DeepCEST 3T: Robust MRI parameter determination and uncertainty quantification with neural networks-application to CEST imaging of the human brain at 3T.

Published on Jul 1, 2020in Magnetic Resonance in Medicine3.635
· DOI :10.1002/MRM.28117
F Glang3
Estimated H-index: 3
(MPG: Max Planck Society),
Anagha Deshmane12
Estimated H-index: 12
(MPG: Max Planck Society)
+ 6 AuthorsMoritz Zaiss28
Estimated H-index: 28
(MPG: Max Planck Society)
Sources
Abstract
PURPOSE: Calculation of sophisticated MR contrasts often requires complex mathematical modeling. Data evaluation is computationally expensive, vulnerable to artifacts, and often sensitive to fit algorithm parameters. In this work, we investigate whether neural networks can provide not only fast model fitting results, but also a quality metric for the predicted values, so called uncertainty quantification, investigated here in the context of multi-pool Lorentzian fitting of CEST MRI spectra at 3T. METHODS: A deep feed-forward neural network including a probabilistic output layer allowing for uncertainty quantification was set up to take uncorrected CEST-spectra as input and predict 3T Lorentzian parameters of a 4-pool model (water, semisolid MT, amide CEST, NOE CEST), including the B0 inhomogeneity. Networks were trained on data from 3 subjects with and without data augmentation, and applied to untrained data from 1 additional subject and 1 brain tumor patient. Comparison to conventional Lorentzian fitting was performed on different perturbations of input data. RESULTS: The deepCEST 3T networks provided fast and accurate predictions of all Lorentzian parameters and were robust to input perturbations because of noise or B0 artifacts. The uncertainty quantification detected fluctuations in input data by increase of the uncertainty intervals. The method generalized to unseen brain tumor patient CEST data. CONCLUSIONS: The deepCEST 3T neural network provides fast and robust estimation of CEST parameters, enabling online reconstruction of sophisticated CEST contrast images without the typical computational cost. Moreover, the uncertainty quantification indicates if the predictions are trustworthy, enabling confident interpretation of contrast changes.
📖 Papers frequently viewed together
2020
9 Authors (F Glang, ..., Moritz Zaiss)
19 Citations
10 Citations
References38
Newest
#1Tanya Nair (McGill University)H-Index: 4
#2Doina Precup (McGill University)H-Index: 47
Last. Tal Arbel (McGill University)H-Index: 26
view all 4 authors...
Abstract Deep learning networks have recently been shown to outperform other segmentation methods on various public, medical-image challenge datasets, particularly on metrics focused on large pathologies. For diseases such as Multiple Sclerosis (MS), however, monitoring all the focal lesions visible on MRI sequences, even very small ones, is essential for disease staging, prognosis, and evaluating treatment efficacy. Small lesion segmentation presents significant challenges to popular deep learn...
128 CitationsSource
#1Steffen Goerke (DKFZ: German Cancer Research Center)H-Index: 14
#2Yannick Soehngen (DKFZ: German Cancer Research Center)H-Index: 1
Last. Mark E. Ladd (DKFZ: German Cancer Research Center)H-Index: 57
view all 13 authors...
PURPOSE: Relaxation-compensated CEST-MRI (i.e., the inverse metrics magnetization transfer ratio and apparent exchange-dependent relaxation) has already been shown to provide valuable information for brain tumor diagnosis at ultrahigh magnetic field strengths. This study aims at translating the established acquisition protocol at 7 T to a clinically relevant magnetic field strength of 3 T. METHODS: Protein model solutions were analyzed at multiple magnetic field strengths to assess the spectral ...
21 CitationsSource
#1Johannes Breitling (DKFZ: German Cancer Research Center)H-Index: 5
#2Anagha Deshmane (MPG: Max Planck Society)H-Index: 12
Last. Moritz Zaiss (MPG: Max Planck Society)H-Index: 28
view all 9 authors...
: High image signal-to-noise ratio (SNR) is required to reliably detect the inherently small chemical exchange saturation transfer (CEST) effects in vivo. In this study, it was demonstrated that identifying spectral redundancies of CEST data by principal component analysis (PCA) in combination with an appropriate data-driven extraction of relevant information can be used for an effective and robust denoising of CEST spectra. The relationship between the number of relevant principal components an...
18 CitationsSource
#1Moritz Zaiss (MPG: Max Planck Society)H-Index: 28
#2Anagha Deshmane (MPG: Max Planck Society)H-Index: 12
Last. Klaus Scheffler (MPG: Max Planck Society)H-Index: 76
view all 10 authors...
PURPOSE: To determine the feasibility of employing the prior knowledge of well-separated chemical exchange saturation transfer (CEST) signals in the 9.4 T Z-spectrum to separate overlapping CEST signals acquired at 3 T, using a deep learning approach trained with 3 T and 9.4 T CEST spectral data from brains of the same subjects. METHODS: Highly spectrally resolved Z-spectra from the same volunteer were acquired by 3D-snapshot CEST MRI at 3 T and 9.4 T at low saturation power of B1 = 0.6 µT. The ...
6 CitationsSource
#1Anagha Deshmane (MPG: Max Planck Society)H-Index: 12
#2Moritz Zaiss (MPG: Max Planck Society)H-Index: 28
Last. Klaus Scheffler (MPG: Max Planck Society)H-Index: 76
view all 9 authors...
PURPOSE: For clinical implementation, a chemical exchange saturation transfer (CEST) imaging sequence must be fast, with high signal-to-noise ratio (SNR), 3D coverage, and produce robust contrast. However, spectrally selective CEST contrast requires dense sampling of the Z-spectrum, which increases scan duration. This article proposes a compromise: using a 3D snapshot gradient echo (GRE) readout with optimized CEST presaturation, sampling, and postprocessing, highly resolved Z-spectroscopy at 3T...
22 CitationsSource
#1Aaron Kujawa (UCL: University College London)H-Index: 3
#2Mina Kim (UCL: University College London)H-Index: 3
Last. Xavier Golay (UCL: University College London)H-Index: 5
view all 7 authors...
Abstract Fitting a model based on the Bloch-McConnell (BM) equations to Chemical Exchange Saturation Transfer (CEST) spectra allows for the quantification of metabolite concentration and exchange rate as well as simultaneous correction of field inhomogeneity, direct water saturation and magnetization transfer. Employing a Bayesian fitting approach permits the integration of prior information into the analysis to incorporate expected parameter distributions and to prevent over-fitting. However, t...
3 CitationsSource
#1Moritz Zaiss (MPG: Max Planck Society)H-Index: 28
#2Kai Herz (MPG: Max Planck Society)H-Index: 8
Last. Klaus Scheffler (MPG: Max Planck Society)H-Index: 76
view all 9 authors...
Abstract Purpose Dynamic CEST studies such as dynamic glucose enhanced imaging, have gained a lot of attention recently. The expected CEST effects after injection are rather small in tissue especially at clinical field strengths (0.5–2%). Small movements during the dynamic CEST measurement together with a subtraction-based evaluation can lead to pseudo CEST effects of the same order of magnitude. These artifacts are studied herein. Methods A brain tumor patient 3D-CEST baseline scan without gluc...
24 CitationsSource
#1Ouri Cohen (Harvard University)H-Index: 5
#2Shuning Huang (Harvard University)H-Index: 12
Last. Christian T. Farrar (Harvard University)H-Index: 30
view all 5 authors...
PURPOSE: To develop a fast magnetic resonance fingerprinting (MRF) method for quantitative chemical exchange saturation transfer (CEST) imaging. METHODS: We implemented a CEST-MRF method to quantify the chemical exchange rate and volume fraction of the Nα -amine protons of L-arginine (L-Arg) phantoms and the amide and semi-solid exchangeable protons of in vivo rat brain tissue. L-Arg phantoms were made with different concentrations (25-100 mM) and pH (pH 4-6). The MRF acquisition schedule varied...
50 CitationsSource
#1Moritz Zaiss (MPG: Max Planck Society)H-Index: 28
#2Mark Schuppert (MPG: Max Planck Society)H-Index: 4
Last. Klaus Scheffler (MPG: Max Planck Society)H-Index: 76
view all 10 authors...
Abstract Purpose The high chemical shift separation at 9.4 T allows for selective saturation of proton pools in exchange with water protons. For the first time, highly selective and comprehensive chemical exchange saturation transfer (CEST) experiments were performed in the human brain at 9.4 T. This work provides insight into CEST signals in the human brain in comparison with existing animal studies, as well as with CEST effects in vivo at lower field strengths. Methods A novel snapshot-CEST me...
17 CitationsSource
Sep 16, 2018 in MICCAI (Medical Image Computing and Computer-Assisted Intervention)
#1Felix J. S. Bragman (UCL: University College London)H-Index: 8
#2Ryutaro Tanno (UCL: University College London)H-Index: 14
Last. M. Jorge Cardoso (UCL: University College London)H-Index: 37
view all 9 authors...
Multi-task neural network architectures provide a mechanism that jointly integrates information from distinct sources. It is ideal in the context of MR-only radiotherapy planning as it can jointly regress a synthetic CT (synCT) scan and segment organs-at-risk (OAR) from MRI. We propose a probabilistic multi-task network that estimates: (1) intrinsic uncertainty through a heteroscedastic noise model for spatially-adaptive task loss weighting and (2) parameter uncertainty through approximate Bayes...
28 CitationsSource
Cited By10
Newest
#1Peter C.M. van Zijl (ICSCI: Kennedy Krieger Institute)H-Index: 109
#2Kevin M. Brindle (University of Cambridge)H-Index: 80
Last. Linda Knutsson (JHUSOM: Johns Hopkins University School of Medicine)H-Index: 20
view all 7 authors...
Abstract null null Access to metabolic information in vivo using magnetic resonance (MR) technologies has generally been the niche of MR spectroscopy (MRS) and spectroscopic imaging (MRSI). Metabolic fluxes can be studied using the infusion of substrates labeled with magnetic isotopes, with the use of hyperpolarization especially powerful. Unfortunately, these promising methods are not yet accepted clinically, where fast, simple, and reliable measurement and diagnosis are key. Recent advances in...
Source
#1Bonnie Lam (University of California, Berkeley)H-Index: 2
#2Michael F. Wendland (University of California, Berkeley)H-Index: 66
Last. Moriel H. Vandsburger (University of California, Berkeley)H-Index: 13
view all 5 authors...
Cardiac chemical exchange saturation transfer-magnetic resonance imaging (CEST-MRI) has been used to probe levels of various metabolites that provide insight into myocardial structure and function. However, imaging of the heart using CEST-MRI is prolonged by the need to repeatedly acquire multiple images for a full Z-spectrum and to perform saturation and acquisition around cardiac and respiratory cycles. Compressed sensing (CS) reconstruction of sparse data enables accelerated acquisition, but ...
Source
#1Christina Gillmann (Leipzig University)H-Index: 4
#2Dorothee Saur (Leipzig University)H-Index: 28
Last. Gerik Scheuermann (Leipzig University)H-Index: 37
view all 4 authors...
3 CitationsSource
#1Kai Herz (MPG: Max Planck Society)H-Index: 8
#2Sebastian Mueller (MPG: Max Planck Society)H-Index: 5
Last. Moritz Zaiss (MPG: Max Planck Society)H-Index: 28
view all 15 authors...
PURPOSE As the field of CEST grows, various novel preparation periods using different parameters are being introduced. At the same time, large, multisite clinical studies require clearly defined protocols, especially across different vendors. Here, we propose a CEST definition standard using the open Pulseq format for a shareable, simple, and exact definition of CEST protocols. METHODS We present the benefits of such a standard in three ways: (1) an open database on GitHub, where fully defined, ...
Source
#6Juergen Herrler (FAU: University of Erlangen-Nuremberg)
Since the seminal works by Brodmann and contemporaries, it is well-known that different brain regions exhibit unique cytoarchitectonic and myeloarchitectonic features. Transferring the approach of classifying brain tissues - and other tissues - based on their intrinsic features to the realm of magnetic resonance (MR) is a longstanding endeavor. In the 1990s, atlas-based segmentation replaced earlier multi-spectral classification approaches because of the large overlap between the class distribut...
1 CitationsSource
#1G. Ninos (UWest: University of the West)H-Index: 1
#2V. Bartzis (UWest: University of the West)H-Index: 6
Last. Ioannis E. Sarris (UWest: University of the West)H-Index: 16
view all 4 authors...
Abstract Background and objective: Human hemodynamic modelling is usually influenced by uncertainties occurring from a considerable unavailability of information linked to the boundary conditions and the physical properties used in the numerical models. Calculating the effect of these uncertainties on the numerical findings along the cardiovascular system is a demanding process due to the complexity of the morphology of the body and the area dynamics. To cope with all these difficulties, Uncerta...
3 CitationsSource
#2Brige Chugh (Sunnybrook Research Institute)H-Index: 1
Last. Angus Z. LauH-Index: 17
view all 7 authors...
MR-only radiation treatment planning is attractive due to the superior soft tissue definition of MRI as compared to CT, and the elimination of the uncertainty introduced by CT-MRI registration. To facilitate MR-only radiation therapy planning, synthetic-CT (sCT) algorithms (for electron density correction) are required for dose calculation. Deep neural networks for sCT generation are useful due to their predictive power, but lack of uncertainty information is a concern for clinical implementatio...
4 CitationsSource
#1Chenlu Guo (Ha Tai: Xiamen University)H-Index: 1
#2Jian Wu (Ha Tai: Xiamen University)H-Index: 3
Last. Congbo Cai (Ha Tai: Xiamen University)H-Index: 15
view all 6 authors...
PURPOSE To develop a method for fast chemical exchange saturation transfer (CEST) imaging. METHODS The periodically rotated overlapping parallel lines enhanced reconstruction (PROPELLER) sampling scheme was introduced to shorten the acquisition time. Deep neural network was employed to reconstruct CEST contrast images. Numerical simulation and experiments on a creatine phantom, hen egg, and in vivo tumor rat brain were performed to test the feasibility of this method. RESULTS The results from nu...
2 CitationsSource
#1Rahel Heule (MPG: Max Planck Society)H-Index: 10
#2Jonas Bause (MPG: Max Planck Society)H-Index: 8
Last. Klaus Scheffler (University of Tübingen)H-Index: 76
view all 4 authors...
PURPOSE Standard relaxation time quantification using phase-cycled balanced steady-state free precession (bSSFP), eg, motion-insensitive rapid configuration relaxometry (MIRACLE), is subject to a considerable underestimation of tissue T1 and T2 due to asymmetric intra-voxel frequency distributions. In this work, an artificial neural network (ANN) fitting approach is proposed to simultaneously extract accurate reference relaxation times (T1 , T2 ) and robust field map estimates ( B1+ , ΔB0 ) from...
1 CitationsSource
#1Or Perlman (Harvard University)H-Index: 7
#2Hirotaka Ito (Brigham and Women's Hospital)H-Index: 15
Last. Christian T. Farrar (Harvard University)H-Index: 30
view all 10 authors...
Oncolytic virotherapy is a promising treatment for high mortality cancers. Non-invasive imaging of the underlying molecular processes is an essential tool for therapy optimization and assessment of viral spread, innate immunity, and therapeutic response. However, previous methods for imaging oncolytic viruses did not correlate with late viral activity or had poor sensitivity and specificity. Similarly, methods developed to image treatment response, such as apoptosis, proved to be slow, nonspecif...
6 CitationsSource