Χ-sepnet(chi-sepnet): Susceptibility source separation using deep neural network
Poster No:
1893
Submission Type:
Abstract Submission
Authors:
Minjun Kim1, Hyeong-Geol Shin2, Sooyeon Ji1, Chungseok Oh3, Jiye Kim4, Jinhee Jang5, Berkin Bilgic6, Jongho Lee1
Institutions:
1Seoul National University, Seoul, Seoul, 2Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, 3Department of Electrical Computer Engineering, Seoul National University, Seoul, Korea, Republic of, 4Seoul National University, Gwanak-gu, Korea, Republic of, 5Seoul St Mary's Hospital, Seoul, Korea, Republic of, 6Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA
First Author:
Co-Author(s):
Hyeong-Geol Shin
Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine
Baltimore, MD
Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine
Baltimore, MD
Chungseok Oh
Department of Electrical Computer Engineering, Seoul National University
Seoul, Korea, Republic of
Department of Electrical Computer Engineering, Seoul National University
Seoul, Korea, Republic of
Berkin Bilgic
Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital
Charlestown, MA
Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital
Charlestown, MA
Introduction:
Recently, χ-separation has been proposed to separate positive and negative susceptibility source distributions1,2,3. These methods should resolve the ill-posed dipole-inversion problem, which creates streaking artifacts. Hence, the results of χ-separation hamper accurate estimation of susceptibility concentrations. The artifacts can be resolved using multiple head orientation datasets as demonstrated in COSMOS QSM4 or a neural network that is trained using COSMOS dataset as shown in QSMnet5. The latter is advantageous because one can infer a COSMOS-quality QSM map from single orientational data once the network is fully trained. In this study, we developed a neural network, χ-sepnet, that generates COSMOS-quality χ-separation results by training the network using multi-orientation χ-separation data. Additionally, another χ-sepnet that only inputs multi-echo GRE data is designed to test the feasibility of using dataset without T2 maps.
Methods:
Multi-echo GRE and multi-echo SE data were acquired from six subjects (see REF.1 for scan parameters; 4:1:1 subjects for train:validation:test; IRB-approved). Each subject GRE dataset consisted of six head orientations. These multi-orientation GRE data and multi-echo SE data were processed using a COSMOS-like χ-separation algorithm that created artifact-free positive (χpos) and negative (χneg) susceptibility maps (χ-sep-COSMOS). These maps were utilized as labels to train the two networks (χ-sepnet-R2' that requires R2 and χ-sepnet-R2* that is free from R2). The original χ-separation algorithm that utilized the MEDI prior is also utilized (χ-sepnet-MEDI).
The networks had same 3D U-net6 structure as QSMnet. For the input, not only local field and R2' (or R2*) but also a QSM map generated from QSMnet was concatenated. The loss function had three losses: L1 loss, gradient loss, and model loss which enforces to learn information from local field maps and R2' maps.
Data augmentation was performed by rotating the images by random angles relative to B0 direction, resulting in a total of 48 training datasets.
For quantitative evaluation, the test subject results were compared for NRMSE, PSNR, and SSIM with the χ-sep-COSMOS results as reference. An ROI analysis performed in 11 ROIs (Fig. 4), reporting the mean and standard deviation of the susceptibility values across the head orientation.
Two additional subjects, one multiple sclerosis (MS) patient and the other subject with calcification (IRB-approved) were inferenced for the evaluation of the networks in clinical practice.
The networks had same 3D U-net6 structure as QSMnet. For the input, not only local field and R2' (or R2*) but also a QSM map generated from QSMnet was concatenated. The loss function had three losses: L1 loss, gradient loss, and model loss which enforces to learn information from local field maps and R2' maps.
Data augmentation was performed by rotating the images by random angles relative to B0 direction, resulting in a total of 48 training datasets.
For quantitative evaluation, the test subject results were compared for NRMSE, PSNR, and SSIM with the χ-sep-COSMOS results as reference. An ROI analysis performed in 11 ROIs (Fig. 4), reporting the mean and standard deviation of the susceptibility values across the head orientation.
Two additional subjects, one multiple sclerosis (MS) patient and the other subject with calcification (IRB-approved) were inferenced for the evaluation of the networks in clinical practice.
Results:
The positive and the negative susceptibility maps from all four methods are summarized in Fig. 2(a). The χpos and χneg maps from the two χ-sepnets show high-quality results that are comparable or even cleaner than the reference maps from χ-sep-COSMOS. Streaking artifacts are observed in the χ-sep-MEDI maps but are less noticeable in the other maps. Quantitative metrics report χ-sepnet-R2' generates the best results while χ-sepnet-R2* provides good outcomes (Fig. 2(b)). The three zoomed-in regions reveal well-known structures of the brain in Fig. 2(c,d,e). Deep gray matter regions are clearly delineated in the χpos and χneg maps. In particular, the hand knob region reveals that the positive susceptibility maps report higher values in the motor cortex than in the sensory cortex, which is well-known iron distribution from histology7.
Conclusions:
In this study, χ-sepnets are designed to reconstruct artifact-free susceptibility source distribution maps using deep learning. The methods only require single head orientation data. Furthermore, χ-sepnet-R2* requires only multi-echo GRE data, reducing the requirement of R2 maps in χ-separation at the cost of slightly decreased accuracy. When applied to the patients, the results reveal expected characteristics of lesions.
Modeling and Analysis Methods:
Methods Development 1
Novel Imaging Acquisition Methods:
Imaging Methods Other 2
Keywords:
Demyelinating
Machine Learning
MRI
Myelin
Structures
1|2Indicates the priority used for review
·Figure 1. (a) Overview of the four χ-separation methods applied (χ-sep-COSMOS and χ-sep-MEDI) or developed (χ-sepnet-R2’ and χ-sep net-R2*) in this study. (b) Processing pipeline of χ-sepnet.
·Figure 2. (a) Reconstruction results of the four χ-separation methods. (b) Quantitative results. (c,d,e) Zoom-in maps of the χ-separation methods at deep gray matter, brainstem, hand knob regions.
Provide references using author date format
2. Chen, J., Gong, N. J., Chaim, K. T., Otaduy, M. C. G., & Liu, C. (2021). Decompose quantitative susceptibility mapping (QSM) to sub-voxel diamagnetic and paramagnetic components based on gradient-echo MRI data. NeuroImage, 242, 118477.
3. Emmerich, J., Bachert, P., Ladd, M. E., & Straub, S. (2021). On the separation of susceptibility sources in quantitative susceptibility mapping: Theory and phantom validation with an in vivo application to multiple sclerosis lesions of different age. Journal of Magnetic Resonance, 330, 107033.
4. Liu, T., Spincemaille, P., De Rochefort, L., Kressler, B., & Wang, Y. (2009). Calculation of susceptibility through multiple orientation sampling (COSMOS): a method for conditioning the inverse problem from measured magnetic field map to susceptibility source image in MRI. Magnetic Resonance in Medicine: An Official Journal of the International Society for Magnetic Resonance in Medicine, 61(1), 196-204.
5. Yoon, J., Gong, E., Chatnuntawech, I., Bilgic, B., Lee, J., Jung, W., ... & Lee, J. (2018). Quantitative susceptibility mapping using deep neural network: QSMnet. Neuroimage, 179, 199-206.
6. Ronneberger, O., Fischer, P., & Brox, T. (2015, October). U-net: Convolutional networks for biomedical image segmentation. In International Conference on Medical image computing and computer-assisted intervention (pp. 234-241). Springer, Cham.
7. Deistung, A., Schäfer, A., Schweser, F., Biedermann, U., Turner, R., & Reichenbach, J. R. (2013). Toward in vivo histology: a comparison of quantitative susceptibility mapping (QSM) with magnitude-, phase-, and R2⁎-imaging at ultra-high magnetic field strength. Neuroimage, 65, 299-314.