Pith. sign in

REVIEW 3 major objections 5 minor 62 references

Adaptive Gate-Aware Mamba Networks for Magnetic Resonance Fingerprinting

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read GAST-Mamba is an end-to-end Mamba-based network that reconstructs T1/T2 maps from 200-frame undersampled MRF signals, reporting 33.12 dB T1 PSNR versus 31.69 dB for the SCQ baseline.

desk verdict A solid Mamba-based MRF reconstruction paper with a credible T1 story, but the T2 headline is undercut by unvalidated T2 = 1.5 x T2* labels and the GAST benefit is smaller than claimed; worth refereeing with revisions. read the letter →

arxiv 2507.03369 v1 pith:6LAGMY35 submitted 2025-07-04 eess.IV cs.LGeess.SP

Tianyi Ding , Hongli Chen , Yang Gao , Zhuang Xiong , Feng Liu , Martijn A. Cloos , Hongfu Sun This is my paper · ORCID
classification eess.IVcs.LGeess.SP
keywords magneticresonancefingerprintingdeeplearningreconstructionMambastructuredstate-spacemodelsquantitativeT1andT2mappingundersampledradialacquisitiongatemechanismimage
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that Magnetic Resonance Fingerprinting (MRF) reconstruction can be done end-to-end by a Mamba-based deep network, replacing the conventional dictionary-matching step with a learned regression from undersampled fingerprint signals to T1 and T2 maps. On 5x accelerated simulated data truncated to 200 frames, the proposed GAST-Mamba reports a T1 PSNR of 33.12 dB and SSIM of 0.9674, outperforming SCQ (31.69 dB) and three other learned baselines, with T2 PSNR 30.62 dB and SSIM 0.9124. The same model remains the best across SNR levels from 10 to 40 dB and across acquisition lengths from 200 to 1000 frames. A single healthy-volunteer in vivo experiment shows qualitatively sharper maps with fewer aliasing artifacts than dictionary matching and the baselines. If these results transfer clinically, the network offers a scalable alternative to dictionary matching for fast quantitative MRI.

What carries the argument

The load-bearing component is the Gate-Aware Spatial-Temporal Adaptive (GAST) module, which runs parallel spatial and temporal pathways and fuses them with a sigmoid gate map computed from multi-scale convolutions, so each spatial location can decide how much weight to give local texture versus temporal context. Around this module, the network uses two MambaIR-style encoders built from Residual State-Space Block Groups, which apply selective structured state-space models along four scan directions and represent the recurrence as a 1D convolution with an $O(N)$ kernel. The Mamba layers supply long-range spatial dependencies at linear complexity, while the gate resolves the trade-off between preserving fine details and maintaining global consistency; SVD compression supplies the compact 100-channel input to the network.

What would settle it

Acquire quantitative reference T1 and T2 maps on a cohort of subjects or a standardized phantom, run GAST-Mamba on matching 5x accelerated in vivo MRF data, and compute voxel-wise bias and error maps. If the learned maps show systematic T2 bias traceable to the 1.5 x T2* scaling, or if per-region errors substantially exceed the tolerances implied by the simulated PSNR, the central claim of clinical transfer fails. A simpler observation is to test the model on in vivo data containing lesions or off-resonance and check whether the reported artifact-free anatomical detail reproduces.

Watch

Extended reading notes

Core claim

The paper claims that MRF parameter estimation can be reformulated as an end-to-end regression problem solved by a structured state-space network, with no per-voxel dictionary search and no exponential growth in memory or compute as more tissue parameters are added. The proposed GAST-Mamba uses two MambaIR-style encoders to extract spatial features from SVD-compressed fingerprint time series, plus a Gate-Aware Spatial-Temporal (GAST) processor that learns a per-pixel gate between fine spatial detail and temporal context. On a simulated test set, the full model achieves the best quantitative scores among all compared methods, with the GAST module contributing the largest gain under strong undersampling. The in vivo results, although qualitative, show improved anatomical delineation and reduced aliasing relative to dictionary matching and all baselines. The paper frames the contribution as the first application of Mamba-based state-space modeling to MRF, giving global receptive fields at linear computational cost.

Load-bearing premise

The central claim rests on the assumption that Bloch-equation simulations seeded by ME-MP2RAGE-derived ground-truth maps, including the approximation T2 = 1.5 x T2*, represent real in vivo MRF physics closely enough that simulated accuracy carries over to clinical scans; the only in vivo check is qualitative and on a single healthy volunteer.

Editorial extensions

If this is right

  • At 5x acceleration, reliable T1 and T2 estimation from 200 frames becomes possible without dictionary matching, removing the exponential dictionary-growth bottleneck for multiparametric mapping.
  • Because the largest gains appear at short acquisition length and low SNR, the method targets clinically relevant regimes where severe undersampling, motion sensitivity, or limited scan time would otherwise degrade quantification.
  • The linear-complexity state-space backbone makes the architecture scalable to higher spatial resolutions or longer temporal series, where Transformer attention becomes prohibitively expensive.
  • Ablation results showing degraded performance when the Mamba encoder is replaced by a CNN support the claim that state-space modeling, not merely network depth, drives the improvement under acceleration.
  • Robustness across SNR levels from 10 to 40 dB suggests the model could work on low-field or portable MRI systems where signal is weaker, a direction the paper names as promising.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The learned gate map, which encodes where spatial context should dominate over temporal signal, could be inspected as a per-pixel reliability or aliasing indicator; correlating gate values with reconstruction errors is a testable extension the paper leaves implicit.
  • If the simulation-to-in-vivo gap turns out to be the limiting factor, then an unsupervised or self-supervised variant trained directly on in vivo signals, which the paper lists as future work, would be a sharper test of the architecture's value than further simulated benchmarks.
  • The same dual-encoder plus gated spatial-temporal fusion recipe could transfer to other quantitative MRI tasks with long temporal evolutions, such as T1rho or magnetization-transfer mapping; this extension is not claimed by the paper.
  • The T2 ground truth is derived by scaling measured T2* by 1.5, so any systematic error in that scaling would directly propagate into both simulated training labels and the reported T2 accuracy, a point the paper does not quantify.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes GAST-Mamba, an end-to-end deep learning framework for magnetic resonance fingerprinting (MRF) reconstruction that combines two Mamba-based encoders (IFE and DSFE) with a Gate-Aware Spatial-Temporal Adaptive (GAST) processor. The network maps 5x-accelerated, SVD-compressed MRF signal time series to T1 and T2 maps. On simulated IR-bSSFP data, the authors report T1 PSNR of 33.12 dB and T2 PSNR of 30.62 dB at t=200, outperforming SCQ, LG-ViT, CONV-ICA, and MRF-Mixer, with an ablation study and a qualitative single-subject in vivo demonstration.

Significance. If the quantitative claims hold, the paper would contribute a scalable, linear-complexity alternative to dictionary matching for accelerated MRF, with a plausible architectural story: Mamba-based global modeling plus a spatially adaptive gate that balances local and temporal cues. The experimental protocol is a genuine strength: all baselines are trained on the same simulated data with matched preprocessing and loss functions, and evaluation covers standard metrics at multiple acceleration levels and SNR values. The ablation is systematic, and the authors are transparent about several limitations. However, the central T2 claim is compromised by the label construction, and the claimed importance of the GAST module under strong undersampling rests on effect sizes well within the reported standard deviations.

major comments (3)
  1. [Section 2.3.1, footnote 1] The T2 labels are not true T2 values: the footnote states that T2* maps were acquired from ME-MP2RAGE and T2 was estimated by scaling T2* by a factor of 1.5. These rescaled T2* maps are used both as the Bloch-simulation parameters (Eq. 16) and as the regression targets. Because the T2/T2* ratio is tissue-dependent and affected by susceptibility and field-inhomogeneity dephasing, the reported T2 PSNR of 30.62 dB and SSIM of 0.9124 quantifies recovery of an unvalidated rescaled T2* map, not quantitative T2 accuracy. The in vivo T2 assessment in Section 3.3 is qualitative and on one healthy volunteer, so there is no independent evidence that the predicted T2 values are physically meaningful. This issue is load-bearing for the T2 half of the central claim; it must be addressed by either reframing the results as T2*-based mapping, validating against an independent T2 reference, or providing tissue-dependent justification for the fixed 1.5 scaling.
  2. [Table 4 and Section 3.4] The paper's claim that the GAST module is particularly important under strong undersampling is not quantitatively supported. At t=200, the full model gives T1 PSNR 33.12±1.09 versus 33.00±1.10 without GAST (A2), a difference of 0.12 dB; for T2 the difference is 30.62±1.54 versus 30.39±1.54, or 0.23 dB. These differences are far smaller than the reported standard deviations, and no paired statistical tests, confidence intervals, or per-sample error bars are provided. The same applies to several other ablation comparisons at t=200 (e.g., A1 and A3 differ from the full model by 0.5 dB or less). Please report paired tests or effect sizes for the key ablation comparisons, and temper the conclusion that GAST provides 'significant benefits' under high acceleration unless such evidence is provided.
  3. [Section 3.3 and Section 4.4] The in vivo evaluation is limited to a single healthy volunteer and is entirely qualitative; no quantitative in vivo metrics, reference maps, or multi-subject variability are reported. The authors acknowledge this in Section 4.4, but the abstract still concludes that the results demonstrate 'effective and robust reconstruction from highly undersampled MRF acquisitions.' Given that the simulation-to-in-vivo gap is a key uncertainty for the entire approach, the central claim should either be restricted to simulated data or supported by a quantitative in vivo validation study. A qualitative single-subject demonstration is not sufficient evidence for the generalizable robustness claimed in the abstract.
minor comments (5)
  1. [Table 3 caption] The caption says 'Ablation study configurations of FoLaMNet architecture' but the method is called GAST-Mamba; please correct this typo.
  2. [Section 2.3.1] The phrase 'Ground-truth parameter maps of T1, T21, and B0' contains a typo ('T21' should be 'T2').
  3. [Figure 4 caption] The caption contains the typo 'accerlated'; also, the statement 'Each row corresponds to a different T1/T2 combination' is vague and should describe what the rows actually show (e.g., different tissue parameter pairs or different slices).
  4. [Section 2.2 and Figure 1] Several module names are used inconsistently in the text and Figure 1 ('Initial Fingerprint Encoder' vs 'Initial Feature Extractor', and the GAST acronym is expanded differently in the abstract and in Section 2.2.2). Please harmonize the terminology throughout.
  5. [General] No code or data availability statement is provided; for a deep learning reconstruction paper, releasing the simulation pipeline and trained models would substantially strengthen reproducibility.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: GAST-Mamba is standard supervised learning on independently acquired ME-MP2RAGE-derived labels, evaluated on held-out simulated data; the T2 labels are an approximation, not a circular input.

full rationale

The central claim is a supervised-learning performance comparison. Ground-truth T1 and T2 maps come from ME-MP2RAGE acquisitions (Sun et al., 2020), an external MRI mapping method, and are not generated by the network or by the authors' prior MRF reconstruction claims. The forward model S_image = Bloch(T1, T2, B0, FAs, TRs) (Eq. 16) is used to create both training and test signals, and the network is evaluated on 90 held-out simulated samples, so the reported PSNR/SSIM values measure generalization on in-distribution data rather than a fitted-input prediction. The one caveat is footnote 1 in Sec. 2.3.1: T2 labels are defined as 1.5 x T2* maps. This makes the 'T2' results internally consistent (signals are simulated with the same scaled T2 and the network regresses to that same target), but it is a physical approximation, not an independently measured T2; this is a validity limitation for clinical T2 claims, not a circular derivation. The self-citations (MRF-Mixer baseline, Ding et al., 2025; ME-MP2RAGE, Sun et al., 2020) are either comparison implementations or data-source references and are not load-bearing arguments that reduce to the paper's own conclusions. The acknowledged single-subject qualitative in vivo evaluation (Sec. 4.4) weakens external generalization but does not introduce circularity. Overall, no specific equation or prediction reduces by construction to its input.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

The central performance claim rests on the Bloch simulation pipeline, the ME-MP2RAGE-derived ground truth, the T2* to T2 scaling constant, the SVD compression choice, and the implicit assumption that simulated training transfers to in vivo scans. These are standard domain assumptions for MRF deep learning, but the T2 scaling and the single-subject qualitative in vivo check are the least supported links. No new physical entities or invented forces are introduced.

free parameters (2)
  • T2 = 1.5 x T2* scaling factor = 1.5
    Section 2.3.1 footnote: T2 values are estimated by scaling T2* from ME-MP2RAGE by a factor of 1.5. This constant is chosen by hand, not derived or validated, and it defines the ground truth that all models are trained and evaluated against, so it directly shapes reported T2 accuracy.
  • SVD compression component count = 50
    Section 2.3.1: the top 50 singular vectors from the MRF dictionary project the 200-frame time series into 50 components. This truncation is a standard but arbitrary choice that affects the information available to the network.
assumptions (6)
  • domain assumption Bloch equation simulation accurately models MRF signal evolution.
    Section 2.3.1, Eq. 16: Simage = Bloch(T1, T2, B0, FAs, TRs). The forward model is assumed correct for generating training signals and for defining the task.
  • domain assumption ME-MP2RAGE-derived parameter maps are reliable ground truth.
    Section 2.3.1: T1, T2 and B0 maps are derived from ME-MP2RAGE scans and used as regression targets. Their accuracy is not independently validated in this paper.
  • domain assumption T2 can be approximated as 1.5 times T2*.
    Section 2.3.1 footnote: T2 values are estimated by scaling T2* with a factor of 1.5. No evidence is provided for this conversion.
  • domain assumption Simulation-to-in-vivo transfer holds.
    Section 3.3: a model trained on simulated data is applied to in vivo data. The in vivo evaluation is qualitative and on one healthy volunteer, so transfer is asserted rather than demonstrated.
  • domain assumption The dictionary-derived SVD basis preserves essential signal information after truncation to 200 frames and 50 components.
    Section 2.3.1: the top 50 singular vectors are used to compress the temporal signal. The sufficiency of this subspace for accurate T1/T2 estimation is assumed.
  • standard math Mamba/SSM recurrence provides linear-complexity sequence modeling with stable training.
    The paper relies on the linear complexity and stability properties of Mamba as established in Gu and Dao (2023) and MambaIR (Guo et al., 2024b). These are accepted results, not derived here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Adaptive Gate-Aware Mamba Networks for Magnetic Resonance Fingerprinting." pith.science (2026). https://pith.science/paper/6LAGMY35

@misc{pith2026250703369,
  author       = {Pith},
  title        = {Pith review of: Adaptive Gate-Aware Mamba Networks for Magnetic Resonance Fingerprinting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6LAGMY35}},
  note         = {Machine review of arXiv:2507.03369}
}
read the original abstract

Magnetic Resonance Fingerprinting (MRF) enables fast quantitative imaging by matching signal evolutions to a predefined dictionary. However, conventional dictionary matching suffers from exponential growth in computational cost and memory usage as the number of parameters increases, limiting its scalability to multi-parametric mapping. To address this, recent work has explored deep learning-based approaches as alternatives to DM. We propose GAST-Mamba, an end-to-end framework that combines a dual Mamba-based encoder with a Gate-Aware Spatial-Temporal (GAST) processor. Built on structured state-space models, our architecture efficiently captures long-range spatial dependencies with linear complexity. On 5 times accelerated simulated MRF data (200 frames), GAST-Mamba achieved a T1 PSNR of 33.12~dB, outperforming SCQ (31.69~dB). For T2 mapping, it reached a PSNR of 30.62~dB and SSIM of 0.9124. In vivo experiments further demonstrated improved anatomical detail and reduced artifacts. Ablation studies confirmed that each component contributes to performance, with the GAST module being particularly important under strong undersampling. These results demonstrate the effectiveness of GAST-Mamba for accurate and robust reconstruction from highly undersampled MRF acquisitions, offering a scalable alternative to traditional DM-based methods.

Figures

Figures reproduced from arXiv: 2507.03369 by the authors.

Figure 1
Figure 1. Proposed GAST-Mamba network structure. The network con [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Inversion-recovery balanced steady-state free precession (IR [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Radial trajectory applied in this work. 2.3 Data Generation and Preprocessing 2.3.1 Simulation Dataset Supervised training data were generated by simulating single-shot radial MRF acquisitions using the inversion-recovery balanced steady-state free precession (IR-bSSFP) sequence. This sequence comprises 1000 repetitions with variable flip angles (FAs) and repetition times (TRs) [PITH_FULL_IMAGE:figures/full_fig_p01… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison of T1 and T2 map reconstructions on 5 [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Quantitative evaluation of model robustness to noise. PSNR and [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: Quantitative evaluation of model performance under different ac [PITH_FULL_IMAGE:figures/full_fig_p019_6.png]
Figure 7
Figure 7. Figure 7: Qualitative comparison of T1 and T2 map reconstructions on [PITH_FULL_IMAGE:figures/full_fig_p020_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

62 extracted references · 54 canonical work pages

  1. [1]

    , author Narayana, P.A

    author Bedell, B.J. , author Narayana, P.A. , year 1998 . title Volumetric analysis of white matter, gray matter, and csf using fractional volume analysis . journal Magnetic resonance in medicine volume 39 , pages 961--969

    work page 1998

  2. [2]

    , author Coppo, S

    author Bipin Mehta, B. , author Coppo, S. , author Frances McGivney, D. , author Ian Hamilton, J. , author Chen, Y. , author Jiang, Y. , author Ma, D. , author Seiberlich, N. , author Gulani, V. , author Alan Griswold, M. , year 2019 . title Magnetic resonance fingerprinting: a technical review . journal Magnetic resonance in medicine volume 81 , pages 25--46

    work page 2019

  3. [3]

    , year 1946

    author Bloch, F. , year 1946 . title Nuclear induction . journal Phys. Rev. volume 70 , pages 460--474 . https://link.aps.org/doi/10.1103/PhysRev.70.460, :10.1103/PhysRev.70.460

    doi:10.1103/physrev.70.460 1946

  4. [4]

    , author Cui, D

    author Cao, P. , author Cui, D. , author Vardhanabhuti, V. , author Hui, E.S. , year 2020 . title Development of fast deep learning quantification for magnetic resonance fingerprinting in vivo . journal Magnetic Resonance Imaging volume 70 , pages 81--90

    work page 2020

  5. [5]

    , author Setsompop, K

    author Cauley, S.F. , author Setsompop, K. , author Ma, D. , author Jiang, Y. , author Ye, H. , author Adalsteinsson, E. , author Griswold, M.A. , author Wald, L.L. , year 2015 . title Fast group matching for mr fingerprinting reconstruction . journal Magnetic resonance in medicine volume 74 , pages 523--528

    work page 2015

  6. [6]

    , author Groves, A.R

    author Chappell, M.A. , author Groves, A.R. , author MacIntosh, B.J. , author Donahue, M.J. , author Jezzard, P. , author Woolrich, M.W. , year 2011 . title Partial volume correction of multiple inversion time arterial spin labeling mri data . journal Magnetic resonance in medicine volume 65 , pages 1173--1183

    work page 2011

  7. [7]

    , author Lu, L

    author Chen, Y. , author Lu, L. , author Zhu, T. , author Ma, D. , year 2022 . title Technical overview of magnetic resonance fingerprinting and its applications in radiation therapy . journal Medical physics volume 49 , pages 2846--2860

    work page 2022

  8. [8]

    , author Haynor, D.R

    author Choi, H.S. , author Haynor, D.R. , author Kim, Y. , year 1991 . title Partial volume tissue classification of multichannel magnetic resonance images-a mixel model . journal IEEE Transactions on Medical Imaging volume 10 , pages 395--407

    work page 1991

Show all 62 references
  1. [9]

    , author Assl \"a nder, J

    author Cloos, M.A. , author Assl \"a nder, J. , author Abbas, B. , author Fishbaugh, J. , author Babb, J.S. , author Gerig, G. , author Lattanzi, R. , year 2019 . title Rapid radial t1 and t2 mapping of the hip articular cartilage with magnetic resonance fingerprinting . journ...

    2019

  2. [10]

    , author Zhu, B

    author Cohen, O. , author Zhu, B. , author Rosen, M.S. , year 2018 . title Mr fingerprinting deep reconstruction network (drone) . journal Magnetic resonance in medicine volume 80 , pages 885--894

    2018

  3. [11]

    , author Cruz, G

    author Coronado, R. , author Cruz, G. , author Castillo-Passi, C. , author Tejos, C. , author Uribe, S. , author Prieto, C. , author Irarrazaval, P. , year 2021 . title A spatial off-resonance correction in spirals for magnetic resonance fingerprinting . journal IEEE Transacti...

    2021

  4. [12]

    , author Cammoun, L

    author Cuadra, M.B. , author Cammoun, L. , author Butz, T. , author Cuisenaire, O. , author Thiran, J.P. , year 2005 . title Comparison and validation of tissue modelization and statistical classification methods in t1-weighted mr brain images . journal IEEE transactions on me...

    2005

  5. [13]

    , author Peters, T.M

    author Deoni, S.C. , author Peters, T.M. , author Rutt, B.K. , year 2005 . title High-resolution t1 and t2 mapping of the brain in a clinically acceptable time with despot1 and despot2 . journal Magnetic Resonance in Medicine: An Official Journal of the International Society f...

    2005

  6. [14]

    , author McGivney, D.F

    author Deshmane, A. , author McGivney, D.F. , author Ma, D. , author Jiang, Y. , author Badve, C. , author Gulani, V. , author Seiberlich, N. , author Griswold, M.A. , year 2019 . title Partial volume mapping using magnetic resonance fingerprinting . journal NMR in Biomedicine...

    2019

  7. [15]

    , author Gao, Y

    author Ding, T. , author Gao, Y. , author Xiong, Z. , author Liu, F. , author Cloos, M.A. , author Sun, H. , year 2025 . title Mrf-mixer: A simulation-based deep learning framework for accelerated and accurate magnetic resonance fingerprinting reconstruction . journal Informat...

    2025

  8. [16]

    , author Cobzas, D

    author Elkady, A.M. , author Cobzas, D. , author Sun, H. , author Seres, P. , author Blevins, G. , author Wilman, A.H. , year 2019 . title Five year iron changes in relapsing-remitting multiple sclerosis deep gray matter compared to healthy controls . journal Multiple Sclerosi...

    2019 doi

  9. [17]

    , author Chen, Y

    author Fang, Z. , author Chen, Y. , author Hung, S.C. , author Zhang, X. , author Lin, W. , author Shen, D. , year 2020 . title Submillimeter mr fingerprinting using deep learning--based tissue quantification . journal Magnetic resonance in medicine volume 84 , pages 579--591

    2020

  10. [18]

    , author Chen, Y

    author Fang, Z. , author Chen, Y. , author Liu, M. , author Xiang, L. , author Zhang, Q. , author Wang, Q. , author Lin, W. , author Shen, D. , year 2019 a. title Deep learning for fast and spatially constrained tissue quantification from highly accelerated data in magnetic re...

    2019

  11. [19]

    , author Chen, Y

    author Fang, Z. , author Chen, Y. , author Nie, D. , author Lin, W. , author Shen, D. , year 2019 b. title Rca-u-net: Residual channel attention u-net for fast tissue quantification in magnetic resonance fingerprinting , in: booktitle Medical Image Computing and Computer Assis...

    2019

  12. [20]

    , author Yan, Y

    author Feng, C.M. , author Yan, Y. , author Chen, G. , author Xu, Y. , author Hu, Y. , author Shao, L. , author Fu, H. , year 2022 . title Multimodal transformer for accelerated mr imaging . journal IEEE Transactions on Medical Imaging volume 42 , pages 2804--2816

    2022

  13. [21]

    , author Nair, G

    author Gao, K.C. , author Nair, G. , author Cortese, I.C. , author Koretsky, A. , author Reich, D.S. , year 2014 . title Sub-millimeter imaging of brain-free water for rapid volume assessment in atrophic brains . journal Neuroimage volume 100 , pages 370--378

    2014

  14. [22]

    , author Xiong, Z

    author Gao, Y. , author Xiong, Z. , author Fazlollahi, A. , author Nestor, P.J. , author Vegh, V. , author Nasrallah, F. , author Winter, C. , author Pike, G.B. , author Crozier, S. , author Liu, F. , et al., year 2022 . title Instant tissue field and magnetic susceptibility m...

    2022

  15. [23]

    , author Chen, D

    author Golbabaee, M. , author Chen, D. , author G \'o mez, P.A. , author Menzel, M.I. , author Davies, M.E. , year 2019 . title Geometry of deep learning for magnetic resonance fingerprinting , in: booktitle ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech a...

    2019

  16. [24]

    , author Cencini, M

    author G \'o mez, P.A. , author Cencini, M. , author Golbabaee, M. , author Schulte, R.F. , author Pirkl, C. , author Horvath, I. , author Fallo, G. , author Peretti, L. , author Tosetti, M. , author Menze, B.H. , et al., year 2020 . title Rapid three-dimensional multiparametr...

    2020

  17. [25]

    , author Dao, T

    author Gu, A. , author Dao, T. , year 2023 . title Mamba: Linear-time sequence modeling with selective state spaces . journal arXiv preprint arXiv:2312.00752

    2023 arXiv

  18. [26]

    , author Dao, T

    author Gu, A. , author Dao, T. , author Ermon, S. , author Rudra, A. , author R \'e , C. , year 2020 . title Hippo: Recurrent memory with optimal polynomial projections . journal Advances in neural information processing systems volume 33 , pages 1474--1487

    2020

  19. [27]

    , author Guo, Y

    author Guo, H. , author Guo, Y. , author Zha, Y. , author Zhang, Y. , author Li, W. , author Dai, T. , author Xia, S.T. , author Li, Y. , year 2024 a. title Mambairv2: Attentive state space restoration . journal arXiv preprint arXiv:2411.15269

    2024 arXiv

  20. [28]

    , author Li, J

    author Guo, H. , author Li, J. , author Dai, T. , author Ouyang, Z. , author Ren, X. , author Xia, S.T. , year 2024 b. title Mambair: A simple baseline for image restoration with state-space model , in: booktitle European Conference on Computer Vision , organization Springer ....

    2024

  21. [29]

    , author Currey, D

    author Hamilton, J.I. , author Currey, D. , author Rajagopalan, S. , author Seiberlich, N. , year 2021 . title Deep learning reconstruction for cardiac magnetic resonance fingerprinting t1 and t2 mapping . journal Magnetic resonance in medicine volume 85 , pages 2127--2135

    2021

  22. [30]

    , author Wang, Y

    author Han, K. , author Wang, Y. , author Chen, H. , author Chen, X. , author Guo, J. , author Liu, Z. , author Tang, Y. , author Xiao, A. , author Xu, C. , author Xu, Y. , et al., year 2022 . title A survey on vision transformer . journal IEEE transactions on pattern analysis...

    2022

  23. [31]

    , author Guttmann, C.R

    author Kikinis, R. , author Guttmann, C.R. , author Metcalf, D. , author Wells III, W.M. , author Ettinger, G.J. , author Weiner, H.L. , author Jolesz, F.A. , year 1999 . title Quantitative follow-up of patients with multiple sclerosis using mri: technical aspects . journal Jo...

    1999

  24. [32]

    , author Hu, Y

    author Li, P. , author Hu, Y. , year 2024 a. title Deep magnetic resonance fingerprinting based on local and global vision transformer . journal Medical Image Analysis volume 95 , pages 103198

    2024

  25. [33]

    , author Hu, Y

    author Li, P. , author Hu, Y. , year 2024 b. title Enhancing mrf reconstruction: A model-based deep learning approach leveraging learned sparsity and physics priors . journal IEEE Transactions on Computational Imaging

    2024

  26. [34]

    , author Hu, Y

    author Li, P. , author Hu, Y. , year 2025 . title Deep graph embedding based on laplacian eigenmaps for mr fingerprinting reconstruction . journal Medical Image Analysis , pages 103481

    2025

  27. [35]

    , author Hamilton, J

    author Liu, Y. , author Hamilton, J. , author Jiang, Y. , author Seiberlich, N. , year 2023 . title Assessment of mrf for simultaneous t 1 and t 2 quantification and water--fat separation in the liver at 0.55 t . journal Magnetic Resonance Materials in Physics, Biology and Med...

    2023

  28. [36]

    , author Tian, Y

    author Liu, Y. , author Tian, Y. , author Zhao, Y. , author Yu, H. , author Xie, L. , author Wang, Y. , author Ye, Q. , author Jiao, J. , author Liu, Y. , year 2024 . title Vmamba: Visual state space model , in: editor Globerson, A. , editor Mackey, L. , editor Belgrave, D. , ...

    2024

  29. [37]

    , author Gulani, V

    author Ma, D. , author Gulani, V. , author Seiberlich, N. , author Liu, K. , author Sunshine, J.L. , author Duerk, J.L. , author Griswold, M.A. , year 2013 . title Magnetic resonance fingerprinting . journal Nature volume 495 , pages 187--192

    2013

  30. [38]

    , author Jiang, Y

    author Ma, D. , author Jiang, Y. , author Chen, Y. , author McGivney, D. , author Mehta, B. , author Gulani, V. , author Griswold, M. , year 2018 . title Fast 3d magnetic resonance fingerprinting for a whole-brain coverage . journal Magnetic resonance in medicine volume 79 , p...

    2018

  31. [39]

    , author Victoria, Y.Y

    author Martinez, J.A. , author Victoria, Y.Y. , author Tringale, K.R. , author Otazo, R. , author Cohen, O. , year 2024 . title Phase-sensitive deep reconstruction method for rapid multiparametric mr fingerprinting and quantitative susceptibility mapping in the brain . journal...

    2024

  32. [40]

    , author Boyacıoğlu, R

    author McGivney, D.F. , author Boyacıoğlu, R. , author Jiang, Y. , author Poorman, M.E. , author Seiberlich, N. , author Gulani, V. , author Keenan, K.E. , author Griswold, M.A. , author Ma, D. , year 2020 . title Magnetic resonance fingerprinting review part 2: Technique and ...

    2020 doi

  33. [41]

    , author Pierre, E

    author McGivney, D.F. , author Pierre, E. , author Ma, D. , author Jiang, Y. , author Saybasili, H. , author Gulani, V. , author Griswold, M.A. , year 2014 . title Svd compression for magnetic resonance fingerprinting in the time domain . journal IEEE transactions on medical i...

    2014

  34. [42]

    , author Hermann, I

    author Nagtegaal, M.A. , author Hermann, I. , author Weing \"a rtner, S. , author Martinez-Heras, E. , author Solana, E. , author Llufriu, S. , author Gass, A. , author Poot, D.H. , author van Osch, M.J. , author Vos, F.M. , et al., year 2023 . title White matter changes measu...

    2023

  35. [43]

    , author Turkstra, L

    author Plante, E. , author Turkstra, L. , year 1991 . title Sources of error in the quantitative analysis of mri scans . journal Magnetic resonance imaging volume 9 , pages 589--595

    1991

  36. [44]

    , author Martin, M.N

    author Poorman, M.E. , author Martin, M.N. , author Ma, D. , author McGivney, D.F. , author Gulani, V. , author Griswold, M.A. , author Keenan, K.E. , year 2020 . title Magnetic resonance fingerprinting part 1: Potential uses, current challenges, and recommendations . journal ...

    2020

  37. [45]

    , author Jaggi, C

    author Ruan, S. , author Jaggi, C. , author Xue, J. , author Fadili, J. , author Bloyet, D. , year 2000 . title Brain tissue classification of magnetic resonance images using partial volume modeling . journal IEEE transactions on Medical Imaging volume 19 , pages 1179--1187

    2000

  38. [46]

    , author Dickey, C.C

    author Shenton, M.E. , author Dickey, C.C. , author Frumin, M. , author McCarley, R.W. , year 2001 . title A review of mri findings in schizophrenia . journal Schizophrenia research volume 49 , pages 1--52

    2001

  39. [47]

    , author Navruz, E

    author Soyak, R. , author Navruz, E. , author Ersoy, E.O. , author Cruz, G. , author Prieto, C. , author King, A.P. , author Unay, D. , author Oksuz, I. , year 2021 . title Channel attention networks for robust mr fingerprint matching . journal IEEE Transactions on Biomedical ...

    2021

  40. [48]

    , author Boudreau, M

    author Stikov, N. , author Boudreau, M. , author Levesque, I.R. , author Tardif, C.L. , author Barral, J.K. , author Pike, G.B. , year 2015 . title On the accuracy of t1 mapping: searching for common ground . journal Magnetic resonance in medicine volume 73 , pages 514--522

    2015

  41. [49]

    , author Cleary, J.O

    author Sun, H. , author Cleary, J.O. , author Glarin, R. , author Kolbe, S.C. , author Ordidge, R.J. , author Moffat, B.A. , author Pike, G.B. , year 2020 . title Extracting more for less: multi-echo mp2rage for simultaneous t1-weighted imaging, t1 mapping, mapping, swi, and q...

    2020

  42. [50]

    title Bunya supercomputer

    author The University of Queensland Research Computing Centre , year 2024 . title Bunya supercomputer . howpublished https://dx.doi.org/10.48610/wf6c-qy55 . note Accessed: 2024

    2024 doi

  43. [51]

    , year 2014

    author Tohka, J. , year 2014 . title Partial volume effect modeling for segmentation and tissue classification of brain magnetic resonance images: A review . journal World journal of radiology volume 6 , pages 855

    2014

  44. [52]

    , author Zijdenbos, A

    author Tohka, J. , author Zijdenbos, A. , author Evans, A. , year 2004 . title Fast and robust parameter estimation for statistical partial volume models in brain mri . journal Neuroimage volume 23 , pages 84--97

    2004

  45. [53]

    , author Shazeer, N

    author Vaswani, A. , author Shazeer, N. , author Parmar, N. , author Uszkoreit, J. , author Jones, L. , author Gomez, A.N. , author Kaiser, . , author Polosukhin, I. , year 2017 . title Attention is all you need . journal Advances in neural information processing systems volume 30

    2017

  46. [54]

    , author Ostenson, J

    author Wang, D. , author Ostenson, J. , author Smith, D.S. , year 2020 . title snapmrf: Gpu-accelerated magnetic resonance fingerprinting dictionary generation and matching using extended phase graphs . journal Magnetic resonance imaging volume 66 , pages 248--256

    2020

  47. [55]

    , author Chen, J

    author Wang, J. , author Chen, J. , author Chen, D. , author Wu, J. , year 2024 a. title Lkm-unet: Large kernel vision mamba unet for medical image segmentation , in: booktitle International Conference on Medical Image Computing and Computer-Assisted Intervention , organizatio...

    2024

  48. [56]

    , author Zheng, J.Q

    author Wang, Z. , author Zheng, J.Q. , author Zhang, Y. , author Cui, G. , author Li, L. , year 2024 b. title Mamba-unet: Unet-like pure visual mamba for medical image segmentation . journal arXiv preprint arXiv:2402.05079

    2024 arXiv

  49. [57]

    , author Veitch, D.P

    author Weiner, M.W. , author Veitch, D.P. , author Aisen, P.S. , author Beckett, L.A. , author Cairns, N.J. , author Green, R.C. , author Harvey, D. , author Jack, C.R. , author Jagust, W. , author Liu, E. , et al., year 2013 . title The alzheimer's disease neuroimaging initia...

    2013

  50. [58]

    , author Ye, T

    author Xing, Z. , author Ye, T. , author Yang, Y. , author Liu, G. , author Zhu, L. , year 2024 . title Segmamba: Long-range sequential modeling mamba for 3d medical image segmentation , in: booktitle International Conference on Medical Image Computing and Computer-Assisted In...

    2024

  51. [59]

    , author Ma, D

    author Yang, M. , author Ma, D. , author Jiang, Y. , author Hamilton, J. , author Seiberlich, N. , author Griswold, M.A. , author McGivney, D. , year 2018 . title Low rank approximation methods for mr fingerprinting with large scale dictionaries . journal Magnetic resonance in...

    2018

  52. [60]

    , author Li, Z

    author Yue, Y. , author Li, Z. , year 2024 . title Medmamba: Vision mamba for medical image classification . journal arXiv preprint arXiv:2403.03849

    2024 arXiv

  53. [61]

    , author Lipton, Z.C

    author Zhang, A. , author Lipton, Z.C. , author Li, M. , author Smola, A.J. , year 2023 . title Dive into deep learning . publisher Cambridge University Press

    2023

  54. [62]

    , author Brady, M

    author Zhang, Y. , author Brady, M. , author Smith, S. , year 2001 . title Segmentation of brain mr images through a hidden markov random field model and the expectation-maximization algorithm . journal IEEE transactions on medical imaging volume 20 , pages 45--57

    2001

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.