pith. sign in

arxiv: 2409.14204 · v4 · submitted 2024-09-21 · 📡 eess.IV · cs.CV

A Unified Deep Learning Framework for Motion Correction in Medical Imaging

Jian Wang , Razieh Faghihpirayesh , Danny Joca , Polina Golland , Ali Gholipour This is my paper

Pith reviewed 2026-05-23 20:34 UTC · model grok-4.3

classification 📡 eess.IV cs.CV
keywords motion correctionmedical image registrationdeep learningcross-modality generalizationrigid motiondeformation correctionfetal MRIunified framework
0
0 comments X

The pith

UniMo corrects motion in medical images from multiple modalities after training once on fetal MRI.

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

The paper presents UniMo as a single deep learning model that handles both global rigid motion and local deformations in medical scans. It trains an equivariant network alongside an encoder-decoder using an alternating optimization scheme and a geometric deformation augmenter that supplies both robustness and extra training data. The model combines intensity and shape information so that the same weights work on new modalities without retraining or fine-tuning. This setup is tested by first training on fetal MRI then applying the fixed model to lung CT, BraTS brain scans, and MedMNIST data. If the approach holds, clinics could maintain one motion-correction network instead of separate ones for each scanner type or body part.

Core claim

UniMo trains once on fetal magnetic resonance images and then, without any retraining, corrects motion in lung CT, brain MRI from BraTS, and other datasets from MedMNIST. The framework alternates between an equivariant neural network that removes global rigid motion and an encoder-decoder that removes local deformations, guided by a unified loss. A geometric deformation augmenter improves global correction by simulating local effects during training and supplies augmented examples. Hybrid use of image intensities and shapes produces stable performance when appearance changes across modalities.

What carries the argument

Alternating optimization of an integrated equivariant network for rigid motion and encoder-decoder for local deformations, driven by a geometric deformation augmenter and hybrid intensity-shape loss.

If this is right

  • The same trained weights achieve higher accuracy than prior motion-correction methods on the tested unseen datasets.
  • One training run on fetal MRI suffices for stable inference on lung CT and brain MRI without fine-tuning.
  • The model handles mixtures of bulk rigid motion and local deformations in a single forward pass.
  • Hybrid intensity and shape features support robustness when image contrast varies across scanners.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Hospitals could reduce the number of separate motion-correction models they maintain if the cross-modality claim generalizes further.
  • The geometric augmenter may be the main driver of stability; an ablation that removes it would test whether the hybrid loss alone is sufficient.
  • Real-time clinical streams with continuous motion could serve as a natural next test case for the same fixed weights.

Load-bearing premise

The specific combination of equivariant network, encoder-decoder, geometric augmenter, and intensity-shape features produces reliable cross-modality generalization without retraining.

What would settle it

Apply the trained UniMo model to a new imaging modality or motion range that differs markedly in appearance statistics and measure whether registration accuracy drops below the level of modality-specific baselines.

Figures

Figures reproduced from arXiv: 2409.14204 by Ali Gholipour, Danny Joca, Jian Wang, Polina Golland, Razieh Faghihpirayesh.

Figure 1
Figure 1. Figure 1: An illustration of the network architecture of our proposed motion correction learning framework, UniMO. [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Two examples of heat maps of TSNR estimated from all motion correction models over [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Statistical results for both transnational and angular errors of all models on single modality. Left: motion correction performance reported over [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Two case studies (left and right half) serve as visualizations of motion tracking results, with the ”target” fetal brains highlighted by red contours for [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Statistical results of Dice comparison on real fMRIs with unknown motions. Left: Motion correction performance with different degrees of motions, [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Motion correction comparison across multiple image modalities for all models. The image modalities from top to bottom are: T2 MRI scans of brains [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: From left to right: transnational and angular errors of motion correction of four image modalities, epoch number of best model performance from [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
read the original abstract

Deep learning has shown significant value in medical image registration for motion correction, however, current techniques are either limited by the type and range of motion they can handle, or require iterative inference and/or retraining for new imaging data. To address these limitations, we introduce UniMo, a Unified Motion Correction framework that leverages deep neural networks to correct for various types of motion in medical imaging. UniMo exploits an alternating optimization scheme for a unified loss function to train an integrated model of 1) an equivariant neural network for global rigid motion correction and 2) an encoder-decoder network to correct local deformations. It features a geometric deformation augmenter that 1) enhances the robustness of global motion correction by addressing any local deformations, and 2) generates augmented data to improve the training process. UniMo is a hybrid model that uses both image intensities and shapes to achieve robust performance amid image appearance variations, and, therefore, it generalizes well to various medical imaging modalities without a need for network retraining. We trained and tested UniMo to track motion in fetal magnetic resonance imaging. Then we tested the trained model, without retraining, on various image modalities from three public datasets, including MedMNIST, lung CT, and BraTS. The results show that UniMo surpassed existing motion correction methods in terms of accuracy, and, notably, it enabled one-time training on a single modality while maintaining high stability and adaptability for inference across multiple unseen imaging datasets. By offering a unified solution, UniMo marks a significant advantage in challenging applications with a mixture of bulk motion and local deformations. https://github.com/IntelligentImaging/UNIMO

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces UniMo, a unified deep learning framework for motion correction in medical imaging. It integrates an equivariant neural network for global rigid motion correction with an encoder-decoder network for local deformations, trained via alternating optimization on a unified loss with a geometric deformation augmenter and hybrid intensity-shape features. The model is trained on fetal MRI and tested without retraining on MedMNIST, lung CT, and BraTS datasets, with the central claim that it achieves superior accuracy and cross-modality generalization.

Significance. If the cross-modality generalization claim is substantiated with quantitative evidence, UniMo would offer a notable advance by enabling single-modality training for robust inference across diverse imaging types, addressing limitations of modality-specific or iterative methods in medical registration.

major comments (2)
  1. [Abstract] Abstract: The assertion that UniMo 'surpassed existing motion correction methods in terms of accuracy' and 'enabled one-time training on a single modality while maintaining high stability and adaptability for inference across multiple unseen imaging datasets' is load-bearing for the central claim yet is unsupported by any quantitative metrics (e.g., registration errors, Dice scores, error bars, or dataset sizes) or ablation results on the test modalities.
  2. [Methods] Methods (description of framework): The alternating optimization scheme and unified loss function are presented without explicit equations defining the loss terms, the integration of equivariant and encoder-decoder components, or how the hybrid intensity-shape features are extracted and combined; this prevents verification of whether the claimed modality-agnostic behavior follows from the architecture or requires unstated preprocessing.
minor comments (1)
  1. The GitHub link is provided but the manuscript does not specify which code components (e.g., augmenter implementation or training scripts) are released or how to reproduce the fetal MRI training and cross-dataset testing.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback. We address each major comment below, agreeing that the abstract and methods sections can be strengthened with additional quantitative details and explicit formulations. We propose revisions accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The assertion that UniMo 'surpassed existing motion correction methods in terms of accuracy' and 'enabled one-time training on a single modality while maintaining high stability and adaptability for inference across multiple unseen imaging datasets' is load-bearing for the central claim yet is unsupported by any quantitative metrics (e.g., registration errors, Dice scores, error bars, or dataset sizes) or ablation results on the test modalities.

    Authors: We agree that the abstract would benefit from explicit quantitative support for these claims. In the revised version, we will incorporate key metrics from our experiments, including registration errors, Dice scores with error bars, and dataset sizes for the fetal MRI training and the MedMNIST, lung CT, and BraTS test sets. This will directly substantiate the accuracy and cross-modality generalization results. revision: yes

  2. Referee: [Methods] Methods (description of framework): The alternating optimization scheme and unified loss function are presented without explicit equations defining the loss terms, the integration of equivariant and encoder-decoder components, or how the hybrid intensity-shape features are extracted and combined; this prevents verification of whether the claimed modality-agnostic behavior follows from the architecture or requires unstated preprocessing.

    Authors: We acknowledge that explicit mathematical definitions would improve verifiability. We will add the equations for the unified loss function, the alternating optimization procedure, the integration between the equivariant network and encoder-decoder, and the extraction/combination of hybrid intensity-shape features. These additions will clarify how the architecture supports modality-agnostic behavior without relying on unstated preprocessing steps. revision: yes

Circularity Check

0 steps flagged

No circularity in derivation chain; empirical framework with external validation

full rationale

The paper presents UniMo as an integrated DL architecture (equivariant network + encoder-decoder + geometric augmenter) trained once on fetal MRI and evaluated zero-shot on MedMNIST, lung CT, and BraTS. No equations, loss functions, or self-citations are shown that reduce any claimed prediction or generalization result to fitted parameters or prior author work by construction. The central claim is an empirical statement about cross-modality performance on public datasets, which is externally falsifiable and does not rely on self-definitional steps or imported uniqueness theorems.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Central claim rests on standard deep-learning training assumptions and the untested premise that the proposed components will transfer across modalities; no explicit free parameters or invented physical entities listed in abstract.

axioms (1)
  • domain assumption Alternating optimization of a unified loss can jointly train rigid and deformable correction networks.
    Invoked in the training scheme description.
invented entities (1)
  • UniMo framework no independent evidence
    purpose: Unified motion correction across modalities
    New proposed integrated model.

pith-pipeline@v0.9.0 · 5844 in / 1197 out tokens · 23652 ms · 2026-05-23T20:34:10.208899+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

60 extracted references · 60 canonical work pages · 1 internal anchor

  1. [1]

    Motion artifacts in mri: A complex problem with many partial solutions,

    M. Zaitsev, J. Maclaren, and M. Herbst, “Motion artifacts in mri: A complex problem with many partial solutions,” Journal of Magnetic Resonance Imaging, vol. 42, no. 4, pp. 887–901, 2015

    work page 2015

  2. [2]

    Prospective motion correction in brain imaging: a review,

    J. Maclaren, M. Herbst, O. Speck, and M. Zaitsev, “Prospective motion correction in brain imaging: a review,” Magnetic resonance in medicine, vol. 69, no. 3, pp. 621–636, 2013

    work page 2013

  3. [3]

    Motion-compensation techniques in neonatal and fetal mr imaging,

    C. Malamateniou et al. , “Motion-compensation techniques in neonatal and fetal mr imaging,” American Journal of Neuroradiology , vol. 34, no. 6, pp. 1124–1136, 2013

    work page 2013

  4. [4]

    Motion estimation and correction in spect, pet and ct,

    A. Z. Kyme and R. R. Fulton, “Motion estimation and correction in spect, pet and ct,” Physics in Medicine & Biology , vol. 66, no. 18, p. 18TR02, 2021

    work page 2021

  5. [5]

    Image registration,

    J. M. Fitzpatrick, D. L. Hill, C. R. Maurer et al., “Image registration,” Handbook of medical imaging , vol. 2, pp. 447–513, 2000

    work page 2000

  6. [6]

    Promo: real-time prospective motion correction in mri using image-based tracking,

    N. White et al. , “Promo: real-time prospective motion correction in mri using image-based tracking,” Magnetic Resonance in Medicine: An Official Journal of the International Society for Magnetic Resonance in Medicine, vol. 63, no. 1, pp. 91–105, 2010

    work page 2010

  7. [7]

    Slice-to-volume medical image registra- tion: A survey,

    E. Ferrante and N. Paragios, “Slice-to-volume medical image registra- tion: A survey,” Medical image analysis , vol. 39, pp. 101–123, 2017

    work page 2017

  8. [8]

    Biomedical image registration,

    A. Gholipour and N. Kehtarnavaz, “Biomedical image registration,” Encyclopedia of Image Processing , pp. 24–39, 2019

    work page 2019

  9. [9]

    Slimm: Slice lo- calization integrated mri monitoring,

    Y . Sui, O. Afacan, A. Gholipour, and S. K. Warfield, “Slimm: Slice lo- calization integrated mri monitoring,” NeuroImage, vol. 223, p. 117280, 2020

    work page 2020

  10. [10]

    Real-time fetal brain tracking for functional fetal mri,

    S. Neves Silva et al., “Real-time fetal brain tracking for functional fetal mri,” Magnetic resonance in medicine , vol. 90, no. 6, pp. 2306–2320, 2023

    work page 2023

  11. [11]

    Deep learning for retrospective motion correction in mri: a comprehensive review,

    V . Spieker et al., “Deep learning for retrospective motion correction in mri: a comprehensive review,” IEEE Transactions on Medical Imaging , 2023

    work page 2023

  12. [12]

    Real-time deep pose estimation with geodesic loss for image-to-template rigid registration,

    S. S. M. Salehi, S. Khan, D. Erdogmus, and A. Gholipour, “Real-time deep pose estimation with geodesic loss for image-to-template rigid registration,” IEEE transactions on medical imaging , vol. 38, no. 2, pp. 470–481, 2018

    work page 2018

  13. [13]

    Keymorph: Robust multi-modal affine registration via unsupervised keypoint de- tection,

    M. Y . Evan, A. Q. Wang, A. V . Dalca, and M. R. Sabuncu, “Keymorph: Robust multi-modal affine registration via unsupervised keypoint de- tection,” in International Conference on Medical Imaging with Deep Learning. PMLR, 2022, pp. 1482–1503

    work page 2022

  14. [14]

    Spaer: Learning spatio-temporal equivariant representations for fetal brain motion tracking,

    J. Wang, R. Faghihpirayesh, P. Golland, and A. Ghoulipour, “Spaer: Learning spatio-temporal equivariant representations for fetal brain motion tracking,” arXiv preprint arXiv:2407.20198 , 2024. 9 Source Target UniMo Equivariant Filter KeyMorph DeepPose Source Target UniMo Equivariant Filter KeyMorph DeepPose BrainsLung CTfMRIMedMN Source Target UniMo Equi...

    work page arXiv 2024

  15. [15]

    Joint motion estimation with geometric deformation correction for fetal echo planar images via deep learning,

    J. Wang, R. Faghihpirayesh, D. Erdogmus, and A. Gholipour, “Joint motion estimation with geometric deformation correction for fetal echo planar images via deep learning,” in Medical Imaging with Deep Learning, 2024

    work page 2024

  16. [16]

    Equivariant filters for efficient tracking in 3d imaging,

    D. Moyer, E. Abaci Turk, P. E. Grant, W. M. Wells, and P. Golland, “Equivariant filters for efficient tracking in 3d imaging,” in Medical Im- age Computing and Computer Assisted Intervention–MICCAI 2021: 24th International Conference, Strasbourg, France, September 27–October 1, 2021, Proceedings, Part IV 24 . Springer, 2021, pp. 193–202

    work page 2021

  17. [17]

    Svort: iterative transformer for slice-to-volume registration in fetal brain mri,

    J. Xu, D. Moyer, P. E. Grant, P. Golland, J. E. Iglesias, and E. Adalsteins- son, “Svort: iterative transformer for slice-to-volume registration in fetal brain mri,” in International Conference on Medical Image Computing and Computer-Assisted Intervention . Springer, 2022, pp. 3–13

    work page 2022

  18. [18]

    Network accelerated motion estimation and reduction (namer): convolutional neural network guided retrospective motion correction using a separable motion model,

    M. W. Haskell et al. , “Network accelerated motion estimation and reduction (namer): convolutional neural network guided retrospective motion correction using a separable motion model,” Magnetic resonance in medicine, vol. 82, no. 4, pp. 1452–1461, 2019

    work page 2019

  19. [19]

    A deep cascade of convolutional neural networks for dynamic mr image reconstruction,

    J. Schlemper, J. Caballero, J. V . Hajnal, A. N. Price, and D. Rueckert, “A deep cascade of convolutional neural networks for dynamic mr image reconstruction,” IEEE transactions on Medical Imaging , vol. 37, no. 2, pp. 491–503, 2017

    work page 2017

  20. [20]

    Deep predictive motion tracking in magnetic resonance imaging: application to fetal imaging,

    A. Singh, S. S. M. Salehi, and A. Gholipour, “Deep predictive motion tracking in magnetic resonance imaging: application to fetal imaging,” IEEE transactions on medical imaging , vol. 39, no. 11, pp. 3523–3534, 2020

    work page 2020

  21. [21]

    Cine cardiac mri motion artifact reduction using a recurrent neural network,

    Q. Lyu et al. , “Cine cardiac mri motion artifact reduction using a recurrent neural network,” IEEE Transactions on Medical Imaging , vol. 40, no. 8, pp. 2170–2181, 2021

    work page 2021

  22. [22]

    Nesvor: implicit neural representation for slice-to-volume reconstruction in mri,

    J. Xu et al., “Nesvor: implicit neural representation for slice-to-volume reconstruction in mri,” IEEE transactions on medical imaging , vol. 42, no. 6, pp. 1707–1719, 2023

    work page 2023

  23. [23]

    Retrospective motion correction of mr images using prior-assisted deep learning,

    S. Chatterjee, A. Sciarra, M. D ¨unnwald, S. Oeltze-Jafra, A. N ¨urnberger, and O. Speck, “Retrospective motion correction of mr images using prior-assisted deep learning,” arXiv preprint arXiv:2011.14134 , 2020

    work page arXiv 2011

  24. [24]

    A knowledge interaction learning for multi-echo mri motion artifact correction towards better enhancement of swi,

    M. A. Al-Masni, S. Lee, A. K. Al-Shamiri, S.-M. Gho, Y . H. Choi, and D.-H. Kim, “A knowledge interaction learning for multi-echo mri motion artifact correction towards better enhancement of swi,” Computers in biology and medicine , vol. 153, p. 106553, 2023

    work page 2023

  25. [25]

    Autofocusing+: noise-resilient motion correction in magnetic resonance imaging,

    E. Kuzmina, A. Razumov, O. Y . Rogov, E. Adalsteinsson, J. White, and D. V . Dylov, “Autofocusing+: noise-resilient motion correction in magnetic resonance imaging,” in International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 2022, pp. 365–375

    work page 2022

  26. [26]

    Deep learning-based motion quantification from k-space for fast model-based magnetic resonance imaging motion cor- rection,

    J. Hossbach et al. , “Deep learning-based motion quantification from k-space for fast model-based magnetic resonance imaging motion cor- rection,” Medical physics, vol. 50, no. 4, pp. 2148–2161, 2023

    work page 2023

  27. [27]

    Physics-informed deep learning for motion-corrected reconstruction of quantitative brain mri,

    H. Eichhorn et al., “Physics-informed deep learning for motion-corrected reconstruction of quantitative brain mri,” in MICCAI, 2024

    work page 2024

  28. [28]

    Joint frequency and image space learning for mri reconstruction and analysis,

    N. M. Singh, J. E. Iglesias, E. Adalsteinsson, A. V . Dalca, and P. Gol- land, “Joint frequency and image space learning for mri reconstruction and analysis,” The journal of machine learning for biomedical imaging , vol. 2022, 2022

    work page 2022

  29. [29]

    Motion artifact reduction for magnetic resonance imaging with deep learning and k-space analysis,

    L. Cui et al., “Motion artifact reduction for magnetic resonance imaging with deep learning and k-space analysis,” PloS one , vol. 18, no. 1, p. e0278668, 2023

    work page 2023

  30. [30]

    Exploiting motion for deep learning reconstruction of extremely- undersampled dynamic mri,

    G. Seegoolam, J. Schlemper, C. Qin, A. Price, J. Hajnal, and D. Rueck- ert, “Exploiting motion for deep learning reconstruction of extremely- undersampled dynamic mri,” in International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 2019, pp. 704–712

    work page 2019

  31. [31]

    End-to-end deep learning nonrigid motion-corrected recon- struction for highly accelerated free-breathing coronary mra,

    H. Qi et al., “End-to-end deep learning nonrigid motion-corrected recon- struction for highly accelerated free-breathing coronary mra,” Magnetic Resonance in Medicine , vol. 86, no. 4, pp. 1983–1996, 2021

    work page 1983

  32. [32]

    Neural implicit k-space for binning-free non-cartesian cardiac mr imaging,

    W. Huang, H. B. Li, J. Pan, G. Cruz, D. Rueckert, and K. Hammernik, “Neural implicit k-space for binning-free non-cartesian cardiac mr imaging,” in International Conference on Information Processing in Medical Imaging. Springer, 2023, pp. 548–560

    work page 2023

  33. [33]

    Iconik: Generating respiratory-resolved abdominal mr reconstructions using neural implicit representations in k-space,

    V . Spieker et al. , “Iconik: Generating respiratory-resolved abdominal mr reconstructions using neural implicit representations in k-space,” in MICCAI. Springer, 2023, pp. 183–192

    work page 2023

  34. [34]

    Two-stage motion correction for super-resolution ultra- sound imaging in human lower limb,

    S. Harput et al., “Two-stage motion correction for super-resolution ultra- sound imaging in human lower limb,” IEEE transactions on ultrasonics, ferroelectrics, and frequency control, vol. 65, no. 5, pp. 803–814, 2018

    work page 2018

  35. [35]

    3d freehand ultrasound without external tracking using deep learning,

    R. Prevost et al., “3d freehand ultrasound without external tracking using deep learning,” Medical image analysis , vol. 48, pp. 187–202, 2018

    work page 2018

  36. [36]

    Automatic inter-frame patient motion correction for dynamic cardiac pet using deep learning,

    L. Shi et al. , “Automatic inter-frame patient motion correction for dynamic cardiac pet using deep learning,” IEEE transactions on medical imaging, vol. 40, no. 12, pp. 3293–3304, 2021

    work page 2021

  37. [37]

    Deep learning based joint pet image reconstruction and motion estimation,

    T. Li, M. Zhang, W. Qi, E. Asma, and J. Qi, “Deep learning based joint pet image reconstruction and motion estimation,” IEEE transactions on medical imaging, vol. 41, no. 5, pp. 1230–1241, 2021

    work page 2021

  38. [38]

    Se (3)-equivariant and noise-invariant 3d motion tracking in medical images,

    B. Billot et al. , “Se (3)-equivariant and noise-invariant 3d motion tracking in medical images,” arXiv preprint arXiv:2312.13534 , 2023

    work page arXiv 2023

  39. [39]

    Computing large deformation metric mappings via geodesic flows of diffeomorphisms,

    M. F. Beg, M. I. Miller, A. Trouv ´e, and L. Younes, “Computing large deformation metric mappings via geodesic flows of diffeomorphisms,” International journal of computer vision , vol. 61, pp. 139–157, 2005

    work page 2005

  40. [40]

    Symmetric diffeomorphic image registration with cross-correlation: evaluating auto- mated labeling of elderly and neurodegenerative brain,

    B. B. Avants, C. L. Epstein, M. Grossman, and J. C. Gee, “Symmetric diffeomorphic image registration with cross-correlation: evaluating auto- mated labeling of elderly and neurodegenerative brain,” Medical image analysis, vol. 12, no. 1, pp. 26–41, 2008

    work page 2008

  41. [41]

    Multi-modal volume registration by maximization of mutual informa- tion,

    W. M. Wells III, P. Viola, H. Atsumi, S. Nakajima, and R. Kikinis, “Multi-modal volume registration by maximization of mutual informa- tion,” Medical image analysis , vol. 1, no. 1, pp. 35–51, 1996

    work page 1996

  42. [42]

    Meta- morph: Learning metamorphic image transformation with appearance changes,

    J. Wang, J. Xing, J. Druzgal, W. M. Wells III, and M. Zhang, “Meta- morph: Learning metamorphic image transformation with appearance changes,” in International Conference on Information Processing in Medical Imaging. Springer, 2023, pp. 576–587

    work page 2023

  43. [43]

    Sur la g ´eom´etrie diff ´erentielle des groupes de Lie de dimension infinie et ses applications `a l’hydrodynamique des fluides parfaits,

    V . I. Arnol’d, “Sur la g ´eom´etrie diff ´erentielle des groupes de Lie de dimension infinie et ses applications `a l’hydrodynamique des fluides parfaits,” Ann. Inst. Fourier, vol. 16, pp. 319–361, 1966

    work page 1966

  44. [44]

    Geodesic shooting for compu- tational anatomy,

    M. I. Miller, A. Trouv ´e, and L. Younes, “Geodesic shooting for compu- tational anatomy,” Journal of mathematical imaging and vision , vol. 24, no. 2, pp. 209–228, 2006

    work page 2006

  45. [45]

    Fast diffeomorphic image registration via fourier-approximated lie algebras,

    M. Zhang and P. T. Fletcher, “Fast diffeomorphic image registration via fourier-approximated lie algebras,” International Journal of Computer Vision, vol. 127, no. 1, pp. 61–73, 2019

    work page 2019

  46. [46]

    3d shape descriptor based on 3d fourier transform,

    D. Vranic and D. Saupe, “3d shape descriptor based on 3d fourier transform,” 2001

    work page 2001

  47. [47]

    Invariant image recognition by zernike moments,

    A. Khotanzad and Y . H. Hong, “Invariant image recognition by zernike moments,” IEEE Transactions on pattern analysis and machine intelli- gence, vol. 12, no. 5, pp. 489–497, 1990

    work page 1990

  48. [48]

    Slerpface: Face template protection via spherical linear interpolation,

    Z. Zhong et al., “Slerpface: Face template protection via spherical linear interpolation,” arXiv preprint arXiv:2407.03043 , 2024

    work page arXiv 2024

  49. [49]

    Spherical linear interpolation and text-anchoring for zero-shot composed image retrieval,

    Y . K. Jang, D. Huynh, A. Shah, W.-K. Chen, and S.-N. Lim, “Spherical linear interpolation and text-anchoring for zero-shot composed image retrieval,” arXiv preprint arXiv:2405.00571 , 2024

    work page arXiv 2024

  50. [50]

    Quatse: Spherical linear interpolation of quaternion for knowledge graph embeddings,

    J. Li, X. Su, X. Ma, and G. Gao, “Quatse: Spherical linear interpolation of quaternion for knowledge graph embeddings,” in CCF International Conference on Natural Language Processing and Chinese Computing . Springer, 2022, pp. 209–220

    work page 2022

  51. [51]

    V oxelmorph: a learning framework for deformable medical image registration,

    G. Balakrishnan, A. Zhao, M. R. Sabuncu, J. Guttag, and A. V . Dalca, “V oxelmorph: a learning framework for deformable medical image registration,” IEEE transactions on medical imaging , vol. 38, no. 8, pp. 1788–1800, 2019

    work page 2019

  52. [52]

    Transmorph: Transformer for unsupervised medical image registration,

    J. Chen, E. C. Frey, Y . He, W. P. Segars, Y . Li, and Y . Du, “Transmorph: Transformer for unsupervised medical image registration,” Medical image analysis, vol. 82, p. 102615, 2022

    work page 2022

  53. [53]

    Diffusemorph: Unsupervised deformable image registration using diffusion model,

    B. Kim, I. Han, and J. C. Ye, “Diffusemorph: Unsupervised deformable image registration using diffusion model,” in European conference on computer vision. Springer, 2022, pp. 347–364

    work page 2022

  54. [54]

    Nocedal and S

    J. Nocedal and S. J. Wright, Numerical optimization. Springer, 1999

    work page 1999

  55. [55]

    Fetal- bet: Brain extraction tool for fetal mri,

    R. Faghihpirayesh, D. Karimi, D. Erdo ˘gmus ¸, and A. Gholipour, “Fetal- bet: Brain extraction tool for fetal mri,” IEEE Open Journal of Engi- neering in Medicine and Biology , 2024

    work page 2024

  56. [56]

    Autosegmentation for thoracic radiation treatment planning: a grand challenge at aapm 2017,

    J. Yang et al. , “Autosegmentation for thoracic radiation treatment planning: a grand challenge at aapm 2017,” Medical physics , vol. 45, no. 10, pp. 4568–4581, 2018

    work page 2017

  57. [57]

    Medmnist classification decathlon: A lightweight automl benchmark for medical image analysis,

    J. Yang, R. Shi, and B. Ni, “Medmnist classification decathlon: A lightweight automl benchmark for medical image analysis,” in 2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI) . IEEE, 2021, pp. 191–195

    work page 2021

  58. [58]

    Medmnist v2-a large-scale lightweight benchmark for 2d and 3d biomedical image classification,

    J. Yang et al., “Medmnist v2-a large-scale lightweight benchmark for 2d and 3d biomedical image classification,” Scientific Data, vol. 10, no. 1, p. 41, 2023

    work page 2023

  59. [59]

    The RSNA-ASNR-MICCAI BraTS 2021 Benchmark on Brain Tumor Segmentation and Radiogenomic Classification

    U. Baid et al. , “The rsna-asnr-miccai brats 2021 benchmark on brain tumor segmentation and radiogenomic classification,” arXiv preprint arXiv:2107.02314, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  60. [60]

    The multimodal brain tumor image segmentation benchmark (brats),

    B. H. Menze et al. , “The multimodal brain tumor image segmentation benchmark (brats),” IEEE transactions on medical imaging , vol. 34, no. 10, pp. 1993–2024, 2014

    work page 1993