Keywords: AI/ML Image Reconstruction, Image Reconstruction, Deep Generative Models, Inverse Problems, Unsupervised Learning, Denoising

Motivation: Publicly available k-space data used for training are inherently noisy with no available ground truth.

Goal(s): To denoise k-space data in an unsupervised manner for downstream applications.

Approach: We use Generalized Stein’s Unbiased Risk Estimate (GSURE) applied to multi-coil MRI to denoise images without access to ground truth. Subsequently, we train a generative model to show improved accelerated MRI reconstruction.

Results: We demonstrate: (1) GSURE can successfully remove noise from k-space; (2) generative priors learned on GSURE-denoised samples produce realistic synthetic samples; and (3) reconstruction performance on subsampled MRI improves using priors trained on denoised images in comparison to training on noisy samples.

Impact: This abstract shows that we can denoise multi-coil data without ground truth and train deep generative models directly on noisy k-space in an unsupervised manner, for improved accelerated reconstruction.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence, Image reconstruction, diffusion models, deep learning

Motivation: Diffusion models can reconstruct high-quality MR images, but their training neglects physical constraints and requires supervision via ground-truth images derived from fully-sampled acquisitions.

Goal(s): Our goal was to devise a diffusion-based method that incorporates physical constraints and that can be trained using undersampled acquisitions.

Approach: We introduced a novel diffusion model (SSDiffRecon) based on a physics-driven unrolled transformer architecture; and self-supervised training was achieved by predicting held-out subsets of acquired k-space data from remaining subsets.

Results: SSDiffRecon achieved superior reconstructions to alternative self-supervised methods, and performed on par with a supervised benchmark trained on fully-sampled acquisitions.

Impact: The improvement in image quality and acquisition speed through SSDiffRecon, combined with the ability to train on undersampled acquisitions, may facilitate adoption of AI-based reconstruction for comprehensive MRI exams in many applications, particularly in pediatric and elderly populations.

Keywords: AI/ML Image Reconstruction, Image Reconstruction, Deep learning clinical adpatation

Motivation: This research aims to address the problem of performance degradation when a reconstruction network and a downstream network are cascaded. The proposed solution, MOST, optimizes a MR reconstruction network for multiple downstream tasks.

Goal(s): Our objective is to sequentially finetune a reconstruction network using losses from multiple downstream tasks while preventing catastrophic forgetting such that the same reconstruction network can be used for the multiple tasks.

Approach: We introduce replay-based continual learning into finetuning for multiple downstream tasks.

Results: Our method successfully circumvents catastrophic forgetting, exhibiting stable performance across all downstream tasks, enabling a single reconstruction network to be used for multiple tasks. 

Impact: When k-space reconstruction and downstream tasks are performed using two separate networks (individually optimized), the cascade may introduce suboptimal results. Here, we propose a solution when multiple downsteam tasks exist, addressing challenges in realistic user environment. 

Keywords: AI/ML Image Reconstruction, Multi-Contrast

Motivation: Scans within an MR exam share redundant information due to the same underlying structures. One contrast can hence be used to guide the reconstruction of another, thereby requiring less measurements.

Goal(s): Multimodal guided reconstruction to reduce scanning times.

Approach: Our method exploits AI-based content/style decomposition in an iterative reconstruction algorithm. We explored this concept via numerical simulation and subsequently validated it on in vivo data.

Results: Compared to a conventional compressed sensing baseline, our method showed consistent improvement in simulations and produced sharper reconstructions from undersampled in vivo data. By enforcing data consistency, it was also more reliable than blind image translation.

Impact: In the clinic, this can potentially enable a reduced MR exam time for a given image quality or improve image quality given a scan time budget. The former can reduce strain on the patient, whereas the latter can improve diagnosis.

Keywords: AI/ML Image Reconstruction, Brain

Motivation: Accelerated MRI acquisitions offer reduced imaging scan times but pose challenges in image reconstructions. Tremendous progress has been made to reconstruct accelerated MRI, but it remains challenging to restore high-frequency image details in highly undersampled data.

Goal(s): Our goal is to develop a solution that can restore subtle structures even for highly accelerated MRI.

Approach: We propose CAMP-Net, a consistency-aware multi-prior framework, that leverages scan-specific features with both image and $$$k$$$-space domain knowledge for MRI reconstruction.

Results: Results on a publicly available brain dataset demonstrated that CAMP-Net can achieve high-quality reconstructions with fine brain anatomical structures even at an acceleration factor of 10X.

Impact: The successful restoration of subtle structures for MRI with high acceleration factors can significantly reduce MRI scan time in clinical routines, benefiting patients, increasing the access to MRI, and significantly reducing healthcare cost of MRI. 

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence, Image reconstruction, High dimensional imaging

Motivation: Multidimensional MR spatiospectral imaging (MD-MRSI) has many applications but is challenging due to high dimensionality and limited SNR. Subspace and learning-based methods have both demonstrated success.

Goal(s): To develop a new MD-MRSI reconstruction method synergizing subspace modeling and a spatiospectral denoiser that can be ‘pre-learned’ without noisy/clean image pairs.

Approach: A self-supervised training strategy was proposed to learn a network-based denoiser combining convolutional, fully-connected, and recurrent features and effectively exploiting multidimensional “correlations”. A plug-and-play ADMM-based algorithm was used to integrate the denoising prior and subspace reconstruction.

Results: Impressive SNR-enhancing reconstruction was demonstrated using simulations and in vivo data from different MD-MRSI acquisitions.

Impact: A new approach is proposed for multidimensional MR spatiospectral image reconstruction integrating low-dimensional modeling and a prelearned denoiser trained via multidimensional interpolation using only noisy data. Potential impacts on quantitative molecular imaging are demonstrated using different MRSI acquisitions.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence

Motivation: Spatial resolution, signal-to-noise ratio, and motion artifacts critically matter in any MRI practices. Current methods focus on a single source of known degradation of imaging. A unified framework is desired, which allows for high-quality reconstruction in the face of multiple unknown sources of degradation.

Goal(s): We reconstruct high-quality brain MRI against degradations by motion, noise, and low resolution, with an image-to-image translation-based deep neural framework.

Approach: We developed a prompt-based learning approach and assessed it on a public brain MRI dataset.

Results: Our method offered remarkably improved reconstructions (PSNR=30.96dB, SSIM=0.9133), as compared to two other state-of-the-art methods.

Impact: We developed a new methodology that enables high-quality MRI reconstruction from scans corrupted by a mixture of multiple unknown sources of degradations, which commonly happen in clinical and research MRI studies, with a unified reconstruction framework.

Keywords: AI/ML Image Reconstruction, Fat, Off-Resonance

Motivation: Accelerated MRI protocols and fat/water separation are critical in clinical imaging but are compromised by off-resonance artifacts from B0 inhomogeneities, particularly in non-Cartesian trajectories with longer readouts.

Goal(s): We aim to develop a deep learning framework that enables off-resonance correction from Center-Out Spiral acquisitions, enhancing scan efficiency and image fidelity without extended acquisition times, with the added value of performing fat/water separation.

Approach: Our physics-informed framework employs a multi-frequency bin model trained on synthetic noise data, enabling off-resonance deblurring and extraction of fat and water components without additional acquisition steps.

Results: We showcase our model's efficacy through phantom and in-vivo reconstructions.

Impact: Our physics-informed deep learning framework offers off-resonance correction in Non-Cartesian Spiral MRI, enabling rapid imaging. Our model handles partial volume effects, with the added value of providing fat/water image separation.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence

Motivation: The high memory demand of model-based deep learning algorithms restricts their application in large-scale (eg., 3D/4D) applications. Moreover, their robustness to input perturbations is not well-studied.

Goal(s): To realize a memory efficient MoDL framework with similar theoretical guarantees as compressed sensing methods, while offering state-of-the-art performance.

Approach: We introduce a memory-efficient deep equilibrium framework with theoretical guarantees on uniqueness, convergence, and robustness.

Results: The proposed scheme offers comparable performance to state of the art methods, while being 10 times more memory-efficient. Additionally, the proposed scheme is significantly more robust to Gaussian and adversarial input perturbations.

Impact:  The proposed approach results in greater than 10x reduction in memory demand, which enables the application of MoDL algorithms in large-scale (3D/4D) applications. The theoretically guaranteed robustness of the proposed algorithm reduces the error amplification in highly under-sampled settings.

Keywords: AI/ML Image Reconstruction, Image Reconstruction, Accelerated imaging, compressed sensing, unsupervised learning

Motivation: Alternative unsupervised training methods are needed for training physics-driven deep learning reconstruction without fully-sampled data.

Goal(s): We propose a novel loss formulation, inspired by compressibility, to evaluate reconstruction quality in supervised, unsupervised and zero-shot settings.

Approach: We leverage reweighted $$$\ell_1$$$-norm, which corresponds to $$$\ell_0$$$-norm of a sparse signal, to evaluate reconstruction quality. In supervised setting, reference weights are used for reweighting, while in unsupervised case, they are updated after each reweighting.

Results: Our findings demonstrate that the networks trained with this loss outperform conventional compressed sensing, while performing similarly to deep learning methods trained using established supervised and unsupervised techniques.

Impact: This work proposes an alternative compressibility-inspired loss formulation that is applicable to supervised, unsupervised and zero-shot learning problems for the training of physics-driven reconstruction neural networks. This approach utilizes compressibility and convexity for learning.