Suremap: Predicting Uncertainty in Cnn-Based Image Reconstructions Using Stein’s Unbiased Risk Estimate

Kitichotkul, Ruangrawee; Metzler, Christopher A.; Ong, Frank; Wetzstein, Gordon

doi:10.1109/icassp39728.2021.9414306

Cited by 7 publications

(2 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These images can lead practitioners to trust the reconstruction, even when they should not. Fortunately, because the last step of D-VDAMP is equivalent to removing colored Gaussian noise with known covariance from the ground-truth signal, Stein's unbiased risk estimate can be combined with D-VDAMP to produce heat maps of the expected squared error per pixel associated with the reconstruction [27,28,29]. This allows practitioners to judge which portions of the reconstruction can be trusted.…”

Section: Discussionmentioning

confidence: 99%

D-VDAMP: Denoising-Based Approximate Message Passing for Compressive MRI

Metzler

Wetzstein

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

Plug and play (P&P) algorithms iteratively apply highly optimized image denoisers to impose priors and solve computational image reconstruction problems, to great effect. However, in general the "effective noise", that is the difference between the true signal and the intermediate solution, within the iterations of P&P algorithms is neither Gaussian nor white. This fact makes existing denoising algorithms suboptimal.In this work, we propose a CNN architecture for removing colored Gaussian noise and combine it with the recently proposed VDAMP algorithm, whose effective noise follows a predictable colored Gaussian distribution. We apply the resulting denoising-based VDAMP (D-VDAMP) algorithm to variable density sampled compressive MRI where it substantially outperforms existing techniques.

show abstract

Section: Discussionmentioning

confidence: 99%

D-VDAMP: Denoising-Based Approximate Message Passing for Compressive MRI

Metzler

Wetzstein

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

show abstract

“…In the recent development, deep learning has emerged as a promising tool to aid CS system in data recovery (Kitichotkul et al, 2021; Lu et al, 2018; Qiao et al, 2020). Deep neural network (DNN) consists of more hidden layers compared to artificial neural network (ANN).…”

Section: Introductionmentioning

confidence: 99%

Efficient reconstruction scheme with deep neural network for highly compressive sensing of fiber Bragg grating spectrum

Lim

Soon

et al. 2023

Transactions of the Institute of Measurement and Control

View full text Add to dashboard Cite

In this work, we propose an efficient reconstruction scheme for compressive sensing (CS) of fiber Bragg grating (FBG) spectrum. Taking advantage of the sparse reflection spectrum of the FBG array network, we have demonstrated the use of CS for compressing the spectrum at an excessively high compression factor up to 64. In addition to that, the spectral difference (SD) of the spectra is used to further enhance their sparsity for the CS model. In this investigation, four different configurations have been devised and tested to compare their performance and effectiveness. Configuration IV that is based on SD and deep neural network offers the best recovery performance. The proposed method is a potential tool for efficient data storage and transmission for FBG sensor network.

show abstract

PixCUE: Joint Uncertainty Estimation and Image Reconstruction in MRI using Deep Pixel Classification

Ekanayake,

Pawar,

Chen

et al. 2024

J Digit Imaging. Inform. med.

View full text Add to dashboard Cite

Deep learning (DL) models are effective in leveraging latent representations from MR data, emerging as state-of-the-art solutions for accelerated MRI reconstruction. However, challenges arise due to the inherent uncertainties associated with undersampling in k-space, coupled with the over- or under-parameterized and opaque nature of DL models. Addressing uncertainty has thus become a critical issue in DL MRI reconstruction. Monte Carlo (MC) inference techniques are commonly employed to estimate uncertainty, involving multiple reconstructions of the same scan to compute variance as a measure of uncertainty. Nevertheless, these methods entail significant computational expenses, requiring multiple inferences through the DL model. In this context, we propose a novel approach to uncertainty estimation during MRI reconstruction using a pixel classification framework. Our method, PixCUE (Pixel Classification Uncertainty Estimation), generates both the reconstructed image and an uncertainty map in a single forward pass through the DL model. We validate the efficacy of this approach by demonstrating that PixCUE-generated uncertainty maps exhibit a strong correlation with reconstruction errors across various MR imaging sequences and under diverse adversarial conditions. We present an empirical relationship between uncertainty estimations using PixCUE and established reconstruction metrics such as NMSE, PSNR, and SSIM. Furthermore, we establish a correlation between the estimated uncertainties from PixCUE and the conventional MC method. Our findings affirm that PixCUE reliably estimates uncertainty in MRI reconstruction with minimal additional computational cost.

show abstract

Suremap: Predicting Uncertainty in Cnn-Based Image Reconstructions Using Stein’s Unbiased Risk Estimate

Cited by 7 publications

References 29 publications

D-VDAMP: Denoising-Based Approximate Message Passing for Compressive MRI

D-VDAMP: Denoising-Based Approximate Message Passing for Compressive MRI

Efficient reconstruction scheme with deep neural network for highly compressive sensing of fiber Bragg grating spectrum

PixCUE: Joint Uncertainty Estimation and Image Reconstruction in MRI using Deep Pixel Classification

Contact Info

Product

Resources

About