View Presentation Video

Zechen Zhou¹, Ryan Chamberlain¹, Praveen Gulaka¹, Enhao Gong¹, Greg Zaharchuk¹, and Ajit Shankaranarayanan^1
1Subtle Medical Inc, Menlo Park, CA, United States

Keywords: Machine Learning/Artificial Intelligence, Machine Learning/Artificial Intelligence

An unrolled network with explicit MR image degradation modeling and physical property constraints, termed PGDNet, is proposed for adaptive image denoising and deblurring. Preliminary evaluation demonstrated that PGDNet can achieve similar/superior image quality enhancement compared to the conventional task-specific networks, and can outperform others in joint denoising and deblurring tasks. PGDNet provides a promising solution for adaptive MR image denoising and deblurring to restore the image quality of accelerated clinical MR scans.

View Presentation Video

Amaresha Shridhar Konar¹, Jaemin Shin², Ramesh Paudyal¹, Abhay Dave³, Maggie Fung², Suchandrima Banerjee⁴, Vaios Hatzoglou⁵, and Amita Shukla-Dave^1,5
1Medical Physics, Memorial Sloan Kettering Cancer Center, New York City, NY, United States, ²GE Healthcare, New York City, NY, United States, ³Touro College of Osteopathic Medicine, New York, NY, United States, ⁴GE Healthcare, Menlo Park, CA, United States, ⁵Radiology, Memorial Sloan Kettering Cancer Center, New York City, NY, United States

Keywords: Machine Learning/Artificial Intelligence, Quantitative Imaging

MRI has excellent extracranial soft-tissue contrast to detect tumors in the head and neck (HN) region. Technical challenges arise due to MRI related artifacts. In routine radiological practice, HN MR imaging protocols are optimized specifically to the subsites. We aimed to evaluate the performance of the HN imaging protocol that include qualitative T₁w, T₂w, and quantitative diffusion MRI powered by a novel deep learning (DL) based reconstruction (recon) using the ACR and QIBA diffusion phantoms. This phantom study showed that qualitative T₁w and T₂w images and multiple b-value DWI data powered with DL recon substantially improves the image quality.

View Presentation Video

Maximilian Zubke^1,2, Filip Klimeš^1,2, Andreas Voskrebenzev^1,2, Marius Wernz^1,2, Till F Kaireit^1,2, Agilo L Kern^1,2, Marcel Gutberlet^1,2, Robin A Müller^1,2, Frank Wacker^1,2, and Jens Vogel-Claussen^1,2
1Institute of Diagnostic and Interventional Radiology, Hannover Medical School, Hannover, Germany, ²Biomedical Research in Endstage and Obstructive Lung Disease Hannover (BREATH), German Center for Lung Research (DZL), Hannover, Germany

Keywords: Machine Learning/Artificial Intelligence, Lung

3D-phase-resolved-functional-lung (3D-PREFUL) MRI enables a non-contrast-enhanced detection of lung ventilation defects. Convolutional, recurrent neural networks (CRNN) can accelerate data acquisition by reconstructing dynamic lung images from undersampled data. However, a remaining bias in reconstruction led to incorrect detection of ventilation defects. Reducing the number of respiratory phases per reconstruction step improved the accuracy of the defect detection. This improvement is demonstrated by comparison of 2 complementary ventilation defect metrics derived from the original data and the 2x, 4x and 6x undersampled data of Asthma, COPD and post-COVID-19 patients reconstructed from a CRNN in partitions of 30,20 and 10 respiratory phases.  

View Presentation Video

Haoyang Pei^1,2, Ding Xia¹, Fang Liu³, and Li Feng^1
1Biomedical Engineering and Imaging Institute and Department of Radiology, Icahn School of Medicine at Mount Sinai, New York, NY, United States, ²Department of Electrical and Computer Engineering, NYU Tandon School of Engineering, New York, NY, United States, ³Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Harvard Medical School, Boston, MA, United States

Keywords: Machine Learning/Artificial Intelligence, Machine Learning/Artificial Intelligence

Golden-angle RAdial Sparse Parallel (GRASP) MRI has recently been extended for rapid, accurate and robust T1 mapping (GraspT1) that can be performed during free breathing. However, GraspT1 implements a conventional T1 mapping framework that reconstructs an image series from undersampled dynamic k-space in the first step and then performs pixel-wise parameter fitting in the second step. This leads to a slow and cumbersome pipeline to obtain T1 maps. In this work, we developed deep learning-based GraspT1 (DeepGraspT1), which directly estimates T1 maps from undersampled k-space and enables additional acceleration that outperforms conventional iterative reconstruction.

View Presentation Video

Chungseok Oh¹, Hongjun An¹, and Jongho Lee^1
1Seoul National University, Seoul, Korea, Republic of

Keywords: Machine Learning/Artificial Intelligence, Machine Learning/Artificial Intelligence

The performance of a deep neural network can be affected by software and hardware setups when training the network and, therefore, can vary from training to training. This issue of reproducibility, which can be referred to as an “intrinsic” reproducibility of deep learning, can be critical for academic research because reproducibility is a key requirement for journal papers. In this study, we explore this intrinsic reproducibility issue for deep learning-powered parallel imaging reconstruction by using a popular end-to-end variational network. This study may provide minimal requirements for reproducible research in network training.  

View Presentation Video

Haozhong Sun¹, Yuze Li¹, Runyu Yang¹, Zhongsen Li¹, and Huijun Chen^1
1Center for Biomedical Imaging Research, Tsinghua University, Beijing, China

Keywords: Machine Learning/Artificial Intelligence, Machine Learning/Artificial Intelligence

Various image reconstruction methods have been proposed to reduce Magnetic resonance (MR) image acquisition time. One of recent trends is convolution neuron network (CNN) based deep learning model. However, most of these CNN models keep architecture of stacking small filters (e.g. 1×1 or 3×3) and the effective receptive field of these networks is limited, which is undesired for reconstruction because the random undersampling pattern causes global artifact. We proposed Fourier convolution block (FCB) to replace regular convolution filters. FCB can achieve both global receptive field and high computing efficiency by multiplication in frequency domain.

View Presentation Video

Gulfam Ahmed Saju¹, Zhiqiang Li², Reza Abiri³, Tianming Liu⁴, and Yuchou Chang^1
1Computer and Information Science, University of Massachusetts Dartmouth, North Dartmouth, MA, United States, ²Neuroradiology, Barrow Neurological Institute, Phoenix, AZ, United States, ³Electrical, Computer and Biomedical Engineering, University of Rhode Island, Kingston, RI, United States, ⁴Computer Science, University of Georgia, Athens, GA, United States

Keywords: Machine Learning/Artificial Intelligence, Parallel Imaging

JSENSE iteratively optimizes sensitivity maps and the image, so the sensitivity profile can be gradually improved during the reconstruction process. The initially reconstructed image in the first iteration is obtained by the initially estimated coil sensitivity maps. The initial coil sensitivity profiles may be inaccurate and therefore degrade the quality of the subsequent image quality and coil sensitivity map estimation in the iterative optimization process. We propose to use unrolled deep network prior to replace the initial reconstruction in the conventional JSENSE for improving the image reconstruction quality. Experimental results show that the proposed method outperforms CG-SENSE and JSENSE.

View Presentation Video

Gulfam Ahmed Saju¹, Zhiqiang Li², Reza Abiri³, Tianming Liu⁴, and Yuchou Chang^1
1Computer and Information Science, University of Massachusetts Dartmouth, North Dartmouth, MA, United States, ²Neuroradiology, Barrow Neurological Institute, Phoenix, AZ, United States, ³Electrical, Computer and Biomedical Engineering, University of Rhode Island, Kingston, RI, United States, ⁴Computer Science, University of Georgia, Athens, GA, United States

Keywords: Machine Learning/Artificial Intelligence, Motion Correction

Periodically Rotated Overlapping ParallEL Lines with Enhanced Reconstruction (PROPELLER) MRI technique enables the correction of motion artifacts resulted from patient motions in a scanner. Undersampling the blades can increase data acquisition speed and reduce potential motions caused by pains in a short time but may degrade image quality. Deep neural networks may support the blade reconstruction with undersampled data but motion patterns are difficult to be acquired for building a training dataset. To avoid the acquisition of training data, this abstract proposes an untrained neural network-based PROPELLER reconstruction technique to enhance image quality with undersampled blades.

View Presentation Video

Jaa-Yeon Lee¹ and Sung-Hong Park^2
1Korea Advanced Institute of Science & Technology, Daejeon, Korea, Republic of, ²Korea Advanced Institute of Science & Technology, Daejoen, Korea, Republic of

Keywords: Machine Learning/Artificial Intelligence, Machine Learning/Artificial Intelligence, Vision Transformer, Masked Image Modeling

Masked image modeling (MIM) skim has recently been shown to pre-train vision transform (ViT) and works as an effective data augmentation. In this study, we proposed MR reconstruction algorithm using ViT with MIM in k-space. The proposed method showed better performance than the original 1D and 2D ViT. Our study showed that MIM can be used to enhance the reconstruction quality to help learn the data distribution and that combining the loss in both k-space and spatial domain reconstructs images with better perceptuality.

View Presentation Video

Yongquan Ye¹, Zhongqi Zhang¹, Eric Z Chen², Xiao Chen², Shanhui Sun², and Jian Xu^1
1United Imaging, Houston, TX, United States, ²United Imaging Intelligence, Cambridge, MA, United States

Keywords: Machine Learning/Artificial Intelligence, Image Reconstruction

The capacity of the deep learning ReconNet3D model for single-channel image reconstruction with highly under-sampled k-spaces is demonstrated. Without the need for coil sensitivity information, the proposed method can achieve an acceleration factor of 4 on a dual-channel VTC coil. Supporting high-factor acceleration with limited coil channels can be very beneficial for imaging with surface coils or preclinical animal scans.

Weiming Feng¹, Lan Zhu¹, Yihan Xia¹, Kangning Wang¹, Yong Zhang², Jiankun Dai³, Guifeng Fu³, and Huan Zhang^1
1Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China, ²GE Healthcare, Shanghai, China, ³GE Healthcare, Beijing, China

Keywords: Machine Learning/Artificial Intelligence, Image Reconstruction, rectum; magnetic resonance imaging; T2-weighted imaging

High-resolution MRI is of much significance in preoperatively staging rectal cancer. However, the motion artifact from intestinal peristalsis inevitably affects image quality then the accuracy of staging. Deep learning reconstruction (DLRecon) that uses artificial neural networks to extract patterns and makes predictions from large data sets, has been verified in related studies for improving image quality and reducing scanning time. In this study, rectum T2-weighted imaging (T2WI) reconstructed with DLRecon and conventional reconstruction were evaluated, and the results indicate that DLRecon could be employed for better image quality without extra scanning time in clinical practice.

View Presentation Video

Xiangchuang Kong¹, Shengzhen Tao¹, Eric H. Middlebrooks¹, Thomas Benkert², Xiangzhi Zhou¹, and Chen Lin^1
1Department of Radiology, Mayo Clinic, Jacksonville, FL, United States, ²Siemens Medical Solutions USA, Inc., Jacksonville, FL, United States, JACKSONVILLE, FL, United States

Keywords: Machine Learning/Artificial Intelligence, Diffusion/other diffusion imaging techniques, deep learning reconstruction; DWI phantom

In a quantitative phantom study, deep learning (DL) reconstruction is shown to improve the SNR ratio of DWI images while preserving measured ADC values.   The average SNR gain is similar to that achieved with better gradient performance between Siemens Prisma and Vida. Such benefits of DL recon should allow better quality and/or shorter scanning of DWI in clinical applications.

View Presentation Video

Michael Carl¹, Rafi Brada², Nir Mazor², Daniel V Litwiller³, and Maggie Fung^4
1GE Healthcare, San Diego, CA, United States, ²GE Research, Herzliya, Israel, ³GE Healthcare, Denver, CO, United States, ⁴GE Healthcare, New York, NY, United States

Keywords: Machine Learning/Artificial Intelligence, Machine Learning/Artificial Intelligence

In this work we use variable-density prospective undersampling of the phase-encode k-space lines (ky) in 2D fast spin-echo (2DFSE) followed by deep learning (DL) reconstruction. We were able to achieve an acceleration of R=4 while maintaining high image quality.

View Presentation Video

Ruiying Liu¹, Jee Hun Kim², Hongyu Li¹, Peizhou Huang¹, Xiaojuan Li², and Leslie Ying^1
1Department of Biomedical Engineering, Department of Electrical Engineering, University at Buffalo, Buffalo, NY, United States, ²Program of Advanced Musculoskeletal Imaging (PAMI), Cleveland Clinic, Cleveland, OH, United States

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence

Deep learning methods have achieved superior reconstruction results in MRI reconstruction. Recently, denoising diffusion probabilistic models (DDPM) have demonstrated great potential in image-processing tasks. In this study, we combine denoising diffusion probabilistic models (DDPM) and GRAPPA for highly accelerated 3D imaging. The method sequentially performs DDPM and GRAPPA with specially designed sampling masks such that the benefits of the diffusion model and the availability of multi-channel data can be utilized jointly. Our results demonstrate that the proposed method can achieve an acceleration factor of up to 16 which is the product of the factors achieved by DDPM and GRAPPA alone. 

View Presentation Video

Alberto Di Biase^1,2, Claudia Prieto^1,2, and Rene Botnar^1,2
1Department of Electrical Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile, ²Millennium Institute for Intelligent Healthcare Engineering (iHEALTH), Santiago, Chile

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, Compressed sensing

We present a Bayesian optimization approach to find the optimal regularization parameters for undersampled MRI reconstructions such as Compressed Sensing (CS). Bayesian optimization can find optimal points for expensive to evaluate functions by efficient sampling the next points to try by maximizing the expected improvement. Additionally, the use of pruning can speed up optimization by stopping early unpromising trials. We demonstrate the effectiveness of this optimization technique by finding optimal parameters for undersampled MRI using Total Variation and CS-Wavelet regularization. The parameters found with the proposed approach are comparable to does found by grid search but requiring shorter computational times.

View Presentation Video

Laura Pfaff^1,2, Fabian Wagner¹, Julian Hossbach^1,2, Elisabeth Preuhs¹, Dominik Nickel², Tobias Wuerfl², and Andreas Maier^1
1Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany, ²Magnetic Resonance, Siemens Healthcare GmbH, Erlangen, Germany

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence

In MRI, the quantitative evaluation of denoising methods is often limited due to the lack of noise-free ground-truth data. We show how to still approximate the quality metrics mean squared error (MSE) and peak signal-to-noise-ratio (PSNR) without access to ground-truth data by using Stein’s unbiased risk estimator (SURE). The proposed method can be employed to evaluate learning- and non-learning-based denoising approaches, assuming an additive Gaussian noise model with known distribution. Our experiments further reveal that the accuracy of our evaluation method increases with the number of test samples available.

View Presentation Video

Hailong Li¹, Vinicius Vieira Alves¹, Amol Pednekar¹, Mary Kate Manhard¹, Joshua Greer¹, Andrew T. Trout¹, Lili He¹, and Jonathan R. Dillman^1
1Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, deep learning reconstruction

Deep learning (DL)-based techniques are increasingly being applied to assist with or improve image reconstruction. However, the impact of DL-based algorithms on radiomics is not well understood. This study aims to evaluate the impact of two commercially available DL-based reconstruction pipelines: (1) SmartSpeed (Philips Healthcare, U.S. FDA-cleared); and (2) SmartSpeed with Super Resolution (SmartSpeed+SuperRes, not U.S. FDA-cleared to date) on MRI radiomic features. Our analysis showed that compared to conventional image reconstruction technique, 42 out of 86 investigated radiomic features from SmartSpeed images were highly correlated whereas only 13 features from SmartSpeed+SuperRes images had high correlations with conventional image features.

View Presentation Video

Dmitrij Kravchenko^1,2, Alexander Isaak^1,2, Narine Mesropyan^1,2, Claus Christian Pieper¹, Daniel Kuetting^1,2, Leon M. Bischoff^1,2, Shuo Zhang³, Christoph Katemann³, Johannes M. Peeters⁴, Oliver Weber³, Ulrike Attenberger¹, and Julian Luetkens^1,2
1Diagnostic and interventional radiology, University Hospital Bonn, Bonn, Germany, ²Quantitative Imaging Laboratory Bonn, Bonn, Germany, ³Philips GmbH Market DACH, Hamburg, Germany, ⁴Philips MR Clinical Science, Best, Netherlands

Keywords: Machine Learning/Artificial Intelligence, Cardiovascular

AI assisted upscaling of low-resolution cine bSSFP images yields comparable image quality to conventional images with no clinically significant difference in volumetric data at a reduction of acquisition time by a factor of 1.5 to 2.

Keywords: Artificial intelligence, acceleration, Superresolution, cardiac MRI

View Presentation Video

Matt Hemsley^1,2, Liam S.P Lawrence^1,2, and Angus Z Lau^1,2
1Medical Biophysics, University of Toronto, Toronto, ON, Canada, ²Physical Sciences Platform, Sunnybrook Research Institute, Toronto, ON, Canada

Keywords: Machine Learning/Artificial Intelligence, Cancer, Denoising, Self-Supervised

In MR guided radiation therapy, images of patients are acquired daily. However, scan times are long to acquire images with an acceptable SNR for treatment planning and adaptation. In this study we test a self-supervised machine learning based approach for denoising data that can utilize previous scans of the same patient to improve quality of the denoised image. Results on a numerical phantom and clinical images are presented and compared to a popular non-machine learning denoising algorithm. 

Zhuo-Xu Cui¹, Sen Jia², Jing Cheng², Qingyong Zhu¹, and Dong Liang^1,3
1Medical AI Research Center, Shenzhen Institutes of Advanced Technology, Chinese Academy of Science, Shenzhen, China, ²Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Science, Shenzhen, China, ³Pazhou Lab, Guangzhou, China

Keywords: Machine Learning/Artificial Intelligence, Image Reconstruction

This study proposes a safeguarded methodology for network unrolling. Specifically, focusing on parallel MR imaging, we unroll a zeroth-order algorithm, of which the network module represents a regularizer itself so that the network output can still be covered by a regularization model. Furthermore, inspired by the idea of deep equilibrium models, before backpropagation, we carry out the unrolled network to converge to a fixed point and then prove that it can tightly approximate the real MR image. In a case where the measurement data contain noise, we prove that the proposed network is robust against noisy interferences.