View Presentation Video

Siying Xu¹, Kerstin Hammernik^2,3, Patrick Krumm¹, Sergios Gatidis^1,4, and Thomas Küstner^1
1Medical Image and Data Analysis (MIDAS.lab), Department of Diagnostic and Interventional Radiology, University Hospital of Tuebingen, Tuebingen, Germany, ²Lab for AI in Medcine, Technical University of Munich, Munich, Germany, ³Department of Computing, Imperial College London, London, United Kingdom, ⁴Max Planck Institute for Intelligent Systems, Tuebingen, Germany

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, Image Reconstruction, Heart

Cardiac CINE MR imaging requires long acquisitions under multiple breath-holds. With the development of deep learning-based reconstruction methods, the acceleration rate and reconstructed image quality have been increased. However, existing methods face several shortcomings, such as limited information-sharing across domains and generalizability which may restrict their clinical adoption. To address these issues, we propose A-LIKNet which incorporates attention mechanisms and maximizes information sharing between low-rank, image, and k-space in an interleaved architecture. Results indicate that the proposed A-LIKNet outperforms other methods for up to 24x accelerated acquisitions within a single breath-hold.

View Presentation Video

Ruimin Feng¹, Qing Wu², Yuyao Zhang², and Hongjiang Wei^1
1School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China, ²School of Information Science and Technology, ShanghaiTech University, Shanghai, China

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, implicit neural representation

Current parallel imaging techniques for MRI acceleration can still not reliably reconstruct a high-quality image from highly reduced k-space measurements with fewer calibration data. In this study, we applied the new insight of implicit neural representation (INR) to parallel MRI reconstruction. The underlying MRI image and coil sensitivities were modeled as continuous functions of spatial coordinates. These functions were simultaneously learned from the sparsely-acquired k-space itself without additional training data. Thanks to the continuous representation and joint estimation scheme, the proposed method outperforms the existing scan-specific methods, demonstrating its potential for further accelerating the MRI acquisition.

View Presentation Video

Aniket Pramanik¹ and Mathews Jacob^1
1Electrical and Computer Engineering, University of Iowa, Iowa City, IA, United States

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence

MRI has been revolutionized by compressed sensing algorithms, which offer guaranteed uniqueness, convergence, and stability. In the recent years, model-based deep learning methods have been emerging as more powerful alternatives for image recovery. The main focus of this paper is to introduce a model based algorithm with similar theoretical guarantees as CS methods. The proposed deep equilibrium formulation is significantly more memory-efficient than unrolled methods, which allows us to apply it to 3D or 2D+time problems that current unrolled algorithms cannot handle. Our results also show that the approach is more robust to input perturbations than current unrolled methods.

View Presentation Video

Yilin Liu¹, Yong Chen², and Pew-thian Yap^3
1Department of Computer Science, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, ²Department of Radiology, Case Western Reserve University, Cleveland, OH, United States, ³Department of Radiology, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence

This work aims to simplify deep image prior (DIP) architectural design decisions in the context of unsupervised accelerated MRI reconstruction, facilitating the deployment of MRI-DIP in real-world settings. We first show that architectures inappropriate for specific MRI datasets (knee, brain) can lead to severe reconstruction artifacts, and then demonstrate that proper network regularization can dramatically improve image quality irrespective of network architectures. The proposed plug-and-play regularization is applicable to any network form without architectural modifications.

View Presentation Video

Maarten Terpstra^1,2, Kerstin Hammernik^3,4, Thomas Küstner⁵, Matteo Maspero^1,2, Cornelis van den Berg^1,2, and Daniel Rueckert^3,4
1Department of Radiotherapy, University Medical Center Utrecht, Utrecht, Netherlands, ²Computational Imaging Group for MR Diagnostics & Therapy, University Medical Center Utrecht, Utrecht, Netherlands, ³Lab for AI in Medcine, Technical University of Munich, Munich, Germany, ⁴Department of Computing, Imperial College London, London, United Kingdom, ⁵Medical Image And Data Analysis (MIDAS.lab), University Hospital of Tübingen, Tübingen, Germany

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence

Machine learning (ML) has become a powerful technique for reconstructing undersampled MRI. However, applying ML to MRI reconstruction requires several essential building blocks, consisting of general operations and MRI-specific operators in the context of reconstruction. While the former are available in generic frameworks such as Keras/TensorFlow or PyTorch, the MR-specific operators are generally custom-implemented. MERLIN was proposed as a generic toolkit for ML-based medical imaging to harmonize the machine learning landscape and to provide complex-valued interfaces for commonly used back-ends. Here, we evaluate MERLIN as a cross-platform toolkit for complex-valued image reconstruction.

View Presentation Video

Prakash Kumar¹ and Krishna S Nayak^1
1Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, United States

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, Real-Time

Real-time MRI captures movements and dynamic processes in human body without reliance on any repetition or synchronization. Many applications require low-latency (typically <200ms) for guidance of interventions or closed-loop feedback. Standard constrained optimization methods are too slow to be implemented “online”. We demonstrate a non-Cartesian deep learning image reconstruction method based on the end-to-end variational network. Training data are created using traditional high latency compressed sensing reconstruction as the reference, with the goal of achieving similar results with low latency. We demonstrate reconstruction latencies of 95ms per frame, with nRMSE of 0.045 and SSIM of 0.95.

View Presentation Video

Frank Zijlstra^1,2 and Peter T While^1,2
1Department of Radiology and Nuclear Medicine, St. Olav's University Hospital, Trondheim, Norway, ²Department of Circulation and Medical Imaging, NTNU - Norwegian University of Science and Technology, Trondheim, Norway

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, synthetic data, compressed sensing, accelerated imaging

This study demonstrates using generative deep learning for transforming magnitude-only images into synthetic raw data for deep-learning-base image reconstruction. Using relatively few raw datasets, a set of neural networks was trained to generate the missing phase and coil sensitivity information in magnitude-only images. These maps are then recombined into synthetic raw data. We trained end-to-end variational networks for 4-fold accelerated compressed sensing reconstruction on the FastMRI dataset, with increasing training set size. Synthetic raw data showed similar improvements as real raw data with increasing data. This shows promise for applying deep-learning-based image reconstruction when raw data is scarce.

View Presentation Video

Alper Güngör^1,2,3, Salman Ul Hassan Dar^1,2, Şaban Öztürk^1,2,4, Yilmaz Korkmaz^1,2, Gokberk Elmas^1,2, Muzaffer Ozbey^1,2, and Tolga Çukur^1,2,5
1Department of Electrical and Electronics Engineering, Bilkent University, Ankara, Turkey, ²National Magnetic Resonance Research Center (UMRAM), Bilkent University, Ankara, Turkey, ³ASELSAN Research Center, Ankara, Turkey, ⁴Amasya University, Amasya, Turkey, ⁵Neuroscience Program, Bilkent University, Ankara, Turkey

Keywords: Image Reconstruction, Image Reconstruction

Learning-based MRI reconstruction is commonly performed using non-adaptive models with frozen weights during inference. Non-adaptive conditional models poorly generalize across variable imaging operators, whereas non-adaptive unconditional models poorly generalize across variations in the image distribution. Here, we introduce a novel adaptive method, AdaDiff, that trains an unconditional diffusion prior for high-fidelity image generations and adapts the prior during inference for improved generalization. AdaDiff outperforms state-of-the-art baselines both visually and quantitatively. 

View Presentation Video

Tae Hyung Kim^1,2,3, Jaejin Cho^2,3, Borjan Gagoski^3,4, Zijing Dong^2,3, Fuyixue Wang^2,3, So Hyun Kang¹, and Berkin Bilgic^2,3,5
1Department of Computer Engineering, Hongik University, Seoul, Korea, Republic of, ²Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston, MA, United States, ³Department of Radiology, Harvard Medical School, Boston, MA, United States, ⁴Fetal-Neonatal Neuroimaging & Developmental Science Center, Boston Children’s Hospital, Boston, MA, United States, ⁵Harvard/MIT Health Sciences and Technology, Cambridge, MA, United States

Keywords: Image Reconstruction, Brain

 The subspace method has been widely used for multi-echo/contrast MRI reconstruction, assuming the temporal MR signal evolution can be compactly represented using a few linear coefficients.  Recently, methods based on artificial neural networks (trained with large datasets) enabled nonlinear representations of temporal or spatio-temporal MR signals and demonstrated improved performance.  This work proposes a novel zero-shot spatio-temporal generative prior for multi-echo/contrast MRI reconstruction, assuming the  spatio-temporal MR signals can be nonlinearly generated using deep generative neural networks without external training data.  The proposed method was evaluated with 3D-QALAS and EPTI data, and exhibited substantial improvement in NRMSE against existing methods.

View Presentation Video

Heng Yu¹, Yamin Arefeen², and Berkin Bilgic^3,4
1The Robotics Institute, Carnegie Mellon University, Pittsburgh, PA, United States, ²Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States, ³Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, ⁴Department of Radiology, Harvard Medical School, Boston, MA, United States

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, Deep Learning

Recently introduced zero-shot self-supervised learning (ZS-SSL) has shown potential in accelerated MRI in a scan-specific scenario, which enabled high quality reconstructions without access to a large training dataset. ZS-SSL has been further combined with the subspace model to accelerate 2D T2-shuffling acquisitions. In this work, we propose a parallel network framework and introduce attention mechanism to improve subspace based zero-shot self-supervised learning and enable higher acceleration factors. We name our method SubZero and demonstrate that it can achieve improved performance compared with current methods in T₁ and T₂ mapping acquisitions .

View Presentation Video

Charles Millard¹ and Mark Chiew^1
1Wellcome Centre for Integrative Neuroimaging, FMRIB, University of Oxford, Oxford, United Kingdom

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, Self-supervised learning

To train a neural network to recover images with sub-sampled examples only, Self-Supervised Learning via Data Undersampling (SSDU) proposes partitioning the sampling mask into two subsets and training a network to recover one from the other. Recent work on the connection between SSDU and the self-supervised denoising method Noisier2Noise has shown that superior reconstruction quality is possible when the distribution of the partition matches the sampling mask. We test this principle on three types of sampling schemes and find that the test set loss is indeed closest to a fully supervised benchmark when the partition distribution is matched.

View Presentation Video

Ismail Arda Vurankaya¹, Yohan Jun^2,3, Jaejin Cho^2,3, and Berkin Bilgic^2,3
1Electrical and Electronics Engineering, Bogazici University, Istanbul, Turkey, ²Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, ³Department of Radiology, Harvard Medical School, Boston, MA, United States

Keywords: Image Reconstruction, Image Reconstruction

We propose a zero-shot self-supervised learning (ZS-SSL) approach for accelerated diffusion MRI reconstruction. Our method builds on the approach in [3] for subject-specific MRI reconstruction. We perform reconstruction across all diffusion directions with a single model, rather than different models for each direction, reducing computation time. We partition the directions as training and validation directions. We train the model on training directions, while keeping track of validation loss. We test our model on entire directions, evaluating the reconstruction quality of a single network across all directions. Jointly trained ZS-SSL provides better reconstructions than standard parallel imaging, while remaining computationally efficient.

View Presentation Video

Seonghyuk Kim¹, Wonil Lee¹, Namho Jeong¹, Jeewon Kim¹, Jongyeon Lee¹, Beomgu Kang¹, and HyunWook Park^1
1Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea, Republic of

Keywords: Parallel Imaging, Machine Learning/Artificial Intelligence, Noise-robust method

We propose a novel loss function that increases noise-robustness in accelerated parallel MR image reconstruction. The loss function is based on the variance of the background area in the noisy undersampled image and that of the difference image between noisy undersampled image and the synthesized undersampled image. The proposed loss function provides stronger regularization and robustness when applied along with other constraints. We show that the application of the proposed loss function boosts performance of the network, yielding improved quality of reconstructed image.

View Presentation Video

Zhehong Zhang¹, Kartiga Selvaganesan¹, Yonghyun Ha², Chenhao Sun², Anja Samardzija¹, Heng Sun¹, Gigi Galiana^1,2, and R. Todd Constable^1,2,3,4
1Department of Biomedical Engineering, School of Engineering and Applied Science, Yale University, New Haven, CT, United States, ²Department of Radiology and Biomedical Imaging, School of Medicine, Yale University, New Haven, CT, United States, ³Interdepartmental Neuroscience Program, School of Medicine, Yale University, New Haven, CT, United States, ⁴Department of Neurosurgery, School of Medicine, Yale University, New Haven, CT, United States

Keywords: Image Reconstruction, Signal Representations, Nonlinear Encoding

Model-based deep learning reconstruction with a nonlinear encoding matrix poses unique challenges to GPU memory, due to the densely connected computational graph nodes in the physics model part. In this work, SVD compression is demonstrated as necessary for such networks, and it is applied to the highly nonlinear case of Bloch-Siegert encoding from a low-field MR scanner. The redundancy across all nonlinear encoding dimensions is exploited for compression. With the compressed encoding matrix, the model-based network is feasible to implement. It outperforms the traditional reconstruction at all levels of simulated Gaussian noise and has advantages over commonly used regularization terms. 

View Presentation Video

Samah Khawaled¹ and Moti Freiman ^2
1The Interdisciplinary Applied mathematics Program, Technion, Haifa, Israel, ²Faculty of Biomedical Engineering, Technion, Haifa, Israel

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence

Deep learning (DL) models are currently employed to reconstruct high quality MRI image from undersampled k-space measurements. Yet, uncertainty quantification in images reconstructed by such models is critical for reliable clinical decision making. We propose NPBREC, a non-parametric Bayesian approach for uncertainty estimation in DL-based MRI reconstruction. We demonstrated the added-value of our Bayesian registration framework on the fastmri multi-coil brain MRI dataset, compared to the baseline E2E-VarNet trained with and without inference-time dropout for uncertainty quantification. Our NPBREC approach demonstrated both improved reconstruction accuracy and a better correlation between reconstruction errors and uncertainty measures.  

View Presentation Video

Zhenghan Fang¹, Kuo-Wei Lai¹, Peter van Zijl¹, Xu Li¹, and Jeremias Sulam^1
1The Johns Hopkins University, Baltimore, MD, United States

Keywords: Image Reconstruction, Susceptibility, Susceptibility Tensor Imaging

The application of STI in human in vivo has been practically infeasible because of its time-consuming acquisition scheme. We propose a novel image reconstruction algorithm for STI that leverages data-driven priors to tackle this issue. Our method, called DeepSTI, learns the data prior implicitly via a deep neural network that resembles the proximal operator of a regularizer function. The dipole inversion problem is then solved iteratively using the learned proximal network. Experimental results demonstrate superior performance of DeepSTI over state-of-the-art methods. DeepSTI is the first reconstruction method to achieve high quality results for human STI with fewer than six orientations.

View Presentation Video

Zhikai Yang^1,2, Liu Yang^1,2, Rohith Saai Pemmasani Prabakaran², AbdulMojeed Olabisi ILYAS², Jianpan Huang¹, and Kannie W. Y. Chan^1,2,3,4,5
1Department of Biomedical Engineering, City University of Hong Kong, Hong Kong, Hong Kong, ²Hong Kong Centre for Cerebro-Cardiovascular Health Engineering (COCHE), Hong Kong, Hong Kong, ³City University of Hong Kong Shenzhen Research Institute, Shenzhen, China, ⁴Russell H. Morgan Department of Radiology and Radiological Science, The Johns Hopkins University School of Medicine, Baltimore, MD, United States, ⁵Tung Biomedical Sciences Centre, City University of Hong Kong, Hong Kong, Hong Kong

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence

We proposed an attention-based multi-offset network to exploit redundant anatomy information for the reconstruction of CEST-MR image (AMO-CEST). To the best of our knowledge, this is the first work using deep learning with varied radial sample patterns and multi-offset slices as input to accelerate CEST-MRI. Compared with other deep learning-based methods on the four times under-sampling mouse brain CEST dataset, the AMO-CEST achieved the best performance with an MMSE of , a PSNR of  dB, and an SSIM . In conclusion, the proposed AMO-CEST network can accelerate the CEST-MRI at high down-sampling rate while maintaining good image quality.

View Presentation Video

Rosie Goodburn¹, Movindu Dassanayake², Bastien Lecoeur¹, Prashant Nair¹, Uwe Oelfke¹, and Andreas Wetscherek^1
1Institute of Cancer Research, London, United Kingdom, ²Imperial College London, London, United Kingdom

Keywords: Image Reconstruction, Radiotherapy

Respiratory resolved (4D)-MRI is expected to benefit MRI-guided radiotherapy for abdominal-thoracic cancers. However, a current limitation of 4D-MRI is that one has to trade-off between artefacts, spatial-temporal resolution, and spatial coverage. Deformable image registration has been employed to enhance image quality of undersampled 4D-MRIs, but such approaches are usually too slow to be used in MR-guided radiotherapy workflows. We investigated the feasibility of employing models trained using the VoxelMorph framework with a view of leveraging a fast computation to minimise 4D-MRI reconstruction time on the MR-Linac. The models performed well and were ~48 times faster than a common registration method.

View Presentation Video

Jae-Hun Lee¹, Dongyeob Han², Jae-Yoon Kim¹, and Dong-Hyun Kim^1
1Department of Electrical and Electronic Engineering, Yonsei Univ., Seoul, Korea, Republic of, ²Siemens Healthineers Ltd, Siemens Korea, Seoul, Korea, Republic of

Keywords: Parallel Imaging, Image Reconstruction, Quantitative Imaging, White Matter

Recently, acceleration of 3D multi-echo gradient-echo (mGRE) acquisition for myelin water imaging (MWI) has been achieved using parallel imaging (PI) or deep learning network. However, these methods typically allow a low acceleration factor (R) for MWI because of the high sensitivity of the MWI estimation routine with respect noise/artifacts. Here, we developed a reconstruction deep learning network called the jointly unrolled cross-domain optimization-based spatio-temporal reconstruction network. According to retrospective and prospective reconstruction results, the proposed method achieved high-fidelity performance on the reconstructed mGRE images and MWI maps.