Space Environment and Debris: A Review of Micro-Meteoroids and Orbital Debris Impact Protection

Binkal Kumar Sharma; Harshitha Baskar

REVIEW 3 major objections 3 minor 24 references

Space Environment and Debris: A Review of Micro-Meteoroids and Orbital Debris Impact Protection

T0 review · 3 major / 3 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 This paper claims that grouping isochromats with shared T1, T2, position, and field offset cuts Bloch simulation time by 3 to 72 times over conventional methods.

desk verdict The real content is a solid but modest MRI simulation acceleration paper; the metadata says it's a space-debris review, and the fidelity evidence is under-powered. read the letter →

arxiv 2508.03957 v1 pith:36W3DAOO submitted 2025-08-05 physics.space-ph

Binkal Kumar Sharma , Harshitha Baskar This is my paper · ORCID

classification physics.space-ph

keywords MRsimulationgroupedisochromatsBlochequationscombinedtransitionsfastspinechoecho-planarimagingisochromatclusteringCPUacceleration

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 argues that MRI simulation need not treat every isochromat separately. Isochromats that share T1, T2, location along one active gradient axis, and field inhomogeneity undergo identical Bloch transitions during RF and ADC windows, so the simulation can combine them into groups. Preparation cost for RF windows drops from $O(N_{\mathrm{RF}} K)$ to $O(N_{\mathrm{RF}} K_{\mathrm{group}})$, and ADC sampling from $O(N_{\mathrm{ADC}} P K)$ to $O(N_{\mathrm{ADC}} P K_{\mathrm{group}})$. On measured brain phantoms preprocessed by clustering, the reported processing times are 3 to 72 times shorter than the conventional combined-transition baseline, reducing a 208-second FSE simulation to 38 seconds and a 66-second EPI simulation to 7 seconds. A reader should care because CPU-only simulation of millions of isochromats becomes practical enough for routine sequence prototyping and parameter-map studies.

What carries the argument

Grouped isochromats: a set of isochromats sharing T1, T2, position along one active gradient axis, and $ΔB_0$, so their Bloch transitions are identical during with-RF and with-ADC subsequences. The key identity is Eq. (10), $A_{\mathrm{block}}(k_{\mathrm{group}},p,t) = \exp(-(t-t_0)/T_2(k_{\mathrm{group}})) \exp(-i \int_{t_0}^t \gamma B_z(k_{\mathrm{group}},t) dt)\, A_{\mathrm{block}}(k_{\mathrm{group}},p,t_0)$, which turns the per-isochromat ADC sum into a per-group update and reduces the dominant complexity terms.

What would settle it

Run the T2W and EPI brain-phantom simulations with 64, 128, and 256 clusters and report a quantitative image-error metric (e.g., normalized RMS difference or SSIM) between clustered and unclustered images; if long-TR images at 256 clusters still exceed a prechosen error threshold, the claimed speedup is bought at a fidelity cost the paper has not quantified.

Watch

Extended reading notes

Core claim

The central discovery is that the conventional assumption—each isochromat must be simulated individually—can be relaxed for a common class of pulse sequences. When isochromats share their T1, T2, position along the single active gradient axis, and $ΔB_0$ offset, their Bloch transitions during with-RF subsequences are identical, so the combined transition matrix can be prepared once per group instead of once per isochromat. For with-ADC subsequences, the k-space signal factorizes: the block signal at time $t$ equals a T2 decay and phase factor times the block signal at $t_0$, allowing the expensive per-isochromat sum to be replaced by per-group amplitudes. On a measured brain phantom clustere

Load-bearing premise

The speedup depends on grouping many isochromats into a few groups by slightly changing their T1, T2, and $ΔB_0$ values, and the paper's only evidence that this changed physics is acceptable is visual inspection of error-scaled images rather than a quantitative error metric.

Editorial extensions

If this is right

If the speedup holds, CPU-only simulation of a 27.5-million-isochromat FSE sequence drops from 208.4 seconds to 38.1 seconds, and EPI from 66.4 seconds to 7.1 seconds, making high-resolution parameter-map simulation practical without GPU hardware.
The method accelerates exactly the subsequences that dominate cost—with-RF and with-ADC windows—so sequences written in the evaluated format (SE, FSE, EPI) inherit the gain whenever those windows have one active gradient.
For measured phantoms, the gain requires preprocessing: clustering parameter maps into 64–256 groups; increasing cluster count trades a few-fold speed advantage for visibly closer images, with errors concentrated in long-TR T2W and EPI sequences.
Because the grouped path is algorithmic rather than hardware-based, it applies in both non-parallel and parallel environments and can be combined with SIMD and multi-threading for additional reduction.

Reading between the lines

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

Beyond the paper: the same grouping logic could accelerate subsequences with two or three active gradients by clustering in a higher-dimensional space of position components, T1, T2, and $ΔB_0$, which the paper leaves as remained work.
Beyond the paper: the clustering step changes each isochromat's physical parameters, so users performing quantitative T1/T2 estimation should check per-voxel parameter error, not only image similarity; the paper reports visual similarity only.
Beyond the paper: since the grouping gain multiplies with SIMD and multi-threading and does not require a GPU, it could be combined with GPU implementations to push million-isochromat simulations into interactive range for pulse-sequence design loops.

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.

Desk Editor's Note

read the letter

Colleague,

Two things you should know before you open this. First, the actual manuscript is a methods paper on fast Bloch-equation MRI simulation, despite the arXiv metadata advertising a review of micro-meteoroid and orbital debris shielding. That mismatch is not a detail to wave away; the abstract has no relation to the body. Second, the central speedup claim—3 to 72× faster—is real and consistent with the complexity analysis, but the fidelity evidence for the grouped-phantom preprocessing is visual only, and the method's utility under realistic off-resonance is untested.

What's new: the paper groups isochromats that share T1, T2, position along one active gradient axis, and ΔB0, so combined-transition preparation and ADC partial sums are computed once per group. The complexity reduction is clean: O(N_RF K) to O(N_RF K_group) for preparation, and O(NADC P K) to O(NADC P K_group) for acquisition. The reported speedups match that arithmetic. The author is honest about the trade-off: clustering coarsens tissue parameters, and the paper shows error images that are worse for long-TR T2W and EPI sequences. The fallback for overlapped gradients is acknowledged as 'remained work.' That's all creditable.

Soft spots, in order. The submission identity is broken: the title, abstract, and author list for this arXiv number describe a space-environment review, but the body is Takeshima's MRI paper internally labeled 2508.03960v1. That alone would normally be a desk-reject red flag, though here it looks like a metadata error rather than a content problem. Second, the fidelity of clustering is not quantified. The paper relies on error images scaled by 10; there is no NRMSE, SSIM, or max-error number. Worse, the measured phantom sets ΔB0 to 0 Hz, which conveniently satisfies the grouping criterion but avoids the realistic case. With a nonzero field-inhomogeneity map, K_group would likely explode toward K, and the speedup would disappear. So the headline number is supported only for a preprocessed phantom whose physical fidelity is unestablished. Third, the baseline is only the author's own prior simulator; no comparison to the public simulators cited in the introduction.

If you're in MRI simulation, this is worth a look as a modest incremental idea with clean complexity analysis. It deserves a serious referee, but the fidelity claim needs quantitative evidence, and the metadata needs fixing before it's usable.

Referee Report

3 major / 3 minor

Summary. This manuscript proposes a fast Bloch-equation MR simulation method based on grouping isochromats that share T1, T2, location along the single active gradient axis, and ΔB0. The grouped isochromats allow combined-transition preparation cost to drop from O(N_RF K) to O(N_RF K_group), and the with-ADC k-space sum to factorize into per-group evolution, reducing the per-sample cost from O(N_ADC P K) to O(N_ADC P K_group). The method is implemented in the VMRscan simulator and evaluated on a numeric circles phantom and a measured brain phantom preprocessed by bisecting K-means clustering (64/128/256 clusters) with ΔB0 set to 0 Hz. For SE, FSE, and EPI sequences, the reported speedups range from 3 to 72× over the author's previous combined-transition method, with and without SIMD. The complexity analysis is clear, and the measured timings are consistent with it.

Significance. If the fidelity concerns are resolved, the grouped-isochromat idea is a genuinely useful algorithmic contribution. It is orthogonal to hardware parallelization, works for sequences whose RF and ADC windows have a single active gradient, and is demonstrated at up to 27.5 million isochromats. The paper ships a concrete implementation and reports actual runtime measurements, which is a strength. The complexity derivation in §2.3 is sound. However, the current evidence for the '3–72×' claim is incomplete for measured phantoms: the clustering preprocessing changes the physics of each isochromat, and no quantitative accuracy budget is provided. The method's behavior under realistic field inhomogeneity is also untested. These gaps prevent the speedup numbers from being a fully meaningful scientific claim as presented.

major comments (3)

[§4.1, Figs. 6–7, S1–S2] The fidelity of the grouped brain phantom is only assessed by visual inspection of error images scaled by 10. The manuscript itself states that 'errors were high when sequences with long TR values (T2W and EPI) were simulated' (§4.1). No quantitative error metric (NRMSE, SSIM, peak absolute error) is reported for any cluster count or sequence. Because the speedup claim is intended to apply to measured phantoms, and the grouping necessarily alters T1/T2/ΔB0, the absence of an accuracy budget makes it impossible to judge whether the 3–72× speedup is achieved at an acceptable fidelity. Please add quantitative errors relative to the ungrouped reference for all sequences and cluster counts, and state a threshold for acceptable fidelity.
[§2.4 and §3.3] The grouping criterion requires isochromats to share an identical ΔB0. In the brain phantom preprocessing (§3.3), ΔB0 is set to 0 Hz, which sidesteps the realistic case of field inhomogeneity. For a continuous ΔB0 map, exact grouping on ΔB0 would leave K_group ≈ K, eliminating the complexity reduction. The paper does not evaluate any nonzero ΔB0 scenario or propose an approximation (e.g., binning). Thus the central speedup claim is only demonstrated under ΔB0=0. Either extend the method to handle smoothly varying off-resonance (or clearly state its absence as a limitation) and test it, or the conclusion should be scoped to ΔB0=0.
[§4.2 and Table 5] The subvoxel scaling study reports processing times but no error analysis for the subvoxel configurations. The supplementary figures S3–S6 show 'noticeable artifacts' and 'strong artifacts' in the 1×1×1 images, and it is unclear whether the grouped method introduces or modifies these artifacts relative to the ungrouped reference. Since the manuscript claims the method is useful for suppressing artifacts (citing [9]), please provide quantitative error metrics for these subvoxel experiments as well.

minor comments (3)

[Abstract] The headline '3 to 72 times faster' lacks context. For the measured brain phantom with SIMD (Table 3), the range is 3.78–8.82×; the 72× figure comes from the idealized circles phantom with 10 groups. Please qualify the headline range to avoid overstating the typical speedup for realistic phantoms.
[§2.3] After Eq. (9), the text says 'reduces K isochromats into PKgroup groups.' This is clear, but it should be more explicit that the first step has complexity O(PK) (touching all K isochromats) while only the subsequent per-sample step is O(N_ADC P K_group). This distinction is central to the complexity claim and deserves emphasis.
[§3.1] The baseline 'conventional method' is the author's own previous simulator [13]. This is a reasonable baseline since it already includes combined transitions and caching, but the paper should state this explicitly in the Introduction or Methods to avoid the impression that the comparison is against a naive simulator.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the speedup claim is a direct timing measurement and the complexity reductions are arithmetic consequences of the grouping definition; the only self-referential element (using the author's prior simulator as baseline) is not load-bearing.

full rationale

The paper's central claim is a measured speedup, not a derived prediction. The complexity reductions in §2.3 follow algebraically from the definition of a group: if isochromats share U(k,t,Δt), then the preparation cost drops from O(N_RF·K) to O(N_RF·K_group), and Eq. (10) factors the with-ADC sum because all members of a group share T2 and Bz. These are exact arithmetic consequences of the grouping definition, not of fitted parameters or of cited results. The 3–72× figures in Tables 3 and 5 are ratios of measured processing times (ungrouped vs. grouped runs of the same VMRscan software), so there is no fitted-input-called-prediction pattern. The K-means preprocessing in §3.3 does alter T1/T2 and zero-fills ΔB0, and the fidelity evidence is only visual inspection of error images scaled by 10 (§4.1, Figs. 6–7, S1–S2); the paper itself concedes 'errors were high when sequences with long TR values (T2W and EPI sequences) were simulated.' This is an unquantified-validation weakness, not circularity, because the speedup claim does not depend on the visual-fidelity claim being treated as a derived theorem. The baseline 'conventional method' is the author's own prior simulator [13]; this is a legitimate, direct comparison since the claim is specifically that the proposed method accelerates that prior implementation, and the times are measured rather than cited. No uniqueness theorem from the authors is invoked; no ansatz is smuggled in via citation; no known empirical result is merely renamed. The limitation for overlapped-gradient subsequences is explicitly listed as remaining work (§5), which further confirms the scope of the speedup claim rather than disguising it. Overall, the derivation chain is self-contained and the central speedup claim reduces to direct measurement plus elementary complexity arithmetic.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central speedup claim depends on (i) standard Bloch/splitting numerics, (ii) static, spatially uniform input phantoms, (iii) single-gradient subsequences, and (iv) the clustering fidelity assumption. The first two are standard in the field; (iii) is an acknowledged scope limit; (iv) is the fragile premise introduced by this paper, and its only support is qualitative visual comparison. 'Groups of isochromats' are a computational abstraction over existing isochromats, not a new physical entity.

free parameters (2)

number of clusters for brain phantom preprocessing = 64, 128, 256
Hand-chosen in §3.5 to demonstrate the speed-fidelity trade-off; reported times and error images depend on it, and the Discussion's recommendation of 'moderately large numbers of clusters such as 256' is a judgment call, not a principled criterion.
T1/T2 parameter-map candidate grids = 100 log-spaced values; T1 in 10-3000 ms, T2 in 10-500 ms
Used in §3.2 to build the measured brain phantom before clustering; a discretization choice that shapes which parameter values exist, but not central to the speedup claim.

assumptions (4)

domain assumption Bloch equations with piecewise-constant B(t) over microsecond steps and asymmetric operator splitting (Appendix A, ref [24]) accurately model the MRI signal.
The entire simulator's fidelity rests on this standard numerical treatment, cited to [24] and not re-derived in this paper.
domain assumption Isochromats are static: positions r(k) do not change with time (§2.1).
The analytic free-precession solutions (Eqs. 4-5) and the grouped Bz sharing both require static isochromats; motion, flow, and diffusion are excluded from the accelerated path.
domain assumption The subsequences to be accelerated have exactly one non-zero gradient component (Gx-only, Gy-only, Gz-only, or no-grads, §2.4).
With two or more active gradients, isochromats sharing a position axis no longer share Bz(t), grouping fails, and the subsequence falls back to the conventional path; the paper lists multi-gradient subsequences as 'remained work' (§5).
ad hoc to paper Clustering tissue parameters into 64-256 groups leaves simulated images acceptably close to the ungrouped reference.
This is the load-bearing fidelity premise of the brain-phantom evaluation (§3.3, §4.1). It is supported only by visual inspection of error images, and the measured phantom's ΔB0 was set to 0 Hz (§3.2), so field inhomogeneity is never exercised.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Space Environment and Debris: A Review of Micro-Meteoroids and Orbital Debris Impact Protection." pith.science (2026). https://pith.science/paper/36W3DAOO

@misc{pith2026250803957,
  author       = {Pith},
  title        = {Pith review of: Space Environment and Debris: A Review of Micro-Meteoroids and Orbital Debris Impact Protection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/36W3DAOO}},
  note         = {Machine review of arXiv:2508.03957}
}

read the original abstract

The threat posed by MMOD to spacecraft has escalated with the growing density of orbital objects, driven by the proliferation of satellite constellations such as Starlink and OneWeb. This paper reviews the current challenges and advancements in MMOD impact protection, emphasizing innovations in shielding technologies. The synthesis of recent developments highlights the role of hybrid materials, additive manufacturing, and international collaboration in ensuring spacecraft resilience while promoting orbital sustainability.

Figures

Figures reproduced from arXiv: 2508.03957 by the authors.

**Figure 2.** Generation of grouped isochromats. (a) A process flow for generating grouped isochro [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** The overview of the proposed method. Each subsequence is classified into three types. In [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: Preprocessing method for ensuring that the numbers of groups are su [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]

**Figure 5.** Figure 5: Reconstructed images using the circles phantom. T1W, T2W, EPI and FSE sequences were [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]

**Figure 6.** Figure 6: Reconstructed images and their error images in the case of an EPI sequence. (a) Recon [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]

**Figure 7.** Figure 7: Reconstructed images and their error images in the case of an FSE sequence. (a) Recon [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages

[9]

R. Kose, K. Kose, Blochsolver: A GPU-optimized fast 3D MRI simulator for experimentally compatible pulse sequences, Journal of Magnetic Reso- nance 281 (2017) 51–65. ��

work page 2017
[1]

Taniguchi, C

Y . Taniguchi, C. Nakaya, Y . Bito, E. Yamamoto, A high-speed MRI sim- ulator using the transition matrix method and periodicity of magnetization (in Japanese), IEICE Transaction on Information and Systems D-II 3 (1994) 556–572

work page 1994
[2]

Taniguchi, C

Y . Taniguchi, C. Nakaya, Y . Bito, E. Yamamoto, A high-speed MRI sim- ulator using the transition matrix method and periodicity of magnetization (translated from Japanese), Systems and Computers in Japan 26 (2) (1995) 54–62. ��

work page 1995
[3]

Benoit-Cattin, G

H. Benoit-Cattin, G. Collewet, B. Belaroussi, H. Saint-Jalmes, C. Odet, The SIMRI project: a versatile and interactive MRI simulator, Journal of Magnetic Resonance 173 (1) (2005) 97–115. �� . 19

work page 2005
[4]

T. H. Jochimsen, A. Sch ¨afer, R. Bammer, M. E. Moseley, E � cient simula- tion of magnetic resonance imaging with Bloch-Torrey equations using intra- voxel magnetization gradients, Journal of Magnetic Resonance 180 (1) (2006) 29–38. ��

work page 2006
[5]

Latta, M

P. Latta, M. L. Gruwel, V . Jell´uˇs, B. Tomanek, Bloch simulations with intra- voxel spin dephasing, Journal of Magnetic Resonance 203 (1) (2010) 44–51. ��

work page 2010
[6]

St ¨ocker, K

T. St ¨ocker, K. Vahedipour, D. Pflugfelder, N. J. Shah, High-performance computing MRI simulations, Magnetic Resonance in Medicine 64 (1) (2010) 186–193. ��

work page 2010
[7]

C. G. Xanthis, I. E. Venetis, A. V . Chalkias, A. H. Aletras, MRISIMUL: A GPU-based parallel approach to MRI simulations, IEEE Transactions on Medical Imaging 33 (3) (2014) 607–617. ��

work page 2014

Show all 24 references

[8]

F. Liu, J. V . Velikina, W. F. Block, R. Kijowski, A. A. Samsonov, Fast realistic MRI simulations based on generalized multi-pool exchange tis- sue model, IEEE Transactions on Medical Imaging 36 (2) (2017) 527–537. ��

2017
[10]

C. G. Xanthis, A. H. Aletras, coremri: A high-performance, publicly avail- able mr simulation platform on the cloud, PLoS ONE 14 (2019) e0216594. ��

2019
[11]

Scholand, C

N. Scholand, C. Graf, M. Uecker, E � cient bloch simulation based on pre- computed state-transition matrices, in: Proceedings of the Joint Annual Meet- ing ISMRM-ESMRMB, 2022, p. 748

2022
[12]

Castillo-Passi, R

C. Castillo-Passi, R. Coronado, G. Varela-Mattatall, C. Alberola-L ´opez, R. Botnar, P. Irarrazaval, Komamri.jl: An open-source framework for general MRI simulations with GPU acceleration, Magnetic Resonance in Medicine 90 (1) (2023) 329–342. ��

2023
[13]

Takeshima, A fast and practical computation method for magnetic reso- nance simulators, Magnetic Resonance in Medicine 90 (2) (2023) 752–760

H. Takeshima, A fast and practical computation method for magnetic reso- nance simulators, Magnetic Resonance in Medicine 90 (2) (2023) 752–760. �� . 20

2023
[14]

Bloch, Nuclear induction, Physical Review 70 (1946) 460–474

F. Bloch, Nuclear induction, Physical Review 70 (1946) 460–474. ��

1946
[15]

H. C. Torrey, Bloch equations with di �usion terms, Physical Review 104 (1956) 563–565. ��

1956
[16]

H. M. McConnell, Reaction rates by nuclear magnetic resonance, The Journal of Chemical Physics 28 (3) (1958) 430–431. ��

1958
[17]

Takeshima, Combined update using grouped isochromats for fast bloch simulation, in: Proceedings of the 2024 ISMRM & ISMRT Annual Meeting, 2024, p

H. Takeshima, Combined update using grouped isochromats for fast bloch simulation, in: Proceedings of the 2024 ISMRM & ISMRT Annual Meeting, 2024, p. 4681

2024
[18]

Haase, J

A. Haase, J. Frahm, W. Hanicke, D. Matthaei, 1h nmr chemical shift selective (CHESS) imaging, Physics in Medicine & Biology 30 (4) (1985) 341. ��

1985
[19]

W. H. Perman, M. A. Bernstein, J. C. Sandstrom, A method for correctly setting the rf flip angle, Magnetic Resonance in Medicine 9 (1) (1989) 16–24. ��

1989
[20]

J. W. Carlson, D. M. Kramer, Rapid radiofrequency calibration in MRI, Magnetic Resonance in Medicine 15 (3) (1990) 438–445. ��

1990
[21]

K. J. Layton, S. Kroboth, F. Jia, S. Littin, H. Yu, J. Leupold, J.-F. Nielsen, T. St ¨ocker, M. Zaitsev, Pulseq: A rapid and hardware-independent pulse sequence prototyping framework, Magnetic Resonance in Medicine 77 (4) (2017) 1544–1552. ��

2017
[22]

Pedregosa, G

F. Pedregosa, G. Varoquaux, A. Gramfort, V . Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V . Dubourg, J. Vanderplas, A. Pas- sos, D. Cournapeau, M. Brucher, M. Perrot, E. Duchesnay, Scikit-learn: Ma- chine learning in Python, Journal of Machine Learni...

2011
[23]

Jackson, C

J. Jackson, C. Meyer, D. Nishimura, A. Macovski, Selection of a convolu- tion function for Fourier inversion using gridding (computerised tomography application), IEEE Transactions on Medical Imaging 10 (3) (1991) 473–478. �� . 21

1991
[24]

��

Accuracy and performance analysis for bloch and bloch-mcconnell simula- tion methods, Journal of Magnetic Resonance 329 (2021) 107011. �� . 22 Appendix B. Supplementary Materials Figure S1. Reconstructed images and their error images...

2021

Pith tools

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

[1] [9]

R. Kose, K. Kose, Blochsolver: A GPU-optimized fast 3D MRI simulator for experimentally compatible pulse sequences, Journal of Magnetic Reso- nance 281 (2017) 51–65. ��

work page 2017

[2] [1]

Taniguchi, C

Y . Taniguchi, C. Nakaya, Y . Bito, E. Yamamoto, A high-speed MRI sim- ulator using the transition matrix method and periodicity of magnetization (in Japanese), IEICE Transaction on Information and Systems D-II 3 (1994) 556–572

work page 1994

[3] [2]

Taniguchi, C

Y . Taniguchi, C. Nakaya, Y . Bito, E. Yamamoto, A high-speed MRI sim- ulator using the transition matrix method and periodicity of magnetization (translated from Japanese), Systems and Computers in Japan 26 (2) (1995) 54–62. ��

work page 1995

[4] [3]

Benoit-Cattin, G

H. Benoit-Cattin, G. Collewet, B. Belaroussi, H. Saint-Jalmes, C. Odet, The SIMRI project: a versatile and interactive MRI simulator, Journal of Magnetic Resonance 173 (1) (2005) 97–115. �� . 19

work page 2005

[5] [4]

T. H. Jochimsen, A. Sch ¨afer, R. Bammer, M. E. Moseley, E � cient simula- tion of magnetic resonance imaging with Bloch-Torrey equations using intra- voxel magnetization gradients, Journal of Magnetic Resonance 180 (1) (2006) 29–38. ��

work page 2006

[6] [5]

Latta, M

P. Latta, M. L. Gruwel, V . Jell´uˇs, B. Tomanek, Bloch simulations with intra- voxel spin dephasing, Journal of Magnetic Resonance 203 (1) (2010) 44–51. ��

work page 2010

[7] [6]

St ¨ocker, K

T. St ¨ocker, K. Vahedipour, D. Pflugfelder, N. J. Shah, High-performance computing MRI simulations, Magnetic Resonance in Medicine 64 (1) (2010) 186–193. ��

work page 2010

[8] [7]

C. G. Xanthis, I. E. Venetis, A. V . Chalkias, A. H. Aletras, MRISIMUL: A GPU-based parallel approach to MRI simulations, IEEE Transactions on Medical Imaging 33 (3) (2014) 607–617. ��

work page 2014