View Presentation Video

Laifa Ma^1,2, Liangjun Chen¹, Fenqiang Zhao¹, Zhengwang Wu¹, Li Wang¹, Weili Lin¹, He Zhang³, Kenli Lin², and Gang Li^1
1University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, ²Hunan University, Changsha, China, ³Fudan University, Shanghai, China

Keywords: Image Reconstruction, Brain

Robust motion correction of fetal brain MRI slices is crucial for fetal brain volume reconstruction. However, conventional methods can only handle a limited range of motion. Hence, a deep learning model based on prior geometric constraints is proposed to predict the motion of 2D slices. It consists of a global and a relative motion estimation network. Sharing features between two networks make the model to learn more unique feature representations for global motion correction. Moreover, we present a control point-based approach to simulate complex fetal motion trajectories. The experimental results demonstrate that the proposed method is effective and efficient.

View Presentation Video

Shengjia Chen¹, Patricia Johnson¹, and Yvonne W. Lui^1
1Department of Radiology, New York University Langone Health, New York, NY, United States

Keywords: Image Reconstruction, Brain

We evaluate image quality in brain MR images with pathology, reconstructed by  a deep learning-based image reconstruction algorithm. We have two main contributions: 1) a  procedure for evaluating the image reconstruction quality of images, both globally and in patches with labelled pathology, and 2) report quantitative differences between two groups of reconstructed images (abnormal vs. normal). The pathology evaluation results find pathology regions have more losses and lower structural similarity when compared to normal patches and entire normal brains.

View Presentation Video

Sudhanya Chatterjee¹, Florintina C¹, Rohan Patil¹, Sajith Rajamani¹, Rajagopalan Sundaresan¹, Uday Patil¹, Preetham Shankapal¹, Suresh Emmanuel Joel¹, Ramesh Venkatesan¹, and Harsh Agarwal^1
1GE Healthcare, Bangalore, India

Keywords: Image Reconstruction, Image Reconstruction, artificial intelligence, abdomen, SSFSE

Single shot fast spin echo (SSFSE) is a popular imaging approach for acquisition of high-resolution MR images in motion sensitive areas such as abdomen. In certain clinical settings, use of multiple coils setup for acquisition is not feasible (such as abdominal scans for obese subjects in non-wide bore MRI scanners). SSFSE imaging with single coil using the popular partial Fourier approach only presents risk of excessive blurring in the images. In this study we present a method to enable SSFSE T2 imaging using single coil. Proposed method is evaluated on prospectively accelerated data.

View Presentation Video

Jie Feng¹, Ming Zhang¹, Ruimin Feng¹, and Hongjiang Wei^1
1School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China

Keywords: Image Reconstruction, Quantitative Susceptibility mapping

This study, SSMo-QSM, introduced a self-supervised strategy to achieve a model-based deep learning QSM reconstruction for dealing with the imperfect ground truth in supervised learning methods. In SSMo-QSM, the direct frequency domain division results between phase and dipole kernel at untruncated areas of TKD are randomly separated into two subsets. One is used as the data consistency of the unrolled network and the other is used to define the loss function for training, respectively. The preliminary results of synthetic data suggested that SSMo-QSM performed comparably against supervised methods on accurate susceptibility mapping with suppressed streaking artifact.

View Presentation Video

Peng Li¹, Xiaodi Li¹, Xin Lu², and Yue Hu^1
1Harbin Institute of Technology, Harbin, China, ²De Montfort University, Leicester, United Kingdom

Keywords: Image Reconstruction, MR Fingerprinting

Magnetic resonance fingerprinting (MRF) can achieve simultaneous imaging of multiple tissue parameters. However, the size of the tissue fingerprint dictionary used in MRF grows exponentially as the number of tissue parameters increases, which may result in prohibitively large dictionaries that require extensive computational resources. Existing CNN-based methods obtain parameter reconstruction patch-wisely, using only local information and resulting in limited reconstruction speed. In this paper, we propose a novel end-to-end local and global vision transformer (LG-VIT) for MRF parameter reconstruction. The proposed method enables significantly fast and accurate end-to-end parameter reconstruction while avoiding the high computational cost of high-dimensional data.

View Presentation Video

Shanshan Shan^1,2, Yang Gao^3,4, Meng Ma⁵, Hongping Gan⁵, David Waddington², Brendan Whelan², Paul Liu², Chunyi Liu¹, Mingyuan Gao¹, and Feng Liu^4
1Center for Molecular Imaging and Nuclear Medicine, State Key Laboratory of Radiation Medicine and Protection,School for Radiological and Interdisciplinary Sciences (RAD-X), Soochow University, Suzhou, China, ²ACRF Image X Institute, Faculty of Medicine and Health, The University of Sydney, Sydney, Australia, ³School of Computer Science and Engineering, Central South University, Changsha, China, ⁴School of Information Technology and Electrical Engineering, University of Queensland, Brisbane, Australia, ⁵School of Software, Northwestern Polytechnical University, Suzhou, China

Keywords: Image Reconstruction, Brain

MRI-Linac systems require real-time anatomical images with high geometric fidelity to localize and track tumours during radiotherapy treatments. Image distortions caused by B0 field inhomogeneity and slow MR acquisition hinder the application of real-time MRI-guided radiotherapy. Here, we develop and investigate a deep learning-based reconstruction pipeline to reconstruct B0 inhomogeneity distortion-corrected images (B0ReconNet) directly from k-space. MR acceleration techniques such as compressed sensing (CS) were integrated into B0ReconNet to further reduce acquisition time. Simulated and experimental data with fully sampled and retrospectively subsampled acquisitions on a 1T open bore MRI-Linac were used to validate the proposed method.

View Presentation Video

Uten Yarach¹, Atita Suwannasak¹, and Prapatsorn Sangpin^2
1Radiologic Technology, Chiang Mai University, Chiang Mai, Thailand, ²Philips Healthcare (Thailand), Bangkok, Thailand

Keywords: Image Reconstruction, Brain

Fast spin echo diffusion magnetic resonance imaging (FSE-DWI) is often referred to as a standard for MRI diagnosis of Cholesteatoma. However, the acquired data require multiple steps during image reconstruction which turn out high residual artifacts. In this work, we develop rapid reconstruction for propeller FSE-DWI to improve its signal-to-noise ratio (SNR) through unrolled deep learning (DL) framework. Results show that the proposed unrolled DL reconstruction enables increasing bout 2x SNR compared to SNR obtained by online reconstructed images. Moreover, its speed is about 200x faster than conventional locally low rank constraint reconstruction.

Qingjia Bao¹, Liyang Xia², Kewen Liu², Xinjie Liu¹, Peng Sun³, Lucio Frydman⁴, and Chaoyang Liu^1
1Innovation Academy for Precision Measurement Science and Technology, Wuhan, China, ²Wuhan University of Technology, Wuhan, China, ³Philips Healthcare, Beijing, China, ⁴Weizmann Institute of Science, Rehovot, Israel

Keywords: Image Reconstruction, Susceptibility

Although a major advantage of SPEN vs EPI is a higher immunity to artifacts, it suffers from Nyquist or motion artifacts. We proposed a new unsupervised CNN model that takes advantage of both physical model and Deep learning. The model consists of three parts:  phase feature extraction module, which can extract the phase features of even/odd phase differences or motion-caused phase differences in multi-shot echo data. Then, the phase maps are generated with these phase difference features. Lastly, the phase correction modules to remove artifacts. The results show that the proposed model can effectively correct Nyquist/motion artifacts in single-shot/multi-shot SPEN.

View Presentation Video

Sangtae Ahn¹, Chitresh Bhushan¹, John Huston², J. Kevin DeMarco³, Robert Y. Shih^3,4, Joshua D. Trzasko², Rafi Brada⁵, Graeme Mckinnon⁶, Isabelle Heukensfeldt Jansen¹, Dan Rettmann⁷, Brian Burns⁸, Ty A. Cashen⁶, Nir Mazor⁵, Xucheng Zhu⁸, and Thomas K. Foo^1
1GE Research, Niskayuna, NY, United States, ²Mayo Clinic College of Medicine, Rochester, MN, United States, ³Walter Reed National Military Medical Center, Bethesda, MD, United States, ⁴Uniformed Services University of the Health Sciences, Bethesda, MD, United States, ⁵GE Research, Herzliya, Israel, ⁶GE HealthCare, Waukesha, WI, United States, ⁷GE HealthCare, Rochester, MN, United States, ⁸GE HealthCare, Menlo Park, CA, United States

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, Brain, Neuro

3D MRI enables thin slices at the cost of long scan times, causing practical challenges. Recently, deep-learning (DL) techniques have successfully accelerated MR scans. However, it is challenging to characterize the image quality (IQ) performance of DL methods by conventional metrics because IQ depends on applications, i.e., how images are used. We evaluate the IQ performance of DL-Speed, our DL-based acceleration method, for 3D T1-weighted MPRAGE brain scans, in 1) post-reconstruction subcortical structure segmentation, and 2) a reader study. The results imply DL-Speed can accelerate scans with reduction factor R=10 while maintaining IQ comparable to standard parallel imaging with R=2.1.

View Presentation Video

Xiaxia Wu ¹, Weiyin Vivian Liu², and Yunfei Zha^1
1Renmin Hospital of Wuhan University, Wuhan, China, ²GE Healthcare,MR Reaearch China,Beijing, Wuhan, China

Keywords: Image Reconstruction, Cartilage

Diagnostic performance was limited to image resolution and contrast between target tissues and surrounding tissues. A rapid knee imaging has been perused but no loss of image quality is critical. This study proposed a rapid knee imaging based on two-dimensional fast spin echo sequence and examined the reliability and diagnostic performance of deep learning-based reconstruction T1-, T2- and PD- weighted images on knee joint pathology via comparison of images with and without deep-learning reconstruction algorithm (DLR). Diagnostic efficacy on knee structural abnormalities of 2D DLR FSE sequence elevated using knee arthroscopy results as the gold standard.

View Presentation Video

Zijian Zhou^1,2, Haikun Qi^1,2, and Peng Hu^1,2
1School of Biomedical Engineering, ShanghaiTech University, Shanghai, China, ²Shanghai Clinical Research and Trial Center, Shanghai, China

Keywords: Image Reconstruction, Motion Correction

Motion-mitigated reconstruction of highly undersampled MRI was achieved by adding a motion estimation module to the data consistency part of the model-based unfolded variation network. The motion estimation module consisted of a pair of convolutional blocks with residual inputs and added only limited number of trainable parameters to the network. The network was trained and tested on synthesized motion-corrupted images from a publicly available knee dataset. The reconstructed images with the proposed motion estimation module were sharper, and details were better recovered, with the structural similarity and peak signal-to-noise ratio significantly improved.

View Presentation Video

Yulia Shcherbakova¹, Tijl A van der Velden^1,2, and Peter R Seevinck^1,2
1Imaging Division, UMC Utrecht, Utrecht, Netherlands, ²MRIguidance B.V., Utrecht, Netherlands

Keywords: Image Reconstruction, Skeletal, AI, acceleration, CS-SENSE, spine, sCT, bone

In this work, we investigated the performance of an AI based algorithm - synthetic CT generation - when subjected to compressed sensing-sensitivity encoding (CS-SENSE) accelerated gradient echo images. We performed MR experiments in five volunteers, using different CS-SENSE acceleration factors for the MR acquisitions. Our results showed that using CS-SENSE factors of 1.45 and 2 increased noise in the MR source images but did not compromise the sCT reconstruction on visual inspection, which was confirmed by quantitative metrics. However, CS-SENSE with a factor of 3 caused artifacts in the sCT images which may affect the safety and diagnostic performance of the product.

View Presentation Video

Peng Li¹, Xiaodi Li¹, Xin Lu², and Yue Hu^1
1Harbin Institute of Technology, Harbin, China, ²De Montfort University, Leicester, United Kingdom

Keywords: Image Reconstruction, MR Fingerprinting, Tensor Low-rank, CP Decomposition, Bloch Response Manifold, non-Cartesian

We propose a deep unrolled network for non-Cartesian MRF reconstruction by unrolling the MRF reconstruction model regularized by the tensor low-rank and the Bloch resonance manifold priors. To avoid computationally burdensome singular value decomposition, we propose a learned CP decomposition module to exploit the tensor low-rank priors of MRF data. Inspired by the MRF imaging mechanism, we also propose a Bloch response manifold module to learn the mapping between reconstructed MRF data and the multiple parameter maps. Numerical experiments show that the proposed network can improve the reconstruction quality of MRF data and multi-parameter maps within significantly reduced computational time.

Yuting Shi¹, Yuyao Zhang², and Hongjiang Wei^1
1Shanghai Jiao Tong University, Shanghai, China, ²ShanghaiTech University, Shanghai, China

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence

Diffusion tensor imaging (DTI) needs a large number of diffusion-weighted images (DWIs) to reliably reconstruct the diffusion measurements of the brain white matter, making the data acquisition time-consuming. Deep learning has emerged as a powerful technique to reduce the number of acquired DWIs. While most existing deep learning methods are supervised and need high-quality ground truth data as the training labels. Here, we proposed an unsupervised and subject-specific DTI reconstruction method called DTI-Net to significantly reduce the required number of DWIs, while also can simultaneously conduct the super-resolution reconstruction of the tensors.

Zejun Wu¹, Jiechao Wang¹, Zunquan Chen¹, Zhigang Wu², Jianfeng Bao³, Shuhui Cai¹, and Congbo Cai^1
1Department of Electronic Science, Xiamen University, Xiamen, China, ²Philips Healthcare, Shenzhen, China, ³the First Affiliated Hospital of Zhengzhou University, Zhengzhou, China

Keywords: Image Reconstruction, Diffusion Tensor Imaging

Deep learning has been used in diffusion tensor imaging (DTI) to fast reconstruct diffusion parameters. However, diffusion-weighted images (DWIs) as network input must maintain diffusion gradient direction consistency during training and testing for deep-learning-based DTI parameter mapping. A dynamic-convolution-based network was developed to achieve generalized DTI parameter mapping for flexible diffusion gradient directions. This proposed method uses dynamic convolution kernels to embed diffusion gradient direction information into feature maps of the corresponding diffusion signal. The results indicate that the proposed method can reconstruct high-quality DTI-derived maps from six diffusion gradient directions.

View Presentation Video

Muhammad Atif¹, Madiha Arshad¹, Yumna Bilal¹, Omair Inam¹, Hassan Shahzad², and Hammad Omer^1
1Medical Image Processing Research Group (MIPRG), Department of Electrical and Computer Engineering, COMSATS University, Islamabad, Pakistan, ²National Centre of Physics (NCP), Islamabad, Pakistan

Keywords: Image Reconstruction, Image Reconstruction

Self-Calibrating GROG (SC-GROG) is a gridding algorithm that maps the k-space MRI data from non-Cartesian to Cartesian domain. The main limitation of SC-GROG is its computational cost to calculate the GROG weights. This paper proposes a customized deep learning framework (based on VGG-16 CNN model) to calculate the 2D-Gridding weight sets for SC-GROG. Initially, the proposed model is trained on human head images, and later fine-tuning is performed using Golden-angle radial Liver Perfusion datasets. The results show that the proposed method significantly reduces the computation time for the estimation of GROG weights while maintaining the image quality.

Caohui Duan¹, Xiangbing Bian¹, Kun Cheng¹, Xiaoyu Wang¹, Jinhao Lyu¹, Xueyang Wang¹, Jianxun Qu², Xin Zhou³, and Xin Lou^1
1Department of Radiology, Chinese PLA General Hospital, Beijing, China, ²MR Collaboration, Siemens Healthineers Ltd., Beijing, China, ³Key Laboratory of Magnetic Resonance in Biological Systems, State Key Laboratory of Magnetic Resonance and Atomic and Molecular Physics, National Center for Magnetic Resonance in Wuhan, Wuhan Institute of Physics and Mathematics, Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences‒Wuhan National Laboratory for Optoelectronics, Wuhan, China

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence

Ultra-high field 7T susceptibility-weighted imaging (SWI) has shown great potential in visualizing and evaluating a broad range of pathology, but suffers from long acquisition times. In this study, a complex-valued convolutional neural network (ComplexNet) model was proposed to reconstruct highly accelerated 7T SWI data. The average reconstruction time of ComplexNet was 0.56 seconds per slice (45.16 seconds per participant). Meanwhile, ComplexNet can provide high-quality 7T SWI for visualizing subtle pathology, including cerebral microbleeds, asymmetric deep medullary veins, and swallow tail sign.

View Presentation Video

Zhijun Wang¹, Huajun She¹, and Yiping P. Du^1
1School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence

Dynamic MRI shows promising clinical values and several applications have been investigated such as cardiac, pulmonary, and hepatic imaging. However, a successful application of dynamic MRI is hampered by its time-consuming acquisition. To improve the performance and interpretability for the accelerating reconstruction methods, we proposed the parallel non-Cartesian Spatial-Temporal Dictionary Learning Neural Networks (stDLNN), which combines the traditional spatial-temporal dictionary learning methods with the deep neural networks for accelerating dynamic MRI. It has favorable interpretability and provides better image quality than the state-of-the-art CS methods (L+S, BCS) and deep learning methods (DCCNN, PNCRNN), especially at high acceleration rate at R=25. 

View Presentation Video

Yinghao Zhang¹, Xiaodi Li¹, Peng Li¹, and Yue Hu^1
1School of Electronics and Information Engineering, Harbin Institute of Technology, Harbin, China, China

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence

Tensor low-rank models have recently emerged as powerful alternative representations for three-dimensional dynamic MR datasets. In this paper, we introduce a novel learned transform-based tensor low-rank network for dynamic MRI based on the tensor singular value decomposition (t-SVD). Instead of manually designing the t-SVD-based transform, we propose to utilize CNN to adaptively learn the relatively optimal transformation from the dynamic MR dataset for more robust and accurate tensor low-rank representations. Experimental results on cardiac cine MRI reconstruction demonstrate the superior performance of the proposed framework compared with the state-of-the-art methods.

View Presentation Video

Thomas Yu^1,2,3, Francesco La Rosa⁴, Gian Franco Piredda^1,5,6, Jonadab Dos Santos Silva⁴, Faye Bourie⁴, Henry Dieckhaus⁷, Govind Nair⁷, Patrick Liebig⁸, Jean-Philippe Thiran^2,3,5, Tobias Kober^1,2,3, Erin Beck^4,7, and Tom Hilbert^1,2,3
1Advanced Clinical Imaging Technology, Siemens Healthineers International AG, Lausanne, Switzerland, ²Signal Processing Laboratory 5 (LTS5), Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, ³Department of Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, ⁴Department of Neurology, Icahn School of Medicine at Mount Sinai, New York, NY, United States, ⁵Centre d’Imagerie Biomédicale (CIBM), EPFL, Lausanne, Switzerland, ⁶Human Neuroscience Platform, Fondation Campus Biotech Geneva, Geneva, Switzerland, ⁷National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, United States, ⁸Siemens Healthineers International AG, Erlangen, Germany

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, Ultra-High Field MRI

High-resolution 3D MR imaging is necessary for the detailed assessment of focal pathologies, such as cortical lesions. However, high-resolution demands tradeoffs with acceleration and SNR, which is difficult to address with standard machine learning reconstructions due to the infeasibility of collecting large datasets of fully sampled data. Using a dataset of high-resolution (0.5mm isotropic), 3D, 7T MP2RAGE scans of multiple sclerosis patients, we show that a self-supervised reconstruction from one scan, requiring no fully sampled data, has higher apparent SNR than a median of three scans, currently used for assessment, with comparable tissue contrast and lesion conspicuity.