconstruction algorithm [15] used in obtaining the SPECT images.…”

Section: Comparison Of Channelized Hotelling and Humanmentioning

confidence: 99%

“…The SA images were then low-pass filtered with a Butterworth filter of order 5 and cutoff of 0.15 pixels . These filter parameters were chosen based on previous work which found optimal filter parameters for similar simulated SPECT images [12], [13].…”

Section: Image Reconstruction and Processingmentioning

confidence: 99%

“…The basic CHO paradigm used in this study is very similar to that used in the previous study by Frey [13]. The reader is referred to that study for a detailed description of the paradigm.…”

Section: E Channelized Hotelling Observer Studymentioning

confidence: 99%

See 4 more Smart Citations

Comparison of channelized hotelling and human observers in determining optimum OSEM reconstruction parameters for myocardial SPECT

Gilland¹,

Qi²,

Tsui³

et al.

2003 IEEE Nuclear Science Symposium. Conference Record (IEEE Cat. No.03CH37515)

Self Cite

View full text Add to dashboard Cite

The performance of the Channelized Hotelling Observer (CHO) was compared to that of human observers for determining optimum parameters for the iterative OS-EM image reconstruction method for the task of defect detection in myocardial SPECT images. The optimum parameters were those that maximized defect detectability in the SPECT images. Low noise, parallel SPECT projection data, with and without an anterior, inferior or lateral LV wall defect, were simulated using the Monte Carlo method. Poisson noise was added to generate noisy realizations. Data were reconstructed using OS-EM at 1 & 4 subsets/iteration and at 1, 3, 5, 7 & 9 iterations. Images were converted to 2D short-axis slices with integer pixel values. The CHO used 3 radially-symmetric, 2D channels, with varying levels of internal observer noise. For each parameter setting, 600 defect-present and 600 defect-absent image vectors were used to calculate the detectability index (dA). The human observers rated the likelihood that a defect was present in a specified location. For each parameter setting, the AUC was estimated from 48 defect-present and 48 defect-absent images. The combined human observer results showed the optimum parameter setting could be in the range 5-36 updates ([number of subsets]/iteration × number of iterations). The CHO results showed the optimum parameter setting to be 4-5 updates. The performance of the CHO was much more sensitive to the reconstruction parameter setting than was that of the human observers. The rankings of the CHO detectability values did not change with varying levels of internal noise. The CHO results showed the optimum parameter setting to be 4-5 updates. The performance of the CHO was much more sensitive to the reconstruction parameter setting than was that of the human observers. The rankings of the CHO detectability values did not change with varying levels of internal noise.

show abstract

Section: Comparison Of Channelized Hotelling and Humanmentioning

confidence: 99%

Section: Comparison Of Channelized Hotelling and Humanmentioning

confidence: 99%

Section: Comparison Of Channelized Hotelling and Humanmentioning

confidence: 99%

Section: Image Reconstruction and Processingmentioning

confidence: 99%

“…The basic CHO paradigm used in this study is very similar to that used in the previous study by Frey [13]. The reader is referred to that study for a detailed description of the paradigm.…”

Section: E Channelized Hotelling Observer Studymentioning

confidence: 99%

See 3 more Smart Citations

Comparison of channelized hotelling and human observers in determining optimum OSEM reconstruction parameters for myocardial SPECT

Gilland¹,

Qi²,

Tsui³

et al.

2003 IEEE Nuclear Science Symposium. Conference Record (IEEE Cat. No.03CH37515)

Self Cite

View full text Add to dashboard Cite

show abstract

Need for objective task‐based evaluation of deep learning‐based denoising methods: A study in the context of myocardial perfusion SPECT

Rahman

Laforest

et al. 2023

Medical Physics

View full text Add to dashboard Cite

Background Artificial intelligence‐based methods have generated substantial interest in nuclear medicine. An area of significant interest has been the use of deep‐learning (DL)‐based approaches for denoising images acquired with lower doses, shorter acquisition times, or both. Objective evaluation of these approaches is essential for clinical application. Purpose DL‐based approaches for denoising nuclear‐medicine images have typically been evaluated using fidelity‐based figures of merit (FoMs) such as root mean squared error (RMSE) and structural similarity index measure (SSIM). However, these images are acquired for clinical tasks and thus should be evaluated based on their performance in these tasks. Our objectives were to: (1) investigate whether evaluation with these FoMs is consistent with objective clinical‐task‐based evaluation; (2) provide a theoretical analysis for determining the impact of denoising on signal‐detection tasks; and (3) demonstrate the utility of virtual imaging trials (VITs) to evaluate DL‐based methods. Methods A VIT to evaluate a DL‐based method for denoising myocardial perfusion SPECT (MPS) images was conducted. To conduct this evaluation study, we followed the recently published best practices for the evaluation of AI algorithms for nuclear medicine (the RELAINCE guidelines). An anthropomorphic patient population modeling clinically relevant variability was simulated. Projection data for this patient population at normal and low‐dose count levels (20%, 15%, 10%, 5%) were generated using well‐validated Monte Carlo‐based simulations. The images were reconstructed using a 3‐D ordered‐subsets expectation maximization‐based approach. Next, the low‐dose images were denoised using a commonly used convolutional neural network‐based approach. The impact of DL‐based denoising was evaluated using both fidelity‐based FoMs and area under the receiver operating characteristic curve (AUC), which quantified performance on the clinical task of detecting perfusion defects in MPS images as obtained using a model observer with anthropomorphic channels. We then provide a mathematical treatment to probe the impact of post‐processing operations on signal‐detection tasks and use this treatment to analyze the findings of this study. Results Based on fidelity‐based FoMs, denoising using the considered DL‐based method led to significantly superior performance. However, based on ROC analysis, denoising did not improve, and in fact, often degraded detection‐task performance. This discordance between fidelity‐based FoMs and task‐based evaluation was observed at all the low‐dose levels and for different cardiac‐defect types. Our theoretical analysis revealed that the major reason for this degraded performance was that the denoising method reduced the difference in the means of the reconstructed images and of the channel operator‐extracted feature vectors between the defect‐absent and defect‐present cases. Conclusions The results show the discrepancy between the evaluation of DL‐based methods with fidelity‐...

show abstract

Collimator-Detector Response Compensation in SPECT

Frey

Tsui

2006

Quantitative Analysis in Nuclear Medicine Imaging

View full text Add to dashboard Cite

Application of task-based measures of image quality to optimization and evaluation of three-dimensional reconstruction-based compensation methods in myocardial perfusion SPECT

Cited by 116 publications

References 32 publications

Comparison of channelized hotelling and human observers in determining optimum OSEM reconstruction parameters for myocardial SPECT

Comparison of channelized hotelling and human observers in determining optimum OSEM reconstruction parameters for myocardial SPECT

Need for objective task‐based evaluation of deep learning‐based denoising methods: A study in the context of myocardial perfusion SPECT

Collimator-Detector Response Compensation in SPECT

Contact Info

Product

Resources

About