Pith. sign in

REVIEW 5 major objections 5 minor 48 references

From Explainable to Explained AI: Ideas for Falsifying and Quantifying Explanations

T0 review · 5 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a partially trainable zero-shot self-supervised network can reconstruct 14-second breath-hold MRCP at R=25 with image quality close to triggered acquisitions and reconstruction times up to 6.7 times shorter than full

desk verdict A real MRCP reconstruction engineering contribution with a head-scratcher of a submission: the arXiv metadata abstract is about pathology explainability, while the actual paper is about zero-shot MRCP, and the abstract's quality claim is contradicted by the paper's own Results. read the letter →

arxiv 2508.09205 v2 pith:73EI4TXS submitted 2025-08-09 eess.IV cs.AIcs.CV

Yoni Schirris , Eric Marcus , Jonas Teuwen , Hugo Horlings , Efstratios Gavves This is my paper
classification eess.IVcs.AIcs.CV
keywords breath-holdMRCPzero-shotself-supervisedlearningdeeplearning-basedMRIreconstructionMRcholangiopancreatographytrainingcompressedsensingunrollednetworkpartialtrainable
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

This paper is trying to establish that zero-shot self-supervised reconstruction—training a network on the single undersampled scan being reconstructed—can make breath-hold MRCP practical at 14 seconds under 25-fold acceleration, without a large training dataset of fully sampled images. The payoff is clinical: current breath-hold MRCP needs around 20 seconds, and respiratory-triggered alternatives can fail or stretch over several minutes when breathing is irregular. The proposed mechanism is to split the unrolled reconstruction network into frozen early stages, initialized from a pretrained backbone, and a small number of trainable late stages; this cuts reconstruction time up to 6.7-fold while the image-quality metric PSNR (peak signal-to-noise ratio) drops only from 38.25 dB to 37.67 dB. If true, this offers a route from scan-specific deep learning into time-constrained clinical workflows.

What carries the argument

The central mechanism is a frozen/trainable split of a 13-stage unrolled reconstruction network, where each stage alternates a residual-network regularization block with a conjugate-gradient data consistency block that enforces the MRI encoding equation. The first $f$ stages are initialized from a self-supervised pretrained backbone and frozen; their input-to-output map is cached after the first epoch, so backpropagation passes only through the $t$ trainable stages. This cuts gradient depth and memory, and the pretrained frozen part supplies a strong starting point that keeps quality high when only a few stages are trained.

What would settle it

Scan a phantom with a known fully sampled ground truth using the same R=25 Poisson-disk/partial-Fourier pattern; if the 12/1 zero-shot reconstruction loses more than the reported ~0.6 dB against that truth, or leaves partition-encoding aliasing behind, the claim that partial training preserves fidelity is falsified. Independently, timing the 0/13 through 12/1 configurations on one GPU would check the claimed approximately linear relationship between reconstruction time and number of trainable stages.

Watch

Extended reading notes

Core claim

The paper's central claim is that zero-shot self-supervised learning can reconstruct 14-second breath-hold MRCP at R=25 with image quality clearly better than compressed sensing and close to successful respiratory-triggered acquisitions. It further claims that the practical blocker—multi-hour zero-shot training—can be lifted by freezing early unrolled stages initialized from a pretrained backbone and training only the last stages; this yields up to a 6.7-fold reduction in reconstruction time with a PSNR change of roughly half a decibel. Results also show that using a pretrained network for the frozen stages beats an $\ell_1$-wavelet compressed-sensing initialization, raising PSNR by more tha

Load-bearing premise

The load-bearing premise is that a compressed-sensing reconstruction of a 40-second breath-hold at R=6 is an acceptable stand-in for ground truth when measuring PSNR at R=25, and that the PSNR numbers come from just two volunteers; if that reference is biased, the reported quality advantage of zero-shot reconstruction is not established.

Editorial extensions

If this is right

  • A 14-second breath-hold MRCP at R=25 becomes feasible, within the 10–15 s range considered tolerable for patients who cannot hold their breath longer.
  • Freezing early stages and training only the last stages cuts reconstruction time up to 6.7-fold with random initialization and 4.3-fold with pretrained initialization, while PSNR moves from 38.25 dB (0/13) to 37.67 dB (12/1).
  • Initializing the frozen stages from a pretrained network instead of from compressed sensing raises PSNR by roughly 6.3–6.7 dB and also shortens reconstruction time.
  • At R=25, zero-shot reconstruction shows better ductal delineation and less aliasing than $\ell_1$-wavelet compressed sensing, though successful respiratory-triggered acquisitions still produce the sharpest images.
  • Because the regularization weight is learned during training, the method avoids per-scan regularization tuning beyond occasional learning-rate adjustment.

Reading between the lines

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

  • Editorial inference: the frozen-stage caching strategy is not MRCP-specific; it should transfer to other scan-specific unrolled reconstructions, such as diffusion-weighted or cardiac MRI, wherever a fully sampled readout dimension can be decoupled, since the speedup comes from reducing backpropagation depth rather than from biliary anatomy.
  • Editorial inference: the residual partition-encoding aliasing seen with the 12/1 configuration suggests the safe number of trainable stages depends on the sampling geometry; with R=2 equidistant undersampling along the partition direction, one trainable stage may be too few to fully suppress aliasing, so the optimal split should be tuned per sampling pattern.
  • Editorial inference: the clinical-translation claim would be directly testable in a blinded reader study on patients with strictures or dilations; the paper's own visual assessment by the authors and the two-volunteer PSNR numbers are not strong enough evidence for diagnostic equivalence.
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

5 major / 5 minor

Summary. The full text (pages 2–24) reports a study on zero-shot self-supervised learning reconstruction for breath-hold magnetic resonance cholangiopancreatography (MRCP). The authors propose combining 2D Poisson-disk and partial-Fourier undersampling to achieve a 14-second breath-hold at R=25, then reconstruct images with a zero-shot unrolled network. To reduce training time, they split the 13 unrolled stages into frozen and trainable parts, initializing frozen stages with a pretrained SSDU backbone and training only the last stages. Experiments on 11 healthy volunteers compare zero-shot reconstructions against compressed sensing and respiratory-triggered acquisitions. The authors report that zero-shot improves over compressed sensing, and they quantify PSNR on two retrospectively undersampled breath-hold acquisitions against a CS reconstruction of a 40-s R=6 acquisition. The abstract claims image quality comparable to successful respiratory-triggered acquisitions, but the Results section explicitly states that no breath-hold acquisition reaches that quality. The paper also contains a title and abstract that are inconsistent with the body: the manuscript header describes an explainable-AI pathology system, while the full text presents the MRCP study.

Significance. If the central claims were established, the proposed partial-trainable zero-shot strategy could be a useful step toward practical breath-hold MRCP with high acceleration and reduced reconstruction time. The idea of freezing early unrolled stages to lower backpropagation depth and cache intermediate k-space is a reasonable technical contribution, and the qualitative figures suggest zero-shot outperforms compressed sensing at R=25. However, the contradiction between the abstract and the Results, the use of a non-ground-truth PSNR reference on only two subjects, and the selection bias in the triggered acquisitions all prevent the paper from supporting its stated comparability claim. The technical idea is potentially salvageable, but the current manuscript is not coherent enough for publication.

major comments (5)
  1. [Title/Abstract vs. Full Text] The manuscript is internally inconsistent at the highest level. The title and the abstract printed before the full text describe a vision-language-model explanation system for computational pathology, while the body is an MRCP reconstruction study with a different title ('Zero-shot self-supervised learning of single breath-hold magnetic resonance cholangiopancreatography (MRCP) reconstruction'). A journal submission must have a title, abstract, and body that refer to the same work. This alone is a blocking defect.
  2. [Abstract vs. Results, Figure 7] The abstract states 'achieved image quality comparable to that of successful respiratory-triggered acquisitions with regular breathing patterns.' Yet Results (Figure 7) states: 'With the exception of residual motion artifacts ... no breath-hold acquisition reaches the image quality of the (successful) triggered acquisition.' The Discussion repeats this. These statements are directly contradictory, and the clinical-translation rationale depends on the comparability claim. The abstract must be corrected to match the actual findings, or the finding must be supported by evidence it currently lacks.
  3. [Methods, Quantitative analysis; Results, Figure 4] The PSNR reference is not a fully sampled ground truth. The 40-s R=6 breath-hold is reconstructed with ℓ1-wavelet compressed sensing, and PSNR is computed against that reconstruction after retrospective undersampling to R=25. This measures agreement with a CS reconstruction, not true image fidelity. Moreover, these numbers are reported for only two volunteers. With n=2 and a non-ground-truth reference, the quantitative claim that 'PSNR decreased only slightly' from 38.25 to 37.67 dB cannot establish image-quality equivalence, and any small differences are not statistically meaningful.
  4. [Methods, Data] The respiratory-triggered reference scans were acquired by repeatedly scanning volunteers until a 'sufficiently regular breathing pattern' was obtained. This selection process biases the reference toward unusually high quality and does not represent the typical or clinical distribution of triggered acquisitions. Using such a cherry-picked reference as the benchmark makes the stated 'comparability' claim even more difficult to support, and the comparison is not fair to the breath-hold method. The paper should report the success rate and the range of triggered image qualities, not only the best cases.
  5. [Throughout Results and Methods] Many numeric values are missing or unreadable in the manuscript text: e.g., reconstruction times in Table 2 and Figure 4, PSNR values in Figure 6, the mean volunteer age in the Data section, and the equations in Methods. These placeholders make the central quantitative claims unverifiable. I understand some of this may be a typesetting artifact, but as submitted, the paper does not allow a reader to check the reported speedups (6.7-fold, 4.3-fold) or the PSNR trade-offs.
minor comments (5)
  1. [Methods, Conventional reconstructions] The compressed-sensing regularization parameter is set to 0.008 because it 'provided visually optimal reconstructions.' This is an ad hoc free parameter; a sensitivity analysis or a systematic selection criterion (e.g., L-curve) should be reported to show the comparison is not biased by an unfavorable CS setting.
  2. [Methods, Zero-shot learning] The mask-split ratios Γ, Θ, and Θ_v and the number of unrolled stages (13) are introduced without justification or sensitivity analysis. Since these are free parameters of the method, the authors should explain how they were chosen and whether results are robust to them.
  3. [Methods, Data] The learning rate was reduced to 0.0001 for volunteer #11 due to instability. This subject-specific tuning is a form of peeking at the test case; its effect on the reported image quality and time should be discussed transparently.
  4. [Table 2] The note contains a typo: 'initizliaed' should be 'initialized.'
  5. [Discussion] The paper states 'the image quality assessment was based solely on visual inspection by the authors, without any involvement of radiologists.' This is a significant limitation for a clinical-imaging claim and should be stated earlier, ideally in the Methods.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: central claims rest on external comparisons (compressed sensing, triggered acquisitions) and on ablations against random initialization; no equation reduces to a fitted value.

full rationale

This is an empirical imaging study, not a derivation chain. The zero-shot self-supervised loss (Eq. 3) is adopted from Yaman et al. (ref 29) and used as an external algorithmic starting point; the paper does not claim to derive it from its own outputs. The partially trainable strategy is evaluated by comparing frozen/trainable configurations (0/13 vs 12/1) with both random and pretrained initialization, and the reported speedup (up to 6.7-fold) is measured, not derived. The pretrained backbone comes from the authors' prior work (ref 27), but it is not load-bearing: random initialization also yields the qualitative and quantitative conclusions, and the prior work is an independent, externally published source rather than an unverified self-citation invoked to forbid alternatives. The quantitative reference for PSNR is a compressed-sensing reconstruction of a 40 s R=6 breath-hold acquisition, retrospectively undersampled to R=25, rather than a fully sampled ground truth; this is a validity limitation, not circularity, because the zero-shot model never sees the R=6 data and the reference is not constructed from the model's own fitted parameters. The abstract's 'comparable to successful respiratory-triggered acquisitions' statement conflicts with the Results statement that no breath-hold acquisition reaches the image quality of the successful triggered acquisition; that is an internal consistency/correctness concern, not a circularity of the derivation. No self-definitional step, fitted-input-called-prediction, uniqueness import, ansatz-smuggling, or renaming of a known result was found. Score 0.

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

The paper has no invented physical entities. It uses standard Fourier/ESPIRiT machinery, plus domain assumptions about reference image validity and single-subject convergence. Free parameters include the CS regularization weight, the learning rate schedule (which had to be retuned for one volunteer), the number of unrolled stages, and the mask-splitting ratios, several of which are unreadable in the provided text.

free parameters (4)
  • Compressed-sensing regularization parameter = 0.008
    Set because it 'provided visually optimal reconstructions' (Methods, Conventional reconstructions).
  • Learning rate = 0.0003 (0.0001 for volunteer #11)
    Default 0.0003; reduced to 0.0001 for subject #11 due to optimization instability (Methods, Discussion).
  • Number of unrolled stages = 13
    Decided for all experiments; the trade-off curve only scans frozen/trainable splits of this fixed total (Methods, Numerical experiments).
  • Training mask split ratios (Gamma, Theta, Theta_v) = not readable in provided text
    The split ratios for the zero-shot subsets are garbled, so this parameter cannot be audited from the manuscript.
assumptions (4)
  • domain assumption Zero-shot self-supervised training on a single acquisition converges to a good reconstruction when combined with a pretrained SSDU backbone.
    The method adopts Yaman et al.'s zero-shot framework (Ref 29) and assumes the split of k-space into training/loss/validation subsets is sufficient; no convergence guarantee is given.
  • domain assumption A compressed-sensing reconstruction of a 40s R=6 breath-hold is a valid PSNR reference for the R=25 target.
    Used as quantitative ground truth because fully sampled data is infeasible; this is stated but not validated.
  • domain assumption Respiratory-triggered MRCP with a regular breathing pattern is a valid image-quality surrogate for diagnostic quality.
    Used as visual reference; scans were repeated until regular breathing was achieved, potentially biasing the reference.
  • domain assumption Decoupling the readout direction and reconstructing 2D phase-encode planes independently does not lose relevant 3D information.
    The method assumes pixel correlation only in phase-encode directions; discontinuities between readout planes are acknowledged as artifacts.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Explainable to Explained AI: Ideas for Falsifying and Quantifying Explanations." pith.science (2026). https://pith.science/paper/73EI4TXS

@misc{pith2026250809205,
  author       = {Pith},
  title        = {Pith review of: From Explainable to Explained AI: Ideas for Falsifying and Quantifying Explanations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/73EI4TXS}},
  note         = {Machine review of arXiv:2508.09205}
}
read the original abstract

Explaining deep learning models is essential for clinical integration of medical image analysis systems. A good explanation highlights if a model depends on spurious features that undermines generalization and harms a subset of patients or, conversely, may present novel biological insights. Although techniques like GradCAM can identify influential features, they are measurement tools that do not themselves form an explanation. We propose a human-machine-VLM interaction system tailored to explaining classifiers in computational pathology, including multi-instance learning for whole-slide images. Our proof of concept comprises (1) an AI-integrated slide viewer to run sliding-window experiments to test claims of an explanation, and (2) quantification of an explanation's predictiveness using general-purpose vision-language models. The results demonstrate that this allows us to qualitatively test claims of explanations and can quantifiably distinguish competing explanations. This offers a practical path from explainable AI to explained AI in digital pathology and beyond. Code and prompts are available at https://github.com/nki-ai/x2x.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

48 extracted references · 40 canonical work pages

  1. [1]

    Magnetic Resonance Cholangiopancreatography

    Barish MA, Ferrucci JT. Magnetic Resonance Cholangiopancreatography. ���������������� ������������������������������������������������������. 1999;341(4):258-264. doi:10.1056/NEJM199907223410407

    work page doi:10.1056/nejm199907223410407 1999

  2. [2]

    MR Cholangiography: Technical Advances and Clinical Applications

    Fulcher AS, Turner MA, Capps GW. MR Cholangiography: Technical Advances and Clinical Applications. ��������������������������. 1999;19(1):25-44. doi:10.1148/radiographics.19.1.g99ja0525

    work page doi:10.1148/radiographics.19.1.g99ja0525 1999

  3. [3]

    A systematic review and economic evaluation of magnetic resonance cholangiopancreatography compared with diagnostic endoscopic retrograde cholangiopancreatography

    Kaltenthaler E, Bravo Vergel Y , Chilcott J, et al. A systematic review and economic evaluation of magnetic resonance cholangiopancreatography compared with diagnostic endoscopic retrograde cholangiopancreatography. ������������������������������������������. 2004;8(10). doi:10.3310/hta8100

    work page doi:10.3310/hta8100 2004

  4. [4]

    Optimal MR cholangiopancreatographic sequence and its clinical application

    Irie H, Honda H, Tajima T, et al. Optimal MR cholangiopancreatographic sequence and its clinical application. ������������������. 1998;206(2):379-387. doi:10.1148/radiology.206.2.9457189

    work page doi:10.1148/radiology.206.2.9457189 1998

  5. [5]

    Yoon LS, Catalano OA, Fritz S, et al. Another Dimension in Magnetic Resonance Cholangiopancreatography: Comparison of 2- and 3-Dimensional Magnetic Resonance Cholangiopancreatography for the Evaluation of Intraductal Papillary Mucinous Neoplasm of the Pancreas. ��������������������������������������������. 2009;33(3):363-368. doi:10.1097/RCT.0b013e3181852193

    work page doi:10.1097/rct.0b013e3181852193 2009

  6. [6]

    Usefulness of the SPACE pulse sequence at 1.5T MR cholangiography: Comparison of image quality and image acquisition time with conventional 3D��TSE sequence

    Nakaura T, Kidoh M, Maruyama N, et al. Usefulness of the SPACE pulse sequence at 1.5T MR cholangiography: Comparison of image quality and image acquisition time with conventional 3D��TSE sequence. ����������������������������������������������������. 2013;38(5):1014-1019. doi:10.1002/jmri.24033

    work page doi:10.1002/jmri.24033 2013

  7. [7]

    3D T2-weighted and Gd-EOB-DTPA-enhanced 3D T1-weighted MR cholangiography for evaluation of biliary anatomy in living liver donors

    Cai L, Yeh BM, Westphalen AC, et al. 3D T2-weighted and Gd-EOB-DTPA-enhanced 3D T1-weighted MR cholangiography for evaluation of biliary anatomy in living liver donors. ������������������������. 2017;42(3):842-850. doi:10.1007/s00261-016-0936-z

    work page doi:10.1007/s00261-016-0936-z 2017

  8. [8]

    Clinical Feasibility of 3-Dimensional Magnetic Resonance Cholangiopancreatography Using Compressed Sensing: Comparison of Image Quality and Diagnostic Performance

    Yoon JH, Lee SM, Kang HJ, et al. Clinical Feasibility of 3-Dimensional Magnetic Resonance Cholangiopancreatography Using Compressed Sensing: Comparison of Image Quality and Diagnostic Performance. ��������������������������. 2017;52(10):612-619. doi:10.1097/RLI.0000000000000380

    work page doi:10.1097/rli.0000000000000380 2017

Show all 48 references
  1. [9]

    Navigator��triggered prospective acquisition correction (PACE) technique vs

    Morita S, Ueno E, Suzuki K, et al. Navigator��triggered prospective acquisition correction (PACE) technique vs. conventional respiratory��triggered technique for free��breathing 3D MRCP: An initial prospective comparative study using healthy volunteers. ������������������ ����...

    2008 doi

  2. [10]

    Breath-held MR Cholangiopancreatography (MRCP) using a 3D Dixon fat–water separated balanced steady state free precession sequence

    Glockner JF, Saranathan M, Bayram E, et al. Breath-held MR Cholangiopancreatography (MRCP) using a 3D Dixon fat–water separated balanced steady state free precession sequence. ����������������������������������������������������. 2013;31(8):1263-

    2013

  3. [11]

    MR cholangiopancreatography at 3.0 T in children: diagnostic quality and ability in assessment of common paediatric pancreatobiliary pathology

    Almehdar A, Chavhan GB. MR cholangiopancreatography at 3.0 T in children: diagnostic quality and ability in assessment of common paediatric pancreatobiliary pathology. ������. 2013;86(1025):20130036. doi:10.1259/bjr.20130036

    2013 doi

  4. [12]

    Possible Biliary Disease: Diagnostic Performance of High-Spatial-Resolution Isotropic 3D T2-weighted MRCP

    Nandalur KR, Hussain HK, Weadock WJ, et al. Possible Biliary Disease: Diagnostic Performance of High-Spatial-Resolution Isotropic 3D T2-weighted MRCP . ������������������. 2008;249(3):883-890. doi:10.1148/radiol.2493080389

    2008 doi

  5. [13]

    Breath-holding 3D MRCP: the time is now? ��������������������

    Zins M. Breath-holding 3D MRCP: the time is now? ��������������������. 2018;28(9):3719-3720. doi:10.1007/s00330-018-5550-8

    2018 doi

  6. [15]

    Magnetic resonance cholangiopancreatography with GRASE sequence at 3.0T: does it improve image quality and acquisition time as compared with 3D TSE? ��������������������

    Yoshida M, Nakaura T, Inoue T, et al. Magnetic resonance cholangiopancreatography with GRASE sequence at 3.0T: does it improve image quality and acquisition time as compared with 3D TSE? ��������������������. 2018;28(6):2436-2443. doi:10.1007/s00330-017- 5240-y

    2018 doi

  7. [16]

    Rapid 3D navigator-triggered MR cholangiopancreatography with SPACE sequence at 3T: only one-third acquisition time of conventional 3D SPACE navigator-triggered MRCP

    Sun B, Chen Z, Duan Q, et al. Rapid 3D navigator-triggered MR cholangiopancreatography with SPACE sequence at 3T: only one-third acquisition time of conventional 3D SPACE navigator-triggered MRCP. ������������������������. 2020;45(1):134-

    2020

  8. [17]

    Three-dimensional MR Cholangiopancreatography in a Breath Hold with Sparsity-based Reconstruction of Highly Undersampled Data

    Chandarana H, Doshi AM, Shanbhogue A, et al. Three-dimensional MR Cholangiopancreatography in a Breath Hold with Sparsity-based Reconstruction of Highly Undersampled Data. ������������������. 2016;280(2):585-594. doi:10.1148/radiol.2016151935

    2016 doi

  9. [18]

    Compressed-Sensing Accelerated 3-Dimensional Magnetic Resonance Cholangiopancreatography: Application in Suspected Pancreatic Diseases

    Zhu L, Wu X, Sun Z, et al. Compressed-Sensing Accelerated 3-Dimensional Magnetic Resonance Cholangiopancreatography: Application in Suspected Pancreatic Diseases. ��������������������������. 2018;53(3):150-157. doi:10.1097/RLI.0000000000000421

    2018 doi

  10. [19]

    Zhu L, Xue H, Sun Z, et al. Modified breath-hold compressed-sensing 3D MR cholangiopancreatography with a small field-of-view and high resolution acquisition: Clinical feasibility in biliary and pancreatic disorders. ������������������������������������������������������������...

    2018 doi

  11. [20]

    Patient-adapted respiratory training: Effect on navigator-triggered 3D MRCP in painful pancreatobiliary disorders

    Zhu L, Sun Z yong, Xue H dan, et al. Patient-adapted respiratory training: Effect on navigator-triggered 3D MRCP in painful pancreatobiliary disorders. �������������������������������������� ��������������. 2018;45:43-50. doi:10.1016/j.mri.2017.09.014

    2018 doi

  12. [21]

    Sparse MRI: The application of compressed sensing for rapid MR imaging

    Lustig M, Donoho D, Pauly JM. Sparse MRI: The application of compressed sensing for rapid MR imaging. ��������������������������������������������������. 2007;58(6):1182-1195. doi:10.1002/mrm.21391

    2007 doi

  13. [22]

    Assessment of reproducibility and stability of different breath-hold maneuvres by dynamic MRI: comparison between healthy adults 23 and patients with pulmonary hypertension

    Plathow C, Ley S, Zaporozhan J, et al. Assessment of reproducibility and stability of different breath-hold maneuvres by dynamic MRI: comparison between healthy adults 23 and patients with pulmonary hypertension. ��������������������. 2006;16(1):173-179. doi:10.1007/s00330-005-2795-9

    2006 doi

  14. [23]

    Breath-hold and free-breathing quantitative assessment of biventricular volume and function using compressed SENSE: a clinical validation in children and young adults

    Kocaoglu M, Pednekar AS, Wang H, et al. Breath-hold and free-breathing quantitative assessment of biventricular volume and function using compressed SENSE: a clinical validation in children and young adults. ���������������������������������������������������������������������...

    2020 doi

  15. [24]

    Combination of compressed sensing and parallel imaging for highly accelerated first-pass cardiac perfusion MRI

    Otazo R, Kim D, Axel L, et al. Combination of compressed sensing and parallel imaging for highly accelerated first-pass cardiac perfusion MRI. �������������������������������������������� ����������������. 2010;64(3):767-776. doi:10.1002/mrm.22463

    2010 doi

  16. [25]

    Accelerated T1ρ acquisition for knee cartilage quantification using compressed sensing and data-driven parallel imaging: A feasibility study

    Pandit P , Rivoire J, King K, et al. Accelerated T1ρ acquisition for knee cartilage quantification using compressed sensing and data-driven parallel imaging: A feasibility study. ������������������������������������������������������������. 2016;75(3):1256-1261. doi:10.1002/mrm.25702

    2016 doi

  17. [26]

    MoDL: Model-Based Deep Learning Architecture for Inverse Problems

    Aggarwal HK, Mani MP , Jacob M. MoDL: Model-Based Deep Learning Architecture for Inverse Problems. ��������������������������������������������. 2019;38(2):394-405. doi:10.1109/TMI.2018.2865356

    2019

  18. [27]

    Deep Learning��Based Accelerated MR Cholangiopancreatography Without Fully��Sampled Data

    Kim J, Nickel MD, Knoll F. Deep Learning��Based Accelerated MR Cholangiopancreatography Without Fully��Sampled Data. ������������������������������������. 2025;38(3):e70002. doi:10.1002/nbm.70002

    2025 doi

  19. [28]

    Self��supervised learning of physics��guided reconstruction neural networks without fully sampled reference data

    Yaman B, Hosseini SAH, Moeller S, et al. Self��supervised learning of physics��guided reconstruction neural networks without fully sampled reference data. ������������������ ��������������������������������. 2020;84(6):3172-3191. doi:10.1002/mrm.28378

    2020 doi

  20. [29]

    Zero-Shot Self-Supervised Learning for MRI Reconstruction

    Yaman B, Hosseini SAH, Akcakaya M. Zero-Shot Self-Supervised Learning for MRI Reconstruction. In: ��������������������������������������������������������������������������������������������������������. 2022. https://openreview.net/forum?id=085y6YPaYjP

    2022

  21. [30]

    Scan��specific robust artificial��neural�� networks for k��space interpolation (RAKI) reconstruction: Database��free deep learning for fast imaging

    Akçakaya M, Moeller S, Weingärtner S, et al. Scan��specific robust artificial��neural�� networks for k��space interpolation (RAKI) reconstruction: Database��free deep learning for fast imaging. ��������������������������������������������������. 2019;81(1):439-453. doi:10.1002...

    2019 doi

  22. [31]

    Residual RAKI: A hybrid linear and non-linear approach for scan-specific k-space deep learning

    Zhang C, Moeller S, Demirel OB, et al. Residual RAKI: A hybrid linear and non-linear approach for scan-specific k-space deep learning. ��������������������. 2022;256:119248. doi:10.1016/j.neuroimage.2022.119248

    2022

  23. [32]

    Deep Image Prior

    Lempitsky V, Vedaldi A, Ulyanov D. Deep Image Prior. In: �������������������������������������������������� ������������������������������������������������������������������������������������. 2018:9446-9454. doi:10.1109/CVPR.2018.00984

    2018

  24. [33]

    Implicit neural representations with periodic activation functions

    Sitzmann V, Martel JNP , Bergman AW, et al. Implicit neural representations with periodic activation functions. In: ���������������������������������������������������������������������������������������������������������������������� ������������������������������������������...

    2020

  25. [34]

    Improved Multi-shot Diffusion-Weighted MRI with Zero-Shot Self-supervised Learning Reconstruction

    Cho J, Jun Y, Wang X, et al. Improved Multi-shot Diffusion-Weighted MRI with Zero-Shot Self-supervised Learning Reconstruction. In: Greenspan H, Madabhushi A, Mousavi P , et al., eds. ���������������������������������������������������������������������������������������������...

    2023

  26. [35]

    Lichy MP, Wietek BM, Mugler JP, et al. Magnetic Resonance Imaging of the Body Trunk Using a Single-Slab, 3-Dimensional, T2-weighted Turbo-Spin-Echo Sequence With High Sampling Efficiency (SPACE) for High Spatial Resolution Imaging: Initial Clinical Experiences. ���������������...

    2005

  27. [36]

    ISMRM Raw data format: A proposed standard for MRI raw datasets

    Inati SJ, Naegele JD, Zwart NR, et al. ISMRM Raw data format: A proposed standard for MRI raw datasets. ��������������������������������������������������. 2017;77(1):411-421. doi:10.1002/mrm.26089

    2017 doi

  28. [37]

    Python port of mapVBVD

    Clarke WT. Python port of mapVBVD. Accessed July 2, 2025. https://github.com/wtclarke/pymapvbvd

    2025

  29. [38]

    ESPIRiT—an eigenvalue approach to autocalibrating parallel MRI: Where SENSE meets GRAPPA

    Uecker M, Lai P, Murphy MJ, et al. ESPIRiT—an eigenvalue approach to autocalibrating parallel MRI: Where SENSE meets GRAPPA. ��������������������������������������������������. 2014;71(3):990-1001. doi:10.1002/mrm.24751

    2014 doi

  30. [39]

    SigPy: A Python Package for High Performance Iterative Reconstruction

    Ong F, Lustig M. SigPy: A Python Package for High Performance Iterative Reconstruction. In: ������������������������������������������������������������������������. 2019. https://cds.ismrm.org/protected/19MProceedings/PDFfiles/4819.html

    2019

  31. [40]

    Deep Residual Learning for Image Recognition

    He K, Zhang X, Ren S, et al. Deep Residual Learning for Image Recognition. In: ���������� ����������������������������������������������������������������������������������������������������������������������������������. IEEE; 2016:770-778. doi:10.1109/CVPR.2016.90

    2016 doi

  32. [41]

    SGDR: STOCHASTIC GRADIENT DESCENT WITH WARM RESTARTS

    Loshchilov I, Hutter F. SGDR: STOCHASTIC GRADIENT DESCENT WITH WARM RESTARTS. In: ������������������. 2017. doi:10.48550/arXiv.1608.03983

    2017 doi arXiv

  33. [42]

    Generalized autocalibrating partially parallel acquisitions (GRAPPA)

    Griswold MA, Jakob PM, Heidemann RM, et al. Generalized autocalibrating partially parallel acquisitions (GRAPPA). ��������������������������������������������������. 2002;47(6):1202-1210. doi:10.1002/mrm.10171

    2002 doi

  34. [43]

    Python implementations of GRAPPA-like algorithms

    McKibben N. Python implementations of GRAPPA-like algorithms. Accessed July 3,

  35. [44]

    Patients’ experiences in magnetic resonance imaging (MRI) and their experiences of breath holding techniques

    Funk E, Thunberg P , Anderzen-Carlsson A. Patients’ experiences in magnetic resonance imaging (MRI) and their experiences of breath holding techniques. ���������������������������������������� ��������������. 2014;70(8):1880-1890. doi:10.1111/jan.12351

    2014 doi

  36. [45]

    Robust partial Fourier reconstruction for diffusion��weighted imaging using a recurrent convolutional neural network

    Gadjimuradov F, Benkert T, Nickel MD, et al. Robust partial Fourier reconstruction for diffusion��weighted imaging using a recurrent convolutional neural network. ������������������ ��������������������������������. 2022;87(4):2018-2033. doi:10.1002/mrm.29100

    2022 doi

  37. [140]

    doi:10.1007/s00261-019-02342-3

    doi

  38. [466]

    doi:10.1007/978-3-031-43907-0_44

    doi

  39. [1270]

    doi:10.1016/j.mri.2013.06.008 22

    2013 doi

  40. [2025]

    https://github.com/mckib2/pygrappa

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.