| Computer # |
| 4484.
 | 1 | Computationally Efficient IMplicit Training Strategy for UNrolled NEtworks (IMUNNE) for Reconstruction of Accelerated Cardiac Dynamic MRI Nikolay Iakovlev1, Florian Andreas Schiffers2, Lexiaozi Fan1, Santiago Lopez Tapia3, KyungPyo Hong1, Dima Bishara1,4, Jane Wilcox5, Daniel C Lee1,5, Aggelos K Katsaggelos3, and Daniel Kim1,4
1Radiology, Northwestern University, Chicago, IL, United States, 2Computer Science, Northwestern University, Evanston, IL, United States, 3Electrical Engineering, Northwestern University, Evanston, IL, United States, 4Biomedical Engineering, Northwestern University, Evanston, IL, United States, 5Medicine, Cardiology Division, Northwestern University, Chicago, IL, United States Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence, Compressed Sensing, Unrolled Network, Implicit Network Motivation: Unrolled networks (UN) achieve state-of-the-art performance in undersampled dynamic MRI reconstruction but suffer from long training times and extensive GPU memory cost. Goal(s): To apply an implicit training strategy for UNs (IMUNNE) in combination with transfer learning to develop an efficient and versatile reconstruction technique for accelerated dynamic cardiac MRI. Approach: We compare IMUNNE with a complex denoiser U-Net and an end-to-end UN on three different highly undersampled dynamic cardiac MRI datasets. Results: For all datasets, we observed that: (1) both unrolled architectures outperform CU-Net with respect to image quality; (2) compared to end-to-end UN, IMUNNE significantly reduced both training and inference times. Impact: This
work has the potential to facilitate a more widespread adoption of highly-accelerated,
cardiac MRI by reducing training time, inference time and memory cost of
state-of-the-art unrolled reconstruction methods, thereby lowering the clinical
hardware requirements and the requisite energy consumption. |
| 4485.
 | 2 | Improving Image Quality of Dynamic Contrast Enhanced Abdominal MRI Using a Novel Deep Learning Reconstruction Eugene Milshteyn1, Soumyadeep Ghosh2, Nabih Nakrour2, Rory L. Cochran2, Nathaniel Mercaldo2, Xinzeng Wang3, Leo L. Tsai2, Arnaud Guidon1, and Mukesh G. Harisinghani2
1GE HealthCare, Boston, MA, United States, 2Department of Radiology, Massachusetts General Hospital, Boston, MA, United States, 3GE HealthCare, Houston, TX, United States Keywords: AI/ML Image Reconstruction, DSC & DCE Perfusion, DISCO-Star, DL Stack-of-stars Motivation: Free breathing DCE imaging utilizes stack-of-stars sampling, which can lead to streak artifacts and noise reduction when too few spokes are used. Goal(s): Our goal was to validate application of deep learning to 3D DISCO-Star DCE imaging in the abdomen via image quality assessment and noise characterization. Approach: DL and conventionally reconstructed images were assessed by two radiologists across different IQ attributes. Noise characteristics were evaluated by calculation of total variation. AUC was also calculated. Results: The radiologists preferred DL across many of the IQ attributes, with noticeably lowered noise and decreased streaks in DL images. AUC was similar between the two reconstructions. Impact: The application of DL to DISCO-Star DCE imaging provides enhanced diagnostic quality, with reduced streaking, higher SNR, and better in-plane resolution. This has the potential to improve care for abdominal patients who have trouble holding their breath. |
| 4486.
 | 3 | Temporally-Aware Neural Networks For Cine MRI Reconstruction From Severely Undersampled Data Niraj Rajesh Mahajan1, Ana Rodríguez-Soto2, Nuri Chung2, Sanjeet Hegde3, Brent L Gordon3, Amanda Potersnak3, Joni Blood3, and Francisco Contijoch2,3
1Department of Computer Science, University of California San Diego, La Jolla, CA, United States, 2Department of Bioengineering, University of California San Diego, La Jolla, CA, United States, 3Rady's Children Hospital, San Diego, San Diego, CA, United States Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence Motivation: MRI guidance of an interventional procedure requires fast image reconstruction. A neural network(NN)-based approach can exploit the similarities between consecutive frames to improve iMRI image reconstruction. Goal(s): We investigate if an LSTM can reconstruct images from just ten spokes per frame in a timeframe compatible with iMRI. Approach: A convolutional (conv)LSTM was trained using the open-source ACDC dataset. Results were compared with Multi-domain convolutional neural network (MD-CNN) - a recently-published 3D NN-based method for undersampled MRI reconstruction. Results: ConvLSTMs can reconstruct frames at ~226 fps (17x faster than MD-CNN ~13 fps). SSIM for the convLSTM was slightly lower than the MD-CNN (0.85 vs 0.89). Impact: With our LSTM-based model, we have achieved a 17x speed-up in the iMRI acquisition process without significant loss in image quality. This suggests that an LSTM-based method could be used to improve iMRI image speed and quality. |
| 4487.
 | 4 | Movieformer: Motion-resolved 4D MRI reconstruction using a network with spatiotemporal attention Anthony Mekhanik1, Victor Murray1, and Ricardo Otazo1,2
1Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, United States, 2Department of Radiology, Memorial Sloan Kettering Cancer Center, New York, NY, United States Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence Motivation: Investigate utility of self-attention deep learning to exploit global temporal information in motion-resolved 4D MR imaging. Goal(s): Design a novel hybrid convolutional-attention network to reconstruct motion-resolved 4D images without explicit k-space data consistency. Approach: A hybrid Unet-style 4D reconstruction network was developed to incorporate windowed multiscale spatiotemporal multihead self-attention. Training and testing were performed on free-breathing data acquired on patients with abdominal tumors. Results: Spatiotemporal attention successfully captured motion in multiple dimensions with improved image quality relative to state-of-the-art XD-GRASP reconstruction. Impact: Self-attention deep learning mechanism can combine long-range spatial learning and global temporal learning to augment capabilities of convolutional networks for improved motion-resolved 4D MRI of mobile tumors. |
| 4488.
 | 5 | Paired Conditional Generative Adversarial Network for Highly Accelerated Liver 4D MRI Di Xu1, Xin Miao2, Yang Yang3, Hengjie Liu4, Jessica E. Scholey1, Wensha Yang1, Mary Feng1, Michael Ohliger1, Yi Lao4, and Ke Sheng1
1Radiation Oncology, UCSF, San Francisco, CA, United States, 2Siemens Healthineers, Boston, MA, United States, 3Radiology, UCSF, San Francisco, CA, United States, 4Radiation Oncology, UCLA, Los Angeles, CA, United States Keywords: AI/ML Image Reconstruction, Radiotherapy, 4D MRI Motivation: Densely sampled k-space leads to high-quality MR but can be impractical due to lengthy scanning time. Accelerating MR acquisition by reducing sampling density can decrease image quality and/or increase reconstruction complexity and time. Goal(s): This work aims to design an algorithm for efficient and high-quality reconstruction of highly accelerated radial-sampling liver 4D MR. Approach: We proposed a novel Paired Conditional generative adversarial network term Re-Con-GAN, evaluated on a 4D liver MR dataset at 3x, 6x, and 10x acceleration ratios. Results: Re-Con-GAN achieved better PSNR, SSIM, and RMSE with sub-second inference speed (0.15s) than compressed sensing (120s) and non-GAN deep learning methods (0.15s-0.73s). Impact: A robust and efficient framework, Re-Con-GAN, is proposed in the current work with sub-second inference speed (0.15s) and promising reconstruction results demonstrated on an in-house curated 4D liver MRI dataset. |
| 4489.
 | 6 | Self-Supervised Low-rank plus Sparse Network for Radial MRI Reconstruction Andrei Mancu1, Wenqi Huang2, Gastao Lima da Cruz3, Daniel Rückert2,4, and Kerstin Hammernik1
1School of Computation, Information and Technology, Technical University of Munich, München, Germany, 2Klinikum Rechts der Isar, Technical University of Munich, München, Germany, 3Department of Radiology, University of Michigan, Ann Arbor, MI, United States, 4Department of Computing, Imperial College London, London, United Kingdom Keywords: AI/ML Image Reconstruction, Image Reconstruction, Inverse Problems, Deep learning, Low-rank, Cardiac MRI, Radial sampling Motivation: Physics-guided self-supervised approaches have proven to be useful in MR image reconstruction from limited Cartesian measurements. However, the potential of radially-sampled k-space data remains largely unexplored. Goal(s): In this context, we introduce a self-supervised learning approach to reconstruct dynamic images from sparsely-sampled radial cardiac data. Approach: The proposed model integrates a novel low-rank and sparse regularizer in its iterative framework to better exploit the characteristics of dynamic images. Results: Our method is compared to iterative reconstruction techniques and other deep neural network approaches in supervised and self-supervised tasks, where the proposed model achieves the best performance for a single and four heartbeat reconstruction. Impact: Self-supervised models for radially sampled cardiac measurements can now be efficiently trained on limited amounts of data to reliably reconstruct high-contrast and low artifact dynamic MR images, even at high acceleration rates for faster acquisition speed. |
| 4490.
 | 7 | Volumetric Reconstruction Resolves Off-Resonance Artifacts in Static and Dynamic PROPELLER MRI Annesha Ghosh1, Gordon Wetzstein2, Mert Pilanci2, and Sara Fridovich-Keil2
1University of California, Berkeley, Berkeley, CA, United States, 2Stanford University, Stanford, CA, United States Keywords: AI/ML Image Reconstruction, Motion Correction, Off-Resonance Correction, Video Reconstruction Motivation: Off-resonance correction typically requires field map measurements or pretraining data, which are slow/difficult to collect. Goal(s): We propose a physics-based strategy to address off-resonance artifacts using only PROPELLER measurements, by modeling an additional spectral dimension. Approach: This strategy exploits an equivalence between measuring a PROPELLER blade at a certain angle, and viewing a relief sculpture at the same angle. In this equivalence, three-dimensional structures (fat) appear shifted along the blade/view direction relative to flatter structures (water).
Results: Our method resolves continuous chemical shift artifacts while allowing for video reconstruction of dynamic tissues. We provide preliminary results on synthetic static and dynamic data. Impact: We use volumetric reconstruction to correct off-resonance artifacts and perform fat/water separation in PROPELLER MRI, without additional field map measurements or pretraining data. We hope our method opens the door to shorter scan times and higher temporal resolution imaging. |
| 4491.
 | 8 | Deep Learning-Based High Frequency Constrained Fast Image Reconstruction for 4D Cardiac MRI Ashmita Deb1, Danielle Kara1, Mary Robakowski1,2, Ojas Potdar1,3, David Chen1, and Christopher T Nguyen1,4,5,6
1Cardiovascular Innovation Research Center, Heart Vascular Thoracic Institute, Cleveland Clinic, Cleveland, OH, United States, 2Chemical and Biomedical Engineering, Cleveland State University, Cleveland, OH, United States, 3Case Western Reserve University, Cleveland, OH, United States, 4Cardiovascular Medicine, Heart Vascular Thoracic Institute, Cleveland Clinic, Cleveland, OH, United States, 5Imaging Institute, Cleveland Clinic, Cleveland, OH, United States, 6Biomedical Engineering, Case Western Reserve University & Cleveland Clinic, Cleveland, OH, United States Keywords: AI/ML Image Reconstruction, Cardiovascular Motivation: Online image reconstructions for prospectively undersampled cardiac MRI data are noisy as the naïve approach fails to remove undersampling artifacts. Compressed sensing (CS) reconstruction reduces artifacts but is time and memory intensive, making it an offline reconstruction option only. Goal(s): Our aim was to address these constraints by training a Deep Learning (DL) model to obtain high resolution online reconstructions. Approach: We achieved this by implementing a spatiotemporal UNET with a weighted high frequency loss. Results: We found the results of the DL model comparable to the CS reconstruction in image quality, with lower computational cost, making it suitable for online reconstructions.
Impact: Our image denoising Deep Learning (DL) model showed similar results to the time consuming, more computationally expensive compressed sensing (CS) reconstruction (gold standard), thus demonstrating its potential for online reconstruction of prospectively undersampled cardiac MRI data.
|
| 4492.
 | 9 | ML models for 4D cine imaging Mark Wrobel1, Vivek Muthurangu1, Javier Montalt1, and Jennifer Steeden1
1Centre for Translational Cardiovascular Imaging, UCL, London, United Kingdom Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence Motivation: The motivation behind this study is to severely shorten scan times for children undergoing cardiac examination. Goal(s): The goal is to be able to create a 4D dataset from a short, free breathing real-time 2D stack of images. Approach: We apply three machine learning models to the 2D stack. The first reconstructs the undersampled image data, the second corrects respiratory artefacts caused by free-breathing and the third model super resolves the images. Results: The image quality is vastly improved after applying the machine learning models. The ventricular volumes are also in good agreement with the reference volumes. Impact: Severely
reduced scan times for comprehensive cardiac examination in CHD without the use
of breath-holds. Machine Learning methods may be able to also be used for other
imaging sequences, also resulting in faster image acquisition. |
| 4493.
 | 10 | Cine Cardiac MRI Motion Correction using Denoising Diffusion Probabilistic Models Yang Liu1, Jiameng Diao1, Zijian Zhou1, Haikun Qi1,2,3, and Peng Hu1,2,3
1ShanghaiTech University, Shanghai, China, 2Shanghai Clinical Research and Trial Center, Shanghai, China, 3United Imaging Healthcare, Shanghai, China Keywords: AI Diffusion Models, Motion Correction Motivation: Cine cardiac MRI is used to evaluate cardiac functions and vascular abnormalities. However, MRI requires a long scan time, which inevitably induces motion artifacts. Goal(s): Develop a cine cardiac MR image motion correction technique to reduce both the scan time and motion artifacts. Approach: We trained a diffusion-based model with simulated data from a public ACDC dataset to reduce the cine cardiac MRI motion artifacts. Results: The proposed method was compared with GAN and U-Net methods in removing motion artifacts. It produced results that closely approach the ground-truth, achieving the highest SSIM and PSNR scores among all the evaluated methods. Impact: Our method demonstrates improvements in motion compensation compared with GAN and U-Net. |
| 4494.
 | 11 | Rapid LAVA imaging with deep learning reconstruction: evaluation of image quality and diagnostic performance in patients with liver cancer Guo Sa1, Qidong Wang1, Desheng Shang1, Qingqing Wen2, Weiqiang Dou2, Zhan Feng1, and Feng Chen1
1Department of Radiology, First Affiliated Hospital,School of Medicine,ZheJiang University, Hangzhou, China, 2MR Research, GE Healthcare, Beijing, China Keywords: AI/ML Image Reconstruction, Cancer Motivation: 3D gradient-echo based liver acceleration volume acquisition (LAVA) sequence is widely used for dynamic contrast imaging in liver. LAVA usually requires breath-holding for over 16 seconds, posing a challenge for individuals with difficulty in prolonged breath-holding. Goal(s): To investigate whether deep learning reconstruction (DLR) allows for LAVA imaging with reduced scan time but without sacrificing image diagnostic quality. Approach: SNR, CNR, and subjective analysis using 5-point Likert scales were compared to evaluate the image quality and diagnostic performance between DLR-LAVA and conventional LAVA. Results: Compared to conventional LAVA, DLR-LAVA showed similar SNR, CNR, and qualitative image quality scores. Impact: Deep
learning reconstruction based rapid LAVA imaging is promising for reducing
breath-hold time while maintaining similar image quality compared with
conventional LAVA imaging. |
| 4495.
 | 12 | Scale Time-Equivariant Convolutional Neural Networks For Dynamic Magnetic Resonance Imaging Yuliang Zhu1,2, Jing Cheng3, Zhuoxu Cui2, Yulin Wang1, Jie Zeng1, Chengbo Wang1, and Dong Liang3
1Department of Electrical and Electronic Engineering, University of Nottingham, Ningbo, China, Ningbo, China, 2Research Center for Medical AI, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, Shen Zhen, China, 3Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, Shen Zhen, China Keywords: AI/ML Image Reconstruction, Image Reconstruction, cardiac, unrolled, deep neural network, equivariance Motivation: The scale symmetry of anatomical structures commonly exists in dynamic magnetic resonance imaging (MRI) data but have rarely been explored. Goal(s): Our goal is to effectively leverage the scale symmetry of local structures in both spatial and temporal dimensions to improve the reconstrcution quality in dynamic MRI. Approach: We present a novel method that incorporates the scale equivariant convolution modules into an unrolled deep neural network. Results: The proposed method was test on the cardiac cine MRI data reconstruction tasks and achieved the improved performance with a PSNR of 43.6967 and a SSIM of 0.9834. Impact: Our method improved the data-efficiency for deep dynamic MRI reconstructions and robustness to drifts in scale of the images that might stem from the variability of patient anatomies or change in field-of-view across different MRI scanners. |
| 4496.
 | 13 | Curation of Training Data for Supervised Deep Learning Reconstruction of Speech Real-Time MRI Kevin Lee1, Prakash Kumar1, Khalil Iskarous2, and Krishna S. Nayak1
1Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, United States, 2Linguistics, University of Southern California, Los Angeles, CA, United States Keywords: AI/ML Image Reconstruction, Data Processing, Dynamic Imaging Motivation: Supervised deep learning (DL) reconstruction requires large training sets and computationally demanding training. Real-time MRI offers large temporal redundancy which yields high reconstruction performance from training on a subset of frames.
Goal(s): To develop a method for curating small DL training datasets that capture the variance of the entire training set and provide performance non-inferior to the entire training set, with reduced training time. Approach: We use clustering for each training speech task followed by selecting a fraction of each cluster to train U-Nets for reconstruction. Results: We achieve improved image quality metrics with comparable image quality metrics with 10x improved training time. Impact: By using curated training data
based on identification and clustering of vocal tract postures, we demonstrate supervised
DL-reconstruction of speech RT-MRI with 10-fold training time reduction and comparable
NRMSE, PSNR, and SSIM. This may be generalized to other dynamic reconstructions. |