Quantitative susceptibility mapping using multi-channel convolutional neural networks with dipole-adaptive multi-frequency inputs
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-susceptibil...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2023-06-01
|
| Series: | Frontiers in Neuroscience |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fnins.2023.1165446/full |
| _version_ | 1797805792743653376 |
|---|---|
| author | Wenbin Si Wenbin Si Yihao Guo Qianqian Zhang Qianqian Zhang Jinwei Zhang Jinwei Zhang Yi Wang Yi Wang Yanqiu Feng Yanqiu Feng Yanqiu Feng |
| author_facet | Wenbin Si Wenbin Si Yihao Guo Qianqian Zhang Qianqian Zhang Jinwei Zhang Jinwei Zhang Yi Wang Yi Wang Yanqiu Feng Yanqiu Feng Yanqiu Feng |
| author_sort | Wenbin Si |
| collection | DOAJ |
| description | 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. |
| first_indexed | 2024-03-13T05:57:26Z |
| format | Article |
| id | doaj.art-e5a361c393c643c1961ee8c15b6ae61a |
| institution | Directory Open Access Journal |
| issn | 1662-453X |
| language | English |
| last_indexed | 2024-03-13T05:57:26Z |
| publishDate | 2023-06-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Neuroscience |
| spelling | doaj.art-e5a361c393c643c1961ee8c15b6ae61a2023-06-13T04:17:04ZengFrontiers Media S.A.Frontiers in Neuroscience1662-453X2023-06-011710.3389/fnins.2023.11654461165446Quantitative susceptibility mapping using multi-channel convolutional neural networks with dipole-adaptive multi-frequency inputsWenbin Si0Wenbin Si1Yihao Guo2Qianqian Zhang3Qianqian Zhang4Jinwei Zhang5Jinwei Zhang6Yi Wang7Yi Wang8Yanqiu Feng9Yanqiu Feng10Yanqiu Feng11School of Biomedical Engineering, Southern Medical University, Guangzhou, ChinaGuangdong Provincial Key Laboratory of Medical Image Processing and Guangdong Province Engineering Laboratory for Medical Imaging and Diagnostic Technology, Southern Medical University, Guangzhou, ChinaDepartment of Radiology, Hainan General Hospital (Hainan Affiliated Hospital of Hainan Medical University), Haikou, Hainan, ChinaSchool of Biomedical Engineering, Southern Medical University, Guangzhou, ChinaGuangdong Provincial Key Laboratory of Medical Image Processing and Guangdong Province Engineering Laboratory for Medical Imaging and Diagnostic Technology, Southern Medical University, Guangzhou, ChinaDepartment of Biomedical Engineering, College of Engineering, Cornell University, Ithaca, NY, United StatesDepartment of Radiology, Weill Cornell Medicine, Cornell University, New York, NY, United StatesDepartment of Biomedical Engineering, College of Engineering, Cornell University, Ithaca, NY, United StatesDepartment of Radiology, Weill Cornell Medicine, Cornell University, New York, NY, United StatesSchool of Biomedical Engineering, Southern Medical University, Guangzhou, ChinaGuangdong Provincial Key Laboratory of Medical Image Processing and Guangdong Province Engineering Laboratory for Medical Imaging and Diagnostic Technology, Southern Medical University, Guangzhou, ChinaGuangdong-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, ChinaQuantitative 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.https://www.frontiersin.org/articles/10.3389/fnins.2023.1165446/fullquantitative susceptibility mappingmagnetic resonance imagingdeep learningconvolutional neural networksimage processing |
| spellingShingle | Wenbin Si Wenbin Si Yihao Guo Qianqian Zhang Qianqian Zhang Jinwei Zhang Jinwei Zhang Yi Wang Yi Wang Yanqiu Feng Yanqiu Feng Yanqiu Feng Quantitative susceptibility mapping using multi-channel convolutional neural networks with dipole-adaptive multi-frequency inputs Frontiers in Neuroscience quantitative susceptibility mapping magnetic resonance imaging deep learning convolutional neural networks image processing |
| title | Quantitative susceptibility mapping using multi-channel convolutional neural networks with dipole-adaptive multi-frequency inputs |
| title_full | Quantitative susceptibility mapping using multi-channel convolutional neural networks with dipole-adaptive multi-frequency inputs |
| title_fullStr | Quantitative susceptibility mapping using multi-channel convolutional neural networks with dipole-adaptive multi-frequency inputs |
| title_full_unstemmed | Quantitative susceptibility mapping using multi-channel convolutional neural networks with dipole-adaptive multi-frequency inputs |
| title_short | Quantitative susceptibility mapping using multi-channel convolutional neural networks with dipole-adaptive multi-frequency inputs |
| title_sort | quantitative susceptibility mapping using multi channel convolutional neural networks with dipole adaptive multi frequency inputs |
| topic | quantitative susceptibility mapping magnetic resonance imaging deep learning convolutional neural networks image processing |
| url | https://www.frontiersin.org/articles/10.3389/fnins.2023.1165446/full |
| work_keys_str_mv | AT wenbinsi quantitativesusceptibilitymappingusingmultichannelconvolutionalneuralnetworkswithdipoleadaptivemultifrequencyinputs AT wenbinsi quantitativesusceptibilitymappingusingmultichannelconvolutionalneuralnetworkswithdipoleadaptivemultifrequencyinputs AT yihaoguo quantitativesusceptibilitymappingusingmultichannelconvolutionalneuralnetworkswithdipoleadaptivemultifrequencyinputs AT qianqianzhang quantitativesusceptibilitymappingusingmultichannelconvolutionalneuralnetworkswithdipoleadaptivemultifrequencyinputs AT qianqianzhang quantitativesusceptibilitymappingusingmultichannelconvolutionalneuralnetworkswithdipoleadaptivemultifrequencyinputs AT jinweizhang quantitativesusceptibilitymappingusingmultichannelconvolutionalneuralnetworkswithdipoleadaptivemultifrequencyinputs AT jinweizhang quantitativesusceptibilitymappingusingmultichannelconvolutionalneuralnetworkswithdipoleadaptivemultifrequencyinputs AT yiwang quantitativesusceptibilitymappingusingmultichannelconvolutionalneuralnetworkswithdipoleadaptivemultifrequencyinputs AT yiwang quantitativesusceptibilitymappingusingmultichannelconvolutionalneuralnetworkswithdipoleadaptivemultifrequencyinputs AT yanqiufeng quantitativesusceptibilitymappingusingmultichannelconvolutionalneuralnetworkswithdipoleadaptivemultifrequencyinputs AT yanqiufeng quantitativesusceptibilitymappingusingmultichannelconvolutionalneuralnetworkswithdipoleadaptivemultifrequencyinputs AT yanqiufeng quantitativesusceptibilitymappingusingmultichannelconvolutionalneuralnetworkswithdipoleadaptivemultifrequencyinputs |