Iterative Low-Dose CT Reconstruction With Priors Trained by Artificial Neural Network

Wu, Dufan; Kim, Kyungsang; Fakhri, Georges El; Li, Quanzheng

doi:10.1109/tmi.2017.2753138

Cited by 210 publications

(125 citation statements)

References 29 publications

Supporting

Mentioning

124

Contrasting

Unclassified

Order By: Relevance

“…Use of a fully IR technique improves image quality without altering hemodynamic parameters on low-dose dynamic CTP when compared with FBP or hybrid IR [39]. Recently, artificial intelligence was used to improve CT image reconstruction [75]. These techniques have the potential to optimize the quality of low-dose myocardial CTP images with shortening of reconstruction times [76,77].…”

Section: Iterative Reconstruction and Other Algorithms For Ctpmentioning

confidence: 99%

Computed tomographic evaluation of myocardial ischemia

et al. 2020

View full text Add to dashboard Cite

Myocardial ischemia is caused by a mismatch between myocardial oxygen consumption and oxygen delivery in coronary artery disease (CAD). Stratification and decision-making based on ischemia improves the prognosis in patients with CAD. Non-invasive tests used to evaluate myocardial ischemia include stress electrocardiography, echocardiography, single-photon emission computed tomography, and magnetic resonance imaging. Invasive fractional flow reserve is considered the reference standard for assessment of the hemodynamic significance of CAD. Computed tomography (CT) angiography has emerged as a first-line imaging modality for evaluation of CAD, particularly in the population at low to intermediate risk, because of its high negative predictive value; however, CT angiography does not provide information on the hemodynamic significance of stenosis, which lowers its specificity. Emerging techniques, e.g., CT perfusion and CT-fractional flow reserve, help to address this limitation of CT, by determining the hemodynamic significance of coronary artery stenosis. CT perfusion involves acquisition during the first pass of contrast medium through the myocardium following pharmacological stress. CTfractional flow reserve uses computational fluid dynamics to model coronary flow, pressure, and resistance. In this article, we review these two functional CT techniques in the evaluation of myocardial ischemia, including their principles, technology, advantages, limitations, pitfalls, and the current evidence.

show abstract

Section: Iterative Reconstruction and Other Algorithms For Ctpmentioning

confidence: 99%

Computed tomographic evaluation of myocardial ischemia

et al. 2020

View full text Add to dashboard Cite

show abstract

“…In other words, the noise-training approach is well-suited for training a network to robustly perform the inverse Radon transform. The incorporation of prior information has proven to be highly powerful, initially through Markov Random Field regularization techniques 34 , later in approaches such as dictionary learning 34 and non-local means 35 , and most recently in deep learning based priors 1,36,37 . Hence, we expect performance can be greatly enhanced by relying entirely on a wide variety of high-quality clinical images and corresponding noisy sinograms.…”

Section: Discussionmentioning

confidence: 99%

A hierarchical approach to deep learning and its application to tomographic reconstruction

Man

2019

15th International Meeting on Fully Three-Dimensional Image Reconstruction in Radiology and Nuclear Medicine

View full text Add to dashboard Cite

Deep learning (DL) has shown unprecedented performance for many image analysis and image enhancement tasks. Yet, solving large-scale inverse problems like tomographic reconstruction remains challenging for DL. These problems involve non-local and space-variant integral transforms between the input and output domains, for which no efficient neural network models have been found. A prior attempt to solve such problems with supervised learning relied on a brute-force fully connected network and applied it to reconstruction for a 128 4 system matrix size. This cannot practically scale to realistic data sizes such as 512 4 and 512 6 for three-dimensional data sets. Here we present a novel framework to solve such problems with deep learning by casting the original problem as a continuum of intermediate representations between the input and output data. The original problem is broken down into a sequence of simpler transformations that can be well mapped onto an efficient hierarchical network architecture, with exponentially fewer parameters than a generic network would need. We applied the approach to computed tomography (CT) image reconstruction for a 512 4 system matrix size. To our knowledge, this enabled the first data-driven DL solver for full-size CT reconstruction without relying on the structure of direct (analytical) or iterative (numerical) inversion techniques. The proposed approach is applicable to other imaging problems such as emission and magnetic resonance reconstruction. More broadly, hierarchical DL opens the door to a new class of solvers for general inverse problems, which could potentially lead to improved signal-to-noise ratio, spatial resolution and computational efficiency in various areas.

show abstract

“…Recently, deep learning methods have demonstrated strong capability in many computer vision areas, and researchers have begun to apply deep learning on CT reconstruction . For instance, regularizations trained by neural networks have been investigated for SIR of low‐dose CT to capture complex image features . It can be expected that more deep learning‐based regularization strategies will be investigated in the near future.…”

Section: Discussionmentioning

confidence: 99%

Regularization strategies in statistical image reconstruction of low‐dose x‐ray CT: A review

et al. 2018

View full text Add to dashboard Cite

Statistical image reconstruction (SIR) methods have shown potential to substantially improve the image quality of low-dose x-ray computed tomography (CT) as compared to the conventional filtered back-projection (FBP) method. According to the maximum a posteriori (MAP) estimation, the SIR methods are typically formulated by an objective function consisting of two terms: (a) a data-fidelity term that models imaging geometry and physical detection processes in projection data acquisition, and (b) a regularization term that reflects prior knowledge or expectations of the characteristics of the to-be-reconstructed image. SIR desires accurate system modeling of data acquisition, while the regularization term also has a strong influence on the quality of reconstructed images. A variety of regularization strategies have been proposed for SIR in the past decades, based on different assumptions, models, and prior knowledge. In this paper, we review the conceptual and mathematical bases of these regularization strategies and briefly illustrate their efficacies in SIR of low-dose CT.

show abstract

Iterative Low-Dose CT Reconstruction With Priors Trained by Artificial Neural Network

Cited by 210 publications

References 29 publications

Computed tomographic evaluation of myocardial ischemia

Computed tomographic evaluation of myocardial ischemia

A hierarchical approach to deep learning and its application to tomographic reconstruction

Regularization strategies in statistical image reconstruction of low‐dose x‐ray CT: A review

Contact Info

Product

Resources

About