1
|
Si W, Guo Y, Zhang Q, Zhang J, Wang Y, Feng Y. Quantitative susceptibility mapping using multi-channel convolutional neural networks with dipole-adaptive multi-frequency inputs. Front Neurosci 2023; 17:1165446. [PMID: 37383103 PMCID: PMC10293650 DOI: 10.3389/fnins.2023.1165446] [Citation(s) in RCA: 1] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Key Words] [Grants] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/14/2023] [Accepted: 05/17/2023] [Indexed: 06/30/2023] Open
Abstract
Quantitative susceptibility mapping (QSM) quantifies the distribution of magnetic susceptibility and shows great potential in assessing tissue contents such as iron, myelin, and calcium in numerous brain diseases. The accuracy of QSM reconstruction was challenged by an ill-posed field-to-susceptibility inversion problem, which is related to the impaired information near the zero-frequency response of the dipole kernel. Recently, deep learning methods demonstrated great capability in improving the accuracy and efficiency of QSM reconstruction. However, the construction of neural networks in most deep learning-based QSM methods did not take the intrinsic nature of the dipole kernel into account. In this study, we propose a dipole kernel-adaptive multi-channel convolutional neural network (DIAM-CNN) method for the dipole inversion problem in QSM. DIAM-CNN first divided the original tissue field into high-fidelity and low-fidelity components by thresholding the dipole kernel in the frequency domain, and it then inputs the two components as additional channels into a multichannel 3D Unet. QSM maps from the calculation of susceptibility through multiple orientation sampling (COSMOS) were used as training labels and evaluation reference. DIAM-CNN was compared with two conventional model-based methods [morphology enabled dipole inversion (MEDI) and improved sparse linear equation and least squares (iLSQR) and one deep learning method (QSMnet)]. High-frequency error norm (HFEN), peak signal-to-noise-ratio (PSNR), normalized root mean squared error (NRMSE), and the structural similarity index (SSIM) were reported for quantitative comparisons. Experiments on healthy volunteers demonstrated that the DIAM-CNN results had superior image quality to those of the MEDI, iLSQR, or QSMnet results. Experiments on data with simulated hemorrhagic lesions demonstrated that DIAM-CNN produced fewer shadow artifacts around the bleeding lesion than the compared methods. This study demonstrates that the incorporation of dipole-related knowledge into the network construction has a potential to improve deep learning-based QSM reconstruction.
Collapse
Affiliation(s)
- Wenbin Si
- School of Biomedical Engineering, Southern Medical University, Guangzhou, China
- Guangdong Provincial Key Laboratory of Medical Image Processing and Guangdong Province Engineering Laboratory for Medical Imaging and Diagnostic Technology, Southern Medical University, Guangzhou, China
| | - Yihao Guo
- Department of Radiology, Hainan General Hospital (Hainan Affiliated Hospital of Hainan Medical University), Haikou, Hainan, China
| | - Qianqian Zhang
- School of Biomedical Engineering, Southern Medical University, Guangzhou, China
- Guangdong Provincial Key Laboratory of Medical Image Processing and Guangdong Province Engineering Laboratory for Medical Imaging and Diagnostic Technology, Southern Medical University, Guangzhou, China
| | - Jinwei Zhang
- Department of Biomedical Engineering, College of Engineering, Cornell University, Ithaca, NY, United States
- Department of Radiology, Weill Cornell Medicine, Cornell University, New York, NY, United States
| | - Yi Wang
- Department of Biomedical Engineering, College of Engineering, Cornell University, Ithaca, NY, United States
- Department of Radiology, Weill Cornell Medicine, Cornell University, New York, NY, United States
| | - Yanqiu Feng
- School of Biomedical Engineering, Southern Medical University, Guangzhou, China
- Guangdong Provincial Key Laboratory of Medical Image Processing and Guangdong Province Engineering Laboratory for Medical Imaging and Diagnostic Technology, Southern Medical University, Guangzhou, China
- Guangdong-Hong Kong-Macao Greater Bay Area Center for Brain Science and Brain-Inspired Intelligence and Key Laboratory of Mental Health of the Ministry of Education, Southern Medical University, Guangzhou, China
| |
Collapse
|
2
|
Dimov AV, Li J, Nguyen TD, Roberts AG, Spincemaille P, Straub S, Zun Z, Prince MR, Wang Y. QSM Throughout the Body. J Magn Reson Imaging 2023; 57:1621-1640. [PMID: 36748806 PMCID: PMC10192074 DOI: 10.1002/jmri.28624] [Citation(s) in RCA: 3] [Impact Index Per Article: 3.0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 12/15/2022] [Revised: 01/19/2023] [Accepted: 01/20/2023] [Indexed: 02/08/2023] Open
Abstract
Magnetic materials in tissue, such as iron, calcium, or collagen, can be studied using quantitative susceptibility mapping (QSM). To date, QSM has been overwhelmingly applied in the brain, but is increasingly utilized outside the brain. QSM relies on the effect of tissue magnetic susceptibility sources on the MR signal phase obtained with gradient echo sequence. However, in the body, the chemical shift of fat present within the region of interest contributes to the MR signal phase as well. Therefore, correcting for the chemical shift effect by means of water-fat separation is essential for body QSM. By employing techniques to compensate for cardiac and respiratory motion artifacts, body QSM has been applied to study liver iron and fibrosis, heart chamber blood and placenta oxygenation, myocardial hemorrhage, atherosclerotic plaque, cartilage, bone, prostate, breast calcification, and kidney stone.
Collapse
Affiliation(s)
- Alexey V. Dimov
- Department of Radiology, Weill Cornell Medicine, New York, NY, United States
| | - Jiahao Li
- Department of Radiology, Weill Cornell Medicine, New York, NY, United States
| | - Thanh D. Nguyen
- Department of Radiology, Weill Cornell Medicine, New York, NY, United States
| | | | - Pascal Spincemaille
- Department of Radiology, Weill Cornell Medicine, New York, NY, United States
| | - Sina Straub
- Department of Radiology, Mayo Clinic, Jacksonville, FL, United States
| | - Zungho Zun
- Department of Radiology, Weill Cornell Medicine, New York, NY, United States
| | - Martin R. Prince
- Department of Radiology, Weill Cornell Medicine, New York, NY, United States
| | - Yi Wang
- Department of Radiology, Weill Cornell Medicine, New York, NY, United States
| |
Collapse
|
3
|
Silva J, Milovic C, Lambert M, Montalba C, Arrieta C, Irarrazaval P, Uribe S, Tejos C. Toward a realistic in silico abdominal phantom for QSM. Magn Reson Med 2023; 89:2402-2418. [PMID: 36695213 PMCID: PMC10952412 DOI: 10.1002/mrm.29597] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/09/2022] [Revised: 12/18/2022] [Accepted: 01/06/2023] [Indexed: 01/26/2023]
Abstract
PURPOSE QSM outside the brain has recently gained interest, particularly in the abdominal region. However, the absence of reliable ground truths makes difficult to assess reconstruction algorithms, whose quality is already compromised by additional signal contributions from fat, gases, and different kinds of motion. This work presents a realistic in silico phantom for the development, evaluation and comparison of abdominal QSM reconstruction algorithms. METHODS Synthetic susceptibility andR 2 * $$ {R}_2^{\ast } $$ maps were generated by segmenting and postprocessing the abdominal 3T MRI data from a healthy volunteer. Susceptibility andR 2 * $$ {R}_2^{\ast } $$ values in different tissues/organs were assigned according to literature and experimental values and were also provided with realistic textures. The signal was simulated using as input the synthetic QSM andR 2 * $$ {R}_2^{\ast } $$ maps and fat contributions. Three susceptibility scenarios and two acquisition protocols were simulated to compare different reconstruction algorithms. RESULTS QSM reconstructions show that the phantom allows to identify the main strengths and limitations of the acquisition approaches and reconstruction algorithms, such as in-phase acquisitions, water-fat separation methods, and QSM dipole inversion algorithms. CONCLUSION The phantom showed its potential as a ground truth to evaluate and compare reconstruction pipelines and algorithms. The publicly available source code, designed in a modular framework, allows users to easily modify the susceptibility,R 2 * $$ {R}_2^{\ast } $$ and TEs, and thus creates different abdominal scenarios.
Collapse
Affiliation(s)
- Javier Silva
- Department of Electrical EngineeringPontificia Universidad Católica de Chile
SantiagoChile
- Biomedical Imaging CenterPontificia Universidad Católica de ChileSantiagoChile
- Millennium Institute for Intelligent Healthcare Engineering (iHEALTH)SantiagoChile
| | - Carlos Milovic
- Millennium Institute for Intelligent Healthcare Engineering (iHEALTH)SantiagoChile
- School of Electrical EngineeringPontificia Universidad Católica de ValparaísoValparaísoChile
| | - Mathias Lambert
- Department of Electrical EngineeringPontificia Universidad Católica de Chile
SantiagoChile
- Biomedical Imaging CenterPontificia Universidad Católica de ChileSantiagoChile
- Millennium Institute for Intelligent Healthcare Engineering (iHEALTH)SantiagoChile
| | - Cristian Montalba
- Biomedical Imaging CenterPontificia Universidad Católica de ChileSantiagoChile
- Millennium Institute for Intelligent Healthcare Engineering (iHEALTH)SantiagoChile
- Department of Radiology, School of MedicinePontificia Universidad Católica de ChileSantiagoChile
| | - Cristóbal Arrieta
- Biomedical Imaging CenterPontificia Universidad Católica de ChileSantiagoChile
- Millennium Institute for Intelligent Healthcare Engineering (iHEALTH)SantiagoChile
| | - Pablo Irarrazaval
- Department of Electrical EngineeringPontificia Universidad Católica de Chile
SantiagoChile
- Biomedical Imaging CenterPontificia Universidad Católica de ChileSantiagoChile
- Millennium Institute for Intelligent Healthcare Engineering (iHEALTH)SantiagoChile
- Institute for Biological and Medical Engineering, Pontificia Universidad Católica de ChileSantiagoChile
| | - Sergio Uribe
- Biomedical Imaging CenterPontificia Universidad Católica de ChileSantiagoChile
- Millennium Institute for Intelligent Healthcare Engineering (iHEALTH)SantiagoChile
- Department of Radiology, School of MedicinePontificia Universidad Católica de ChileSantiagoChile
| | - Cristian Tejos
- Department of Electrical EngineeringPontificia Universidad Católica de Chile
SantiagoChile
- Biomedical Imaging CenterPontificia Universidad Católica de ChileSantiagoChile
- Millennium Institute for Intelligent Healthcare Engineering (iHEALTH)SantiagoChile
| |
Collapse
|
4
|
Boehm C, Schlaeger S, Meineke J, Weiss K, Makowski MR, Karampinos DC. On the water-fat in-phase assumption for quantitative susceptibility mapping. Magn Reson Med 2023; 89:1068-1082. [PMID: 36321543 DOI: 10.1002/mrm.29516] [Citation(s) in RCA: 1] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/02/2022] [Revised: 10/06/2022] [Accepted: 10/15/2022] [Indexed: 11/07/2022]
Abstract
PURPOSE To (a) define multi-peak fat model-based effective in-phase echo times for quantitative susceptibility mapping (QSM) in water-fat regions, (b) analyze the relationship between fat fraction, field map quantification bias and susceptibility bias, and (c) evaluate the susceptibility mapping performance of the proposed effective in-phase echoes in comparison to single-peak in-phase echoes and water-fat separation for regions where both water and fat are present. METHODS Effective multipeak in-phase echo times for a bone marrow and a liver fat spectral model were derived from a single voxel simulation. A Monte Carlo simulation was performed to assess the field map estimation error as a function of fat fraction for the different in-phase echoes. Additionally, a phantom scan and in vivo scans in the liver, spine, and breast were performed and evaluated with respect to quantification accuracy. RESULTS The use of single-peak in-phase echoes can introduce a worst-case susceptibility bias of 0.43 $$ 0.43 $$ ppm. The use of effective multipeak in-phase echoes shows a similar quantitative performance in the numerical simulation, the phantom and in all in vivo anatomies when compared to water-fat separation-based QSM. CONCLUSION QSM based on the proposed effective multipeak in-phase echoes can alleviate the quantification bias present in QSM based on single-peak in-phase echoes. When compared to water-fat separation-based QSM the proposed effective in-phase echo times achieve a similar quantitative performance while drastically reducing the computational expense for field map estimation.
Collapse
Affiliation(s)
- Christof Boehm
- Department of Diagnostic and Interventional Radiology, School of Medicine, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany
| | - Sarah Schlaeger
- Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany
| | | | | | - Marcus R Makowski
- Department of Diagnostic and Interventional Radiology, School of Medicine, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany
| | - Dimitrios C Karampinos
- Department of Diagnostic and Interventional Radiology, School of Medicine, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany
| |
Collapse
|
5
|
Sun C, Ghassaban K, Song J, Chen Y, Zhang C, Qu F, Zhu J, Wang G, Haacke EM. Quantifying calcium changes in the fetal spine using quantitative susceptibility mapping as extracted from STAGE imaging. Eur Radiol 2023; 33:606-614. [PMID: 36044065 PMCID: PMC10662431 DOI: 10.1007/s00330-022-09042-5] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 03/18/2022] [Revised: 06/11/2022] [Accepted: 07/19/2022] [Indexed: 11/25/2022]
Abstract
OBJECTIVES To evaluate calcium deposition in the fetal spine in vivo during the second and third trimesters using quantitative susceptibility mapping (QSM). METHODS Fifty-four pregnant women in their second and third trimesters underwent a 2D multi-echo STrategically Acquired Gradient Echo (STAGE) MR imaging protocol at 3T covering the fetal spine. The first echo data was used for QSM processing. A linear regression model was used to assess the correlation between magnetic susceptibility and gestational age (GA). A paired sample t-test was used to compare the consistency of QSM measurements from each sequence. RESULTS The magnetic susceptibility of the fetal spine decreased linearly with advancing GA, with a slope of -52.3 parts per billion (ppb)/week and a Pearson correlation coefficient (r) of 0.83 (p < 0.001). In 37 subjects for whom the STAGE local QSM data were available from both flip angles, the average magnetic susceptibility values were -1111 ± 278 ppb and -1081 ± 262 ppb for FA = 8° and FA = 40°, respectively. These means were not statistically different according to a paired sample t-test (p = 0.156). CONCLUSIONS QSM is a reliable technique for evaluating calcium deposition and bone mineral density of fetal vertebrae. Our results demonstrate an increase in fetal calcium levels as a function of GA. These measures might be able to provide reference values for calcium content in the fetal spine during the second and third trimesters. KEY POINTS • Calcium deposition and mineralization in the fetal spine, evaluated by vertebral magnetic susceptibility, increased with advancing gestational age. • Our results provide reference values for calcium content in the fetal spine during the second and third trimesters.
Collapse
Affiliation(s)
- Cong Sun
- Department of Radiology, Shandong Provincial Hospital, Shandong University, Jinan, China
- Department of Radiology, Beijing Hospital, National Center of Gerontology, Institute of Geriatric Medicine, Chinese Academy of Medical Sciences, Beijing, China
| | - Kiarash Ghassaban
- Department of Radiology, Wayne State University, Detroit, MI, USA
- SpinTech MRI Inc., Bingham Farms, MI, USA
| | - Jiaguang Song
- Department of Ultrasound, Shandong Provincial Hospital Affiliated to Shandong First Medical University, Jinan, China
| | - Yufan Chen
- Department of Radiology, Shandong Provincial Hospital, Shandong University, Jinan, China
| | - Chao Zhang
- Department of Radiology, Shandong Provincial Hospital, Shandong University, Jinan, China
| | - Feifei Qu
- MR Collaboration, Siemens Healthineers Ltd., Shanghai, China
| | - Jinxia Zhu
- MR Collaboration, Siemens Healthineers Ltd., Beijing, China
| | - Guangbin Wang
- Department of Radiology, Shandong Provincial Hospital Affiliated to Shandong First Medical University, 324, Jingwu Road, Jinan, 250021, Shandong, China.
| | - E Mark Haacke
- Department of Radiology, Wayne State University, Detroit, MI, USA.
- SpinTech MRI Inc., Bingham Farms, MI, USA.
| |
Collapse
|
6
|
Bachrata B, Trattnig S, Robinson SD. Quantitative susceptibility mapping of the head-and-neck using SMURF fat-water imaging with chemical shift and relaxation rate corrections. Magn Reson Med 2022; 87:1461-1479. [PMID: 34850446 PMCID: PMC7612304 DOI: 10.1002/mrm.29069] [Citation(s) in RCA: 2] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/03/2021] [Revised: 09/23/2021] [Accepted: 10/15/2021] [Indexed: 12/19/2022]
Abstract
PURPOSE To address the challenges posed by fat-water chemical shift artifacts and relaxation rate discrepancies to quantitative susceptibility mapping (QSM) outside the brain, and to generate accurate susceptibility maps of the head-and-neck at 3 and 7 Tesla. METHODS Simultaneous Multiple Resonance Frequency (SMURF) imaging was extended to 7 Tesla and used to acquire head-and-neck gradient echo images at both 3 and 7 Tesla. Separated fat and water images were corrected for Type 1 (displacement) and Type 2 (phase discrepancy) chemical shift artefacts, and for the bias resulting from differences in T1 and T 2 ∗ relaxation rates, recombined and used as the basis for QSM. A novel phase signal-based masking approach was used to generate head-and-neck masks. RESULTS SMURF generated well-separated fat and water images of the head-and-neck. Corrections for chemical shift artefacts and relaxation rate differences removed overestimation of the susceptibility values, blurring in the susceptibility maps, and the disproportionate influence of fat in mixed voxels. The resulting susceptibility maps showed high correspondence between the paramagnetic areas and the locations of fatty tissues and the susceptibility estimates were similar to literature values. The proposed masking approach was shown to provide a simple means of generating head-and-neck masks. CONCLUSION Corrections for Type 1 and Type 2 chemical shift artefacts and for fat-water relaxation rate differences, mainly in T1 , were shown to be required for accurate susceptibility mapping of fatty-body regions. SMURF made it possible to apply these corrections and generate high-quality susceptibility maps of the entire head-and-neck at both 3 and 7 Tesla.
Collapse
Affiliation(s)
- Beata Bachrata
- High Field MR Centre, Department of Biomedical Imaging and Image-Guided Therapy, Medical University of Vienna, Vienna, Austria
- Karl Landsteiner Institute for Clinical Molecular MR in Musculoskeletal Imaging, Vienna, Austria
| | - Siegfried Trattnig
- High Field MR Centre, Department of Biomedical Imaging and Image-Guided Therapy, Medical University of Vienna, Vienna, Austria
- Karl Landsteiner Institute for Clinical Molecular MR in Musculoskeletal Imaging, Vienna, Austria
| | - Simon Daniel Robinson
- High Field MR Centre, Department of Biomedical Imaging and Image-Guided Therapy, Medical University of Vienna, Vienna, Austria
- Karl Landsteiner Institute for Clinical Molecular MR in Musculoskeletal Imaging, Vienna, Austria
- Centre of Advanced Imaging, University of Queensland, Brisbane, Australia
- Department of Neurology, Medical University of Graz, Graz, Austria
| |
Collapse
|
7
|
Cai Z, Tao Q, Scotti A, Yi P, Feng Y, Cai K. Early detection of increased marrow adiposity with age in rats using Z-spectral MRI at ultra-high field (7 T). NMR IN BIOMEDICINE 2022; 35:e4633. [PMID: 34658086 DOI: 10.1002/nbm.4633] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Subscribe] [Scholar Register] [Received: 02/19/2021] [Revised: 09/16/2021] [Accepted: 09/18/2021] [Indexed: 06/13/2023]
Abstract
BACKGROUND Nowadays, the drive towards high-field MRI is fueled by the pursuit of higher signal-to-noise ratio, spatial resolution, and imaging speed. However, high field strength is associated with field inhomogeneity, acceleration of T2 * decay, and increased chemical shift, which may pose challenges to conventional MRI for fat quantification in complex tissues such as bone marrow. With proton MRI spectroscopy (1 H-MRS), on the other hand, it is difficult to produce high resolution. As a novel alternative fat quantification method, high-resolution Z-spectral MRI (ZS-MRI) can achieve fat quantification by acquiring direct saturated images of both fat and water under the same TE , which may be less affected by T2 * decay and field inhomogeneity. PURPOSE To demonstrate ZS-MRI for marrow adipose tissue (MAT) quantification in rat's lumbar spine and the early detection of MAT changes with age. METHODS The accuracy of ZS-MRI for fat quantification at ultra-high-field MRI (7 T) was verified with MRS and conventional Dixon MRI in water-oil mixed phantoms with varying fat fraction (FF). Dixon MRI data were processed with iterative decomposition of water and fat with echo asymmetry and least-squares estimation. ZS-MRI was then used to longitudinally monitor the adiposity in the lumbar spine of young healthy rats at 13, 17, and 21 weeks to detect the early changes of FF with age in MAT. Hematoxylin-eosin staining of lumbar spines from separated rat groups was performed for verification. RESULTS In ex vivo phantom experiments, both Dixon MRI and ZS-MRI were well correlated with 1 H-MRS for the quantification of FF at 7 T (R > 0.99). Compared with Dixon MRI, ZS-MRI showed reduced image artifacts due to field inhomogeneity and presented better agreement with 1 H-MRS for the early detection of increased MAT due to age at 7 T (ZS-MRI R = 0.78 versus Dixon MRI R = 0.34). The increased MAT FF due to age was confirmed by histology. CONCLUSION ZS-MRI proves itself as an alternative fat quantification method for bone marrow in rats at 7 T.
Collapse
Affiliation(s)
- Zimeng Cai
- School of Biomedical Engineering, Southern Medical University, Guangzhou, Guangdong, China
- Guangdong Provincial Key Laboratory of Medical Image Processing, Southern Medical University, Guangzhou, Guangdong, China
| | - Quan Tao
- School of Biomedical Engineering, Southern Medical University, Guangzhou, Guangdong, China
- Guangdong Provincial Key Laboratory of Medical Image Processing, Southern Medical University, Guangzhou, Guangdong, China
| | - Alessandro Scotti
- Department of Radiology, College of Medicine, University of Illinois at Chicago, Chicago, Illinois, USA
- Department of Bioengineering, University of Illinois at Chicago, Chicago, Illinois, USA
| | - Peiwei Yi
- School of Biomedical Engineering, Southern Medical University, Guangzhou, Guangdong, China
- Guangdong Provincial Key Laboratory of Medical Image Processing, Southern Medical University, Guangzhou, Guangdong, China
| | - Yanqiu Feng
- School of Biomedical Engineering, Southern Medical University, Guangzhou, Guangdong, China
- Guangdong Provincial Key Laboratory of Medical Image Processing, Southern Medical University, Guangzhou, Guangdong, China
| | - Kejia Cai
- Department of Radiology, College of Medicine, University of Illinois at Chicago, Chicago, Illinois, USA
- Department of Bioengineering, University of Illinois at Chicago, Chicago, Illinois, USA
| |
Collapse
|
8
|
Jafari R, Spincemaille P, Zhang J, Nguyen TD, Luo X, Cho J, Margolis D, Prince MR, Wang Y. Deep neural network for water/fat separation: Supervised training, unsupervised training, and no training. Magn Reson Med 2021; 85:2263-2277. [PMID: 33107127 PMCID: PMC7809709 DOI: 10.1002/mrm.28546] [Citation(s) in RCA: 21] [Impact Index Per Article: 7.0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/16/2020] [Revised: 09/12/2020] [Accepted: 09/14/2020] [Indexed: 12/22/2022]
Abstract
PURPOSE To use a deep neural network (DNN) for solving the optimization problem of water/fat separation and to compare supervised and unsupervised training. METHODS The current T 2 ∗ -IDEAL algorithm for solving water/fat separation is dependent on initialization. Recently, DNN has been proposed to solve water/fat separation without the need for suitable initialization. However, this approach requires supervised training of DNN using the reference water/fat separation images. Here we propose 2 novel DNN water/fat separation methods: 1) unsupervised training of DNN (UTD) using the physical forward problem as the cost function during training, and 2) no training of DNN using physical cost and backpropagation to directly reconstruct a single dataset. The supervised training of DNN, unsupervised training of DNN, and no training of DNN methods were compared with the reference T 2 ∗ -IDEAL. RESULTS All DNN methods generated consistent water/fat separation results that agreed well with T 2 ∗ -IDEAL under proper initialization. CONCLUSION The water/fat separation problem can be solved using unsupervised deep neural networks.
Collapse
Affiliation(s)
- Ramin Jafari
- Meinig School of Biomedical Engineering, Cornell University, Ithaca, NY
- Department of Radiology, Weill Cornell Medicine, New York, NY
| | | | - Jinwei Zhang
- Meinig School of Biomedical Engineering, Cornell University, Ithaca, NY
- Department of Radiology, Weill Cornell Medicine, New York, NY
| | - Thanh D. Nguyen
- Department of Radiology, Weill Cornell Medicine, New York, NY
| | - Xianfu Luo
- Department of Radiology, Weill Cornell Medicine, New York, NY
- Department of Radiology, Northern Jiangsu People's Hospital, Yangzhou, China
| | - Junghun Cho
- Meinig School of Biomedical Engineering, Cornell University, Ithaca, NY
- Department of Radiology, Weill Cornell Medicine, New York, NY
| | - Daniel Margolis
- Department of Radiology, Weill Cornell Medicine, New York, NY
| | | | - Yi Wang
- Meinig School of Biomedical Engineering, Cornell University, Ithaca, NY
- Department of Radiology, Weill Cornell Medicine, New York, NY
| |
Collapse
|