Collimator optimization in myocardial perfusion SPECT using the ideal observer and realistic background variability for lesion detection and joint detection and localization tasks

Ghaly, Michael; Du, Yong; Links, Jonathan M.; Frey, Eric

doi:10.1088/0031-9155/61/5/2048

Cited by 12 publications

(7 citation statements)

References 49 publications

(69 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

Section: Introductionmentioning

confidence: 99%

“…This realistic phantom includes parameterized models for anatomy, which allows the generation of a series of phantoms with different anatomical variations. These phantoms have been used in Nuclear Medicine imaging and CT research, [6][7][8][9][10][11][12] as well as in various applications of deep learning. [13][14][15] In the XCAT phantom, changing the values of parameters that control organ anatomy can be used to vary the volumes and shapes of some tissues.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Generating anthropomorphic phantoms using fully unsupervised deformable image registration with convolutional neural networks

Chen

et al. 2020

Medical Physics

Self Cite

View full text Add to dashboard Cite

Purpose: Computerized phantoms have been widely used in nuclear medicine imaging for imaging system optimization and validation. Although the existing computerized phantoms can model anatomical variations through organ and phantom scaling, they do not provide a way to fully reproduce the anatomical variations and details seen in humans. In this work, we present a novel registration-based method for creating highly anatomically detailed computerized phantoms. We experimentally show substantially improved image similarity of the generated phantom to a patient image. Methods: We propose a deep-learning-based unsupervised registration method to generate a highly anatomically detailed computerized phantom by warping an XCAT phantom to a patient computed tomography (CT) scan. We implemented and evaluated the proposed method using the NURBSbased XCAT phantom and a publicly available low-dose CT dataset from TCIA. A rigorous tradeoff analysis between image similarity and deformation regularization was conducted to select the loss function and regularization term for the proposed method. A novel SSIM-based unsupervised objective function was proposed. Finally, ablation studies were conducted to evaluate the performance of the proposed method (using the optimal regularization and loss function) and the current state-ofthe-art unsupervised registration methods. Results: The proposed method outperformed the state-of-the-art registration methods, such as SyN and VoxelMorph, by more than 8%, measured by the SSIM and less than 30%, by the MSE. The phantom generated by the proposed method was highly detailed and was almost identical in appearance to a patient image. Conclusions: A deep-learning-based unsupervised registration method was developed to create anthropomorphic phantoms with anatomies labels that can be used as the basis for modeling organ properties. Experimental results demonstrate the effectiveness of the proposed method. The resulting anthropomorphic phantom is highly realistic. Combined with realistic simulations of the image formation process, the generated phantoms could serve in many applications of medical imaging research.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Generating anthropomorphic phantoms using fully unsupervised deformable image registration with convolutional neural networks

Chen

et al. 2020

Medical Physics

Self Cite

View full text Add to dashboard Cite

show abstract

“…Thus, evaluation of these approaches should be based on the clinical task for which imaging was performed. Multiple methods have been developed to perform this task‐based assessment of image quality, 1–3 and the efficacy of this evaluation procedure has been demonstrated in multiple studies 4–10 …”

Section: Introductionmentioning

confidence: 99%

“…Multiple methods have been developed to perform this task-based assessment of image quality, [1][2][3] and the efficacy of this evaluation procedure has been demonstrated in multiple studies. [4][5][6][7][8][9][10] Currently, DL-based denoising methods for medical imaging are typically evaluated using figures of merit (FoMs) such as root mean squared error (RMSE) and structural similarity index measure (SSIM). [11][12][13] These FoMs measure fidelity between the images obtained using DL-based denoising approaches as compared to some reference images.…”

Section: Introductionmentioning

confidence: 99%

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

“…A large number of images with known truth is needed to provide reliable estimates of these quantities (Fukunaga and Hayes, 1989a, Kupinski et al, 2007, Ge et al, 2014). Furthermore, in large optimization and evaluation studies, task performance is assessed for many different combinations of system parameters and methods, such as different collimator designs (Yihuan et al, 2014, Ghaly et al, 2016), reconstruction methods and parameters (Frey et al, 2002, Gilland et al, 2006, He et al, 2006), and post-reconstruction filters and processing techniques (Frey et al, 2002, Sankaran et al, 2002). Thus, ensemble techniques require an enormous number of images to be obtained and stored; which often limits the number of parameters that can be explored.…”

Section: Introductionmentioning

confidence: 99%

A comparison of resampling schemes for estimating model observer performance with small ensembles

et al. 2017

Self Cite

View full text Add to dashboard Cite

In objective assessment of image quality, an ensemble of images is used to compute the 1st and 2nd order statistics of the data. Often, only a finite number of images is available, leading to the issue of statistical variability in numerical observer performance. Resampling-based strategies can help overcome this issue. In this paper, we compared different combinations of resampling schemes (the leave-one-out (LOO) and the half-train/half-test (HT/HT)) and model observers (the conventional channelized Hotelling observer (CHO), channelized Linear discriminant (CLD) and channelized Quadratic discriminant (CQD)). Observer performance was quantified by the area under the ROC curve (AUC). For a binary classification task and for each observer, the AUC value for an ensemble size of 2000 samples per class served as a gold standard for that observer. Results indicated that each observer yielded different performance depending on the ensemble size and the resampling scheme. For a small ensemble size, the combination [CHO, HT/HT] had more accurate rankings than the combination [CHO, LOO]. Using the LOO scheme, the CLD and CHO had similar performance for large ensembles. However, the CLD outperformed the CHO and gave more accurate rankings for smaller ensembles. As the ensemble size decreased, the performance of [CHO, LOO] combination was seriously deteriorated as opposed to the [CLD, LOO] combination. Thus, it might be desirable to use the CLD with the LOO scheme when smaller ensemble size is available.

show abstract

Collimator optimization in myocardial perfusion SPECT using the ideal observer and realistic background variability for lesion detection and joint detection and localization tasks

Cited by 12 publications

References 49 publications

Generating anthropomorphic phantoms using fully unsupervised deformable image registration with convolutional neural networks

Generating anthropomorphic phantoms using fully unsupervised deformable image registration with convolutional neural networks

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

A comparison of resampling schemes for estimating model observer performance with small ensembles

Contact Info

Product

Resources

About