IMPULSED model based cytological feature estimation with U-Net: Application to human brain tumor at 3T
Jian Wu
Department of Electronic Science, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, China
Search for more papers by this authorTaishan Kang
Department of Radiology, Zhongshan Hospital of Xiamen University, School of Medicine, Xiamen University, Xiamen, China
Search for more papers by this authorXinli Lan
Department of Electronic Science, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, China
Search for more papers by this authorXinran Chen
Department of Electronic Science, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, China
Search for more papers by this authorZhigang Wu
MSC Clinical & Technical Solutions, Philips Healthcare, Beijing, China
Search for more papers by this authorJiazheng Wang
MSC Clinical & Technical Solutions, Philips Healthcare, Beijing, China
Search for more papers by this authorLiangjie Lin
MSC Clinical & Technical Solutions, Philips Healthcare, Beijing, China
Search for more papers by this authorCongbo Cai
Department of Electronic Science, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, China
Search for more papers by this authorJianzhong Lin
Department of Radiology, Zhongshan Hospital of Xiamen University, School of Medicine, Xiamen University, Xiamen, China
Search for more papers by this authorXin Ding
Department of Pathology, Zhongshan Hospital of Xiamen University, School of Medicine, Xiamen University, Xiamen, China
Search for more papers by this authorCorresponding Author
Shuhui Cai
Department of Electronic Science, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, China
Correspondence
Shuhui Cai, Department of Electronic Science, Xiamen University, Xiamen 361005, China.
Email: [email protected]
Search for more papers by this authorJian Wu
Department of Electronic Science, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, China
Search for more papers by this authorTaishan Kang
Department of Radiology, Zhongshan Hospital of Xiamen University, School of Medicine, Xiamen University, Xiamen, China
Search for more papers by this authorXinli Lan
Department of Electronic Science, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, China
Search for more papers by this authorXinran Chen
Department of Electronic Science, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, China
Search for more papers by this authorZhigang Wu
MSC Clinical & Technical Solutions, Philips Healthcare, Beijing, China
Search for more papers by this authorJiazheng Wang
MSC Clinical & Technical Solutions, Philips Healthcare, Beijing, China
Search for more papers by this authorLiangjie Lin
MSC Clinical & Technical Solutions, Philips Healthcare, Beijing, China
Search for more papers by this authorCongbo Cai
Department of Electronic Science, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, China
Search for more papers by this authorJianzhong Lin
Department of Radiology, Zhongshan Hospital of Xiamen University, School of Medicine, Xiamen University, Xiamen, China
Search for more papers by this authorXin Ding
Department of Pathology, Zhongshan Hospital of Xiamen University, School of Medicine, Xiamen University, Xiamen, China
Search for more papers by this authorCorresponding Author
Shuhui Cai
Department of Electronic Science, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, China
Correspondence
Shuhui Cai, Department of Electronic Science, Xiamen University, Xiamen 361005, China.
Email: [email protected]
Search for more papers by this authorJian Wu and Taishan Kang contributed equally to this work.
Funding information: National Natural Science Foundation of China, Grant/Award Numbers: 11775184; 82071913; 82102021; U1805261; Science and Technology Project of Fujian Province of China, Grant/Award Number: 2019Y0001
Click here for author-reader discussions
Abstract
Purpose
This work introduces and validates a deep-learning-based fitting method, which can rapidly provide accurate and robust estimation of cytological features of brain tumor based on the IMPULSED (imaging microstructural parameters using limited spectrally edited diffusion) model fitting with diffusion-weighted MRI data.
Methods
The U-Net was applied to rapidly quantify extracellular diffusion coefficient (Dex), cell size (d), and intracellular volume fraction (vin) of brain tumor. At the training stage, the image-based training data, synthesized by randomizing quantifiable microstructural parameters within specific ranges, was used to train U-Net. At the test stage, the pre-trained U-Net was applied to estimate the microstructural parameters from simulated data and the in vivo data acquired on patients at 3T. The U-Net was compared with conventional non-linear least-squares (NLLS) fitting in simulations in terms of estimation accuracy and precision.
Results
Our results confirm that the proposed method yields better fidelity in simulations and is more robust to noise than the NLLS fitting. For in vivo data, the U-Net yields obvious quality improvement in parameter maps, and the estimations of all parameters are in good agreement with the NLLS fitting. Moreover, our method is several orders of magnitude faster than the NLLS fitting (from about 5 min to <1 s).
Conclusion
The image-based training scheme proposed herein helps to improve the quality of the estimated parameters. Our deep-learning-based fitting method can estimate the cell microstructural parameters fast and accurately.
CONFLICT OF INTEREST
Zhigang Wu, Jiazheng Wang, and Liangjie Lin are employed by Philips Healthcare China. All other authors declare no competing financial interests.
Open Research
DATA AVAILABILITY STATEMENT
The code used in this study is available at https://github.com/wjgxw/ogse. All data necessary for this study are available from the corresponding author upon reasonable request.
Supporting Information
| Filename | Description |
|---|---|
| mrm29429-sup-0001-Supinfo.pdfPDF document, 2.1 MB | FIGURE S1. Simulated results of the influence of different choice of Din value on d value estimation when SNR ≈ 25 dB. The green solid lines represent zero error. The top and bottom of the boxes are the 25th and 75th percentiles of the datasets, and dots are outliers. From left to right and top to bottom (A − I), Din = 0.6, 0.9, 1.2, 1.5, 1.8, 2.1, 2.4, 2.7, 3.0 μm2/ms. FIGURE S2. Fitting results when the SD of Rician noise is 0. The green solid line represents the zero error. The black short lines represent the error between the target and the fitted values. FIGURE S3. The flowchart of training data generation. (A) An example of the synthetic TDS. (B, C) Intermediate DW images during the generation of training data. Different columns correspond to different b values. (D) A representative sample used in network training. The b values (s/mm2) are labeled at the top of each DW image. FIGURE S4. Training error and validation error curves of (A) Dex, (B) d, and (C) vin. FIGURE S5. Variation of SNR with signal intensity when SD of Rician noise is 6.6 × 10−3. Figure S6. The results obtained with inappropriate parameter range. (A) The SD of Rician noise of training dataset was 3.3 × 10−3. The green solid lines represent zero error. The top and bottom of the boxes are the 25th and 75th percentiles of the datasets, and dots are outliers. (B) Dex map obtained from NLLS fitting. (C) Dex map obtained from a U-Net trained with a dataset whose parameters were uniformly sampled in the following ranges: d: 0–35 μm, vin: 0%–100%, and Dex: 0–3 μm2/ms. (D) Dex map obtained from a U-Net trained with a dataset whose parameters were uniformly sampled in the following ranges: d: 0–50 μm, vin: 0%–100%, and Dex: 0–3 μm2/ms. The SDs of Rician noise of both datasets were 3.3 × 10−3. FIGURE S7. (A) Diagrammatic representation of the FCN used for comparison. (B) The parametric maps estimated with U-Net, NLLS fitting, and FCN. FIGURE S8. (A) Comparison of the estimated parametric maps obtained from U-Net and DenseNet. (B) Comparison of the estimated parametric maps when the U-Net was trained with different loss functions. (C) Comparison of the estimated parametric maps when the training data were cropped to different sizes. (D) Comparison of the estimated parametric maps when the U-Net was trained with different batch sizes. TABLE S1. Information of brain tumor patients. |
Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.
REFERENCES
- 1Jiang X, Li H, Zhao P, et al. Early detection of treatment-induced mitotic arrest using temporal diffusion magnetic resonance spectroscopy. Neoplasia. 2016; 18: 387-397.
- 2Jiang X, Mckinley ET, Xie J, Li H, Gore J. In vivo magnetic resonance imaging of treatment-induced apoptosis. Sci Rep. 2019; 9: 9540.
- 3Wu D, Martin LJ, Northington FJ, Zhang J. Oscillating-gradient diffusion magnetic resonance imaging detects acute subcellular structural changes in the mouse forebrain after neonatal hypoxia-ischemia. J Cereb Blood Flow Metab. 2019; 39: 1336-1348.
- 4Reynaud O. Time-dependent diffusion MRI in cancer: tissue modeling and applications. Front Phys. 2017; 5: 58.
- 5Fukutomi H, Glasser MF, Zhang H, et al. Neurite imaging reveals microstructural variations in human cerebral cortical gray matter. NeuroImage. 2018; 182: 488-499.
- 6Palombo M, Ianus A, Guerreri M, et al. SANDI: a compartment-based model for non-invasive apparent soma and neurite imaging by diffusion MRI. NeuroImage. 2020; 215:116835.
- 7Panagiotaki E, Walker-Samuel S, Siow B, et al. Noninvasive quantification of solid tumor microstructure using VERDICT MRI. Cancer Res. 2014; 74: 1902-1912.
- 8Panagiotaki E, Chan RW, Dikaios N, et al. Microstructural characterization of normal and malignant human prostate tissue with vascular, extracellular, and restricted diffusion for cytometry in tumours magnetic resonance imaging. Investig Radiol. 2015; 50: 218-227.
- 9Xu J, Jiang X, Li H, et al. Magnetic resonance imaging of mean cell size in human breast tumors. Magn Reson Med. 2020; 83: 2002-2014.
- 10Jiang X, Li H, Xie J, Zhao P, Gore JC, Xu J. Quantification of cell size using temporal diffusion spectroscopy. Magn Reson Med. 2016; 75: 1076-1085.
- 11Jiang X, Li H, Devan SP, Gore JC, Xu J. MR cell size imaging with temporal diffusion spectroscopy. Magn Reson Imaging. 2020; 77: 109-123.
- 12Xu J. Probing neural tissues at small scales: recent progress of oscillating gradient spin echo (OGSE) neuroimaging in humans. J Neurosci Methods. 2020; 349:109024.
- 13Reynaud O, Winters KV, Hoang DM, Wadghiri YZ, Novikov DS, Kim SG. Pulsed and oscillating gradient MRI for assessment of cell size and extracellular space (POMACE) in mouse gliomas. NMR Biomed. 2016; 29: 1350-1363.
- 14Zaccagna F, Riemer F, Priest AN, et al. Non-invasive assessment of glioma microstructure using VERDICT MRI: correlation with histology. Eur Radiol. 2019; 29: 5559-5566.
- 15Xu J, Jiang X, Devan SP, et al. MRI-cytometry: mapping nonparametric cell size distributions using diffusion MRI. Magn Reson Med. 2021; 85: 748-761.
- 16Jiang X, Xu J, Gore JC. Mapping hepatocyte size in vivo using temporal diffusion spectroscopy MRI. Magn Reson Med. 2020; 84: 2671-2683.
- 17Wu D, Jiang K, Li H, et al. Time-dependent diffusion MRI for quantitative microstructural mapping of prostate cancer. Radiology. 2022; 303: 578-587.
- 18El-Fiki M, El-Henawy Y, Abdel-Rahman N. Cystic meningioma. Acta Neurochir. 1996; 138: 811-817.
- 19Van Cauter S, Veraart J, Sijbers J, et al. Gliomas: diffusion kurtosis MR imaging in grading. Radiology. 2012; 263: 492-501.
- 20Lin X, DeAngelis LM. Treatment of brain metastases. J Clin Oncol. 2015; 33: 3475-3484.
- 21Schachter M, Does M, Anderson A, Gore J. Measurements of restricted diffusion using an oscillating gradient spin-echo sequence. J Magn Reson Imaging. 2000; 147: 232-237.
- 22Bertleff M, Domsch S, Weingärtner S, et al. Diffusion parameter mapping with the combined intravoxel incoherent motion and kurtosis model using artificial neural networks at 3T. NMR Biomed. 2017; 30:e3833.
- 23Barbieri S, Gurney-Champion OJ, Klaassen R, Thoeny HC. Deep learning how to fit an intravoxel incoherent motion model to diffusion-weighted MRI. Magn Reson Med. 2020; 83: 312-321.
- 24Kaandorp MP, Barbieri S, Klaassen R, et al. Improved unsupervised physics-informed deep learning for intravoxel incoherent motion modeling and evaluation in pancreatic cancer patients. Magn Reson Med. 2021; 86: 2250-2265.
- 25Chen L, Schär M, Chan KW, et al. In vivo imaging of phosphocreatine with artificial neural networks. Nat Commun. 2020; 11: 1072.
- 26Perlman O, Ito H, Herz K, et al. Quantitative imaging of apoptosis following oncolytic virotherapy by magnetic resonance fingerprinting aided by deep learning. Nat Biomed Eng. 2021; 6: 648-657.
- 27Glang F, Deshmane A, Prokudin S, et al. DeepCEST 3T: robust MRI parameter determination and uncertainty quantification with neural networks—application to CEST imaging of the human brain at 3T. Magn Reson Med. 2020; 84: 450-466.
- 28Gyori NG, Clark CA, Alexander DC, Kaden E. On the potential for mapping apparent neural soma density via a clinically viable diffusion MRI protocol. NeuroImage. 2021; 239:118303.
- 29Gyori NG, Palombo M, Clark C, Zhang H, Alexander D. Training data distribution significantly impacts the estimation of tissue microstructure with machine learning. Magn Reson Med. 2022; 87: 932-947.
- 30Hill I, Palombo M, Santin M, et al. Machine learning based white matter models with permeability: an experimental study in cuprizone treated in-vivo mouse model of axonal demyelination. NeuroImage. 2021; 224:117425.
- 31Golkov V, Dosovitskiy A, Sperl JI, et al. Q-space deep learning: twelve-fold shorter and model-free diffusion MRI scans. IEEE Trans Med Imaging. 2016; 35: 1344-1351.
- 32Chandra SS, Bran Lorenzana M, Liu X, Liu S, Bollmann S, Crozier S. Deep learning in magnetic resonance image reconstruction. J Med Imaging Radiat Oncol. 2021; 65: 564-577.
- 33Park G, Hong J, Duffy BA, Lee J-M, Kim H. White matter hyperintensities segmentation using the ensemble U-net with multi-scale highlighting foregrounds. NeuroImage. 2021; 237:118140.
- 34Kofler A, Dewey M, Schaeffter T, Wald C, Kolbitsch C. Spatio-temporal deep learning-based undersampling artefact reduction for 2D radial cine MRI with limited training data. IEEE Trans Med Imaging. 2019; 39: 703-717.
- 35Hu Y, Xu Y, Tian Q, et al. RUN-UP: accelerated multishot diffusion-weighted MRI reconstruction using an unrolled network with U-net as priors. Magn Reson Med. 2021; 85: 709-720.
- 36Liu C, Liang C, Liu Z, Zhang S, Huang B. Intravoxel incoherent motion (IVIM) in evaluation of breast lesions: comparison with conventional DWI. Eur J Radiol. 2013; 82: 782-789.
- 37Jiang X, Li H, Xie J, et al. In vivo imaging of cancer cell size and cellularity using temporal diffusion spectroscopy. Magn Reson Med. 2017; 78: 156-164.
- 38Tanner JE. Transient diffusion in a system partitioned by permeable barriers. Application to NMR measurements with a pulsed field gradient. J Chem Phys. 1978; 69: 1748-1754.
- 39Nilsson M, Ltt J, Westen DV, et al. Noninvasive mapping of water diffusional exchange in the human brain using filter-exchange imaging. Magn Reson Med. 2013; 69: 1572-1580.
- 40Li H, Jiang X, Xie J, Gore JC, Xu J. Impact of transcytolemmal water exchange on estimates of tissue microstructural properties derived from diffusion MRI. Magn Reson Med. 2017; 77: 2239-2249.
- 41Schad LR, Brix G, Zuna I, Harle W, Lorenz WJ, Semmler W. Multiexponential proton spin-spin relaxation in MR imaging of human brain tumors. J Comput Assist Tomogr. 1989; 13: 577-587.
- 42Li K, Li H, Zhang XY, et al. Influence of water compartmentation and heterogeneous relaxation on quantitative magnetization transfer imaging in rodent brain tumors. Magn Reson Med. 2016; 76: 635-644.
- 43Dortch RD, Yankeelov TE, Yue Z, Quarles CC, Does M. Evidence of multiexponential T2 in rat glioblastoma. NMR Biomed. 2009; 22: 609-618.
- 44Aggarwal M, Jones MV, Calabresi PA, Mori S, Zhang J. Probing mouse brain microstructure using oscillating gradient diffusion MRI. Magn Reson Med. 2012; 67: 98-109.
- 45Gore JC, Xu J, Colvin DC, Yankeelov TE, Parsons EC, Does MD. Characterization of tissue structure at varying length scales using temporal diffusion spectroscopy. NMR Biomed. 2010; 23: 745-756.
- 46Parsons EC, Does MD, Gore JC. Temporal diffusion spectroscopy: theory and implementation in restricted systems using oscillating gradients. Magn Reson Med. 2006; 55: 75-84.
- 47Ronneberger O, Fischer P, Brox T. U-net: convolutional networks for biomedical image segmentation. In: N Navab, J Hornegger, W Wells, A Frangi, eds. Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015, Part III, Lecture Notes in Computer Science. Vol 9351; 2015: 234-241.
10.1007/978-3-319-24574-4_28 Google Scholar
- 48Grøvik E, Yi D, Iv M, et al. Handling missing MRI sequences in deep learning segmentation of brain metastases: a multicenter study. NPJ Digit Med. 2021; 4: 33.
- 49Park J, Jung W, Choi EJ, et al. DIFFnet: diffusion parameter mapping network generalized for input diffusion gradient schemes and b-value. IEEE Trans Med Imaging. 2022; 41: 491-499.
- 50Dar SUH, Özbey M, Çatlı AB, Çukur T. A transfer-learning approach for accelerated MRI using deep neural networks. Magn Reson Med. 2020; 84: 663-685.
- 51Yoon J, Gong E, Chatnuntawech I, Bilgic B, Lee J. Quantitative susceptibility mapping using deep neural network: QSMnet. NeuroImage. 2018; 179: 199-206.
- 52Hu Z, Wang Y, Zhang Z, et al. Distortion correction of single-shot EPI enabled by deep-learning. NeuroImage. 2020; 221:117170.
- 53Zhang H, Wang C, Chen W, et al. Deep learning based multiplexed sensitivity-encoding (DL-MUSE) for high-resolution multi-shot DWI. NeuroImage. 2021; 244:118632.
- 54Huang G, Liu Z, Van Der Maaten L, Weinberger KQ. Densely connected convolutional networks. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2017: 2261-2269.
- 55Valsamis JJ, Dubovan PI, Baron CA. Characterization and correction of time-varying eddy currents for diffusion MRI. Magn Reson Med. 2022; 87: 2209-2223.
