pith. machine review for the scientific record. sign in

arxiv: 2605.00916 · v1 · submitted 2026-04-29 · 💻 cs.CV

Recognition: unknown

SAMamba3D: adapting Segment Anything for generalizable 3D segmentation of multiphase pore-scale images

Rui Zhang , Xianzhi Song , Linqi Zhu , Branko Bijeljic , Gensheng Li , Martin J. Blunt
Authors on Pith no claims yet

Pith reviewed 2026-05-09 20:05 UTC · model grok-4.3

classification 💻 cs.CV
keywords 3D segmentationSegment Anything ModelMambapore-scale imagingmultiphase fluidsX-ray CTrock microstructuregeneralizable models
0
0 comments X

The pith

SAMamba3D adapts a largely frozen SAM encoder with Mamba-based volumetric context modeling and progressive cross-scale feature interaction to achieve generalizable 3D segmentation of multiphase pore-scale images.

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

The paper aims to overcome the limitation of current 3D segmentation methods that require retraining or heavy fine-tuning for each new rock type, fluid pattern, scanner, or acquisition condition in multiphase pore-scale X-ray images. It proposes a parameter-efficient adaptation of the Segment Anything Model that keeps most of the original 2D encoder frozen while adding Mamba components for 3D context and cross-scale interactions. This setup is tested on sandstone and carbonate datasets with varying fluids, wettability, and scanning conditions, where it matches or beats existing baselines. If the approach holds, it would support faster, more consistent analysis of large 3D images while retaining accurate measures of fluid saturation, connectivity, and interfaces.

Core claim

SAMamba3D is a parameter-efficient framework that adapts the Segment Anything Model for 3D pore-scale segmentation by coupling its largely frozen encoder with Mamba-based volumetric context modeling and progressive cross-scale feature interaction. For sandstone and carbonate datasets with different fluids, wettability, and scanning conditions, it matches or outperforms current 3D baselines while reducing the need for case-specific retraining. The resulting segmentations preserve physically meaningful descriptors including fluid saturation, connectivity, and interface morphology.

What carries the argument

The SAMamba3D framework, which couples a largely frozen SAM encoder with Mamba-based volumetric context modeling and progressive cross-scale feature interaction to extend 2D boundary priors into generalizable 3D segmentation.

If this is right

  • Matches or outperforms current 3D segmentation baselines on sandstone and carbonate datasets under varied fluid, wettability, and scanning conditions.
  • Reduces the need for case-specific retraining when rock type, fluid pattern, or acquisition conditions change.
  • Produces segmentations that preserve fluid saturation, connectivity, and interface morphology.
  • Supports faster and more reliable analysis of large 3D multiphase images.

Where Pith is reading between the lines

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

  • The same frozen-encoder plus efficient 3D module pattern could be applied to other volumetric imaging domains such as materials or biological samples if the Mamba and cross-scale components transfer well.
  • Minimizing retraining opens the possibility of on-the-fly segmentation during ongoing experiments where scanner settings shift.
  • Additional tests on datasets with greater resolution or noise extremes would clarify the boundary of the claimed generalizability.

Load-bearing premise

Coupling a largely frozen SAM encoder with Mamba-based volumetric context modeling and progressive cross-scale feature interaction produces generalizable 3D segmentations across varying rock types, fluids, and acquisition conditions without extensive retraining.

What would settle it

A new multiphase pore-scale dataset from an unseen rock type, fluid combination, or scanning condition where SAMamba3D segmentation accuracy or physical descriptor preservation falls below that of a fully retrained 3D baseline.

Figures

Figures reproduced from arXiv: 2605.00916 by Branko Bijeljic, Gensheng Li, Linqi Zhu, Martin J. Blunt, Rui Zhang, Xianzhi Song.

Figure 1
Figure 1. Figure 1: Architecture of SAMamba3D. (a) A 3D-adapted SAM image encoder extracts boundary-aware representations [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Accuracy–efficiency trade-off across the seven 3D segmentation models evaluated [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Model performance on test set (Bentheimer, water-wet, see Table 1). A representative 2D cross-section of a [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Model performance on a test dataset (Bentheimer, mixed-wet, see Table 1). A representative 2D cross-section [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Model performance on a test dataset (Estaillades carbonate, oil-wet, see Table 1). A representative 2D [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of the distribution of measured interfacial curvature for the oil–brine interface reconstructed by [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of in situ contact angle distributions for the mixed-wet sample reconstructed by SAMamba3D and the base-case segmentation. (a) SAMamba3D. (b) Base case [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Generalization of SAMamba3D across diverse rock types and wetting conditions. Segmentation results are [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Generalization of SAMamba3D across different fluid systems. Segmentation results on four unseen datasets [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗
read the original abstract

Reliable segmentation of multiphase pore-scale X-ray images of rocks is necessary to quantify fluid saturation, connectivity, and interfacial geometry. However, current 3D segmentation methods are typically dataset-specific, requiring retraining or extensive fine-tuning whenever rock type, fluid pattern, scanner, or acquisition conditions change. Foundation models such as the Segment Anything Model (SAM) provide strong 2D boundary priors, but they are not directly applicable to 3D data. We present SAMamba3D, a parameter-efficient framework that adapts a largely frozen SAM encoder to generalizable 3D pore-scale segmentation by coupling it with Mamba-based volumetric context modeling and progressive cross-scale feature interaction. For sandstone and carbonate datasets, with different fluids, wettability, and scanning conditions, SAMamba3D matches or outperforms current 3D baselines while reducing the need for case-specific retraining. The resulting segmented images preserve physically meaningful descriptors, including fluid saturation, connectivity, and interface morphology, enabling more reliable and rapid analysis of large 3D multiphase images.

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 / 3 minor

Summary. The manuscript introduces SAMamba3D, a parameter-efficient framework that adapts a largely frozen SAM encoder for 3D segmentation of multiphase pore-scale X-ray images by coupling it with Mamba-based volumetric context modeling and progressive cross-scale feature interaction. It claims that, for sandstone and carbonate datasets with varying fluids, wettability, and scanning conditions, the method matches or outperforms existing 3D baselines, reduces the need for case-specific retraining, and preserves physically meaningful descriptors such as fluid saturation, connectivity, and interface morphology.

Significance. If the empirical results and generalization claims hold after clarification of the training protocol, the work would be significant for digital rock physics and porous-media analysis. It addresses a practical bottleneck (dataset-specific retraining) by leveraging 2D foundation models in a 3D setting with modest added parameters, potentially enabling faster, more reliable quantification of multiphase images across diverse acquisition conditions.

major comments (2)
  1. [Methods / Experimental Setup] The central generalization claim (reduced case-specific retraining while matching/outperforming baselines across rock/fluid/scanner variations) is load-bearing but rests on an unverified assumption about the training protocol. The manuscript must explicitly state whether the Mamba volumetric and cross-scale modules were trained once on a mixed or representative set and then applied with minimal change to held-out conditions, or whether separate training runs were performed per dataset/condition. Without this detail (likely in the Methods or Experimental Setup section), the reduction in retraining is not demonstrated beyond standard parameter-efficient fine-tuning.
  2. [Results / Abstract] The abstract asserts performance parity or improvement and preservation of physical descriptors but supplies no quantitative metrics, specific baselines, error analysis, or dataset sizes. This prevents verification of the central claim; the full results section should include tables or figures with Dice/IoU scores, saturation errors, connectivity metrics, and statistical comparisons against at least two current 3D baselines on each rock type.
minor comments (3)
  1. The abstract would be strengthened by including one or two key quantitative results (e.g., average Dice improvement or saturation error) to support the performance claims.
  2. [Figures] Ensure all figure captions explicitly describe the comparison (e.g., which baseline is shown in each panel) and label axes/units for physical descriptors such as saturation or interfacial area.
  3. [Methods] Clarify the exact number of trainable parameters added by the Mamba and cross-scale modules relative to the frozen SAM encoder; this supports the 'parameter-efficient' claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive review and for recognizing the potential significance of SAMamba3D for digital rock physics. We address each major comment below and will revise the manuscript accordingly to strengthen clarity and verifiability.

read point-by-point responses

  1. Referee: [Methods / Experimental Setup] The central generalization claim (reduced case-specific retraining while matching/outperforming baselines across rock/fluid/scanner variations) is load-bearing but rests on an unverified assumption about the training protocol. The manuscript must explicitly state whether the Mamba volumetric and cross-scale modules were trained once on a mixed or representative set and then applied with minimal change to held-out conditions, or whether separate training runs were performed per dataset/condition. Without this detail (likely in the Methods or Experimental Setup section), the reduction in retraining is not demonstrated beyond standard parameter-efficient fine-tuning.

    Authors: We agree that explicit description of the training protocol is necessary to substantiate the generalization claim. In the reported experiments, the Mamba volumetric context modeling and progressive cross-scale feature interaction modules were trained once on a mixed representative set drawn from both sandstone and carbonate images (encompassing variations in fluids, wettability, and scanning conditions). The trained modules were then applied to held-out test conditions with only minimal or no further fine-tuning. This single-training protocol underpins the reduction in case-specific retraining. We will add a dedicated paragraph in the Methods section (and reference it in the Experimental Setup) that clearly states this training procedure, including the composition of the mixed training set and the minimal adaptation applied to held-out data. revision: yes

  2. Referee: [Results / Abstract] The abstract asserts performance parity or improvement and preservation of physical descriptors but supplies no quantitative metrics, specific baselines, error analysis, or dataset sizes. This prevents verification of the central claim; the full results section should include tables or figures with Dice/IoU scores, saturation errors, connectivity metrics, and statistical comparisons against at least two current 3D baselines on each rock type.

    Authors: We acknowledge that the abstract is concise and does not contain numerical values, which is standard practice, but we agree that the full results must enable direct verification. The revised manuscript will expand the Results section with consolidated tables (new Table 2) and accompanying figures that report Dice and IoU scores, fluid saturation errors, connectivity metrics (e.g., Euler characteristic and cluster size distributions), and statistical comparisons (paired t-tests) against at least two 3D baselines (3D U-Net and nnU-Net) for each rock type. Dataset sizes, acquisition parameters, and error analyses will be explicitly tabulated. These additions will be cross-referenced from the abstract and discussion to ensure the central claims are quantitatively supported. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical adaptation of existing models

full rationale

The paper presents SAMamba3D as a parameter-efficient architectural adaptation that couples a frozen SAM encoder with Mamba-based volumetric modeling and cross-scale interaction modules. No equations, derivations, or first-principles results are claimed; performance claims rest on empirical evaluation across sandstone and carbonate datasets with varying conditions. No self-citations are load-bearing for any derivation, no fitted inputs are relabeled as predictions, and no uniqueness theorems or ansatzes are smuggled in. The framework is self-contained as a practical engineering contribution whose validity is tested externally via segmentation metrics rather than by construction from its own inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no specific free parameters, axioms, or invented entities are identifiable from the provided text.

pith-pipeline@v0.9.0 · 5504 in / 1141 out tokens · 41424 ms · 2026-05-09T20:05:38.781678+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

51 extracted references · 5 canonical work pages · 2 internal anchors

  1. [1]

    Bashir, M

    A. Bashir, M. Ali, S. Patil, M. S. Aljawad, M. Mahmoud, D. Al-Shehri, H. Hoteit, and M. S. Kamal. Comprehensive review of CO2 geological storage: Exploring principles, mechanisms, and prospects.Earth-Science Reviews, 249:104672, 2024. 16 Zhang, R., et al

    2024

  2. [2]

    J. M. Miocic, N. Heinemann, K. Edlmann, J. Scafidi, F. Molaei, and J. Alcalde. Underground hydrogen storage: A review.Geological Society, London, Special Publications, 528(1):73–86, 2023

    2023

  3. [3]

    Wildenschild and A

    D. Wildenschild and A. P. Sheppard. X-ray imaging and analysis techniques for quantifying pore-scale structure and processes in subsurface porous medium systems.Advances in Water Resources, 51:217–246, 2013

    2013

  4. [4]

    M. J. Blunt, B. Bijeljic, H. Dong, O. Gharbi, S. Iglauer, P. Mostaghimi, A. Paluszny, and C. Pentland. Pore-scale imaging and modelling.Advances in Water Resources, 51:197–216, 2013

    2013

  5. [5]

    Iassonov, T

    P. Iassonov, T. Gebrenegus, and M. Tuller. Segmentation of X-ray computed tomography images of porous materials: A crucial step for characterization and quantitative analysis of pore structures.Water Resources Research, 45:W09415, 2009

    2009

  6. [6]

    Schlüter, A

    S. Schlüter, A. Sheppard, K. Brown, and D. Wildenschild. Image processing of multiphase images obtained via X-ray microtomography: A review.Water Resources Research, 50(4):3615–3639, 2014

    2014

  7. [7]

    Scanziani, K

    A. Scanziani, K. Singh, M. J. Blunt, and A. Guadagnini. Automatic method for estimation of in situ effective contact angle from X-ray micro tomography images of two-phase flow in porous media.Journal of Colloid and Interface Science, 496:51–59, 2017

    2017

  8. [8]

    Huang, A

    R. Huang, A. L. Herring, and A. Sheppard. Effect of saturation and image resolution on representative elementary volume and topological quantification: An experimental study on Bentheimer sandstone using micro-CT.Transport in Porous Media, 137:489–518, 2021

    2021

  9. [9]

    Andrew, B

    M. Andrew, B. Bijeljic, and M. J. Blunt. Pore-scale contact angle measurements at reservoir conditions using X-ray microtomography.Advances in Water Resources, 68(1):24–31, 2014

    2014

  10. [10]

    AlRatrout, A

    A. AlRatrout, A. Q. Raeini, B. Bijeljic, and M. J. Blunt. Automatic measurement of contact angle in pore-space images.Advances in Water Resources, 109:158–169, 2017

    2017

  11. [11]

    Q. Lin, B. Bijeljic, R. Pini, M. J. Blunt, and S. Krevor. Imaging and measurement of pore-scale interfacial curvature to determine capillary pressure simultaneously with relative permeability.Water Resources Research, 54(9):7046–7060, 2018

    2018

  12. [12]

    Y . Niu, P. Mostaghimi, M. Shabaninejad, P. Swietojanski, and R. T. Armstrong. Digital rock segmentation for petrophysical analysis with reduced user bias using convolutional neural networks.Water Resources Research, 56(2):e2019WR026597, 2020

    2020

  13. [13]

    N. J. Alqahtani, Y . Niu, Y . D. Wang, T. Chung, Z. Lanetc, A. Zhuravljov, R. T. Armstrong, and P. Mostaghimi. Super-resolved segmentation of X-ray images of carbonate rocks using deep learning.Transport in Porous Media, 143(2):497–525, 2022

    2022

  14. [14]

    H. Wang, R. Guo, L. E. Dalton, D. Crandall, S. A. Hosseini, M. Fan, and C. Chen. Comparative assessment of U-net-based deep learning models for segmenting microfractures and pore spaces in digital rocks.SPE Journal, 29(11):5779–5791, 2024

    2024

  15. [15]

    Siavashi, M

    J. Siavashi, M. Mahdaviara, M. J. Shojaei, M. Sharifi, and M. J. Blunt. Segmentation of two-phase flow X-ray tomography images to determine contact angle using deep autoencoders.Energy, 288:129698, 2024

    2024

  16. [16]

    Y . Gao, S. Foroughi, Z. Ma, S. Yuan, L. Xiao, B. Bijeljic, and M. J. Blunt. Gradient information enhanced image segmentation and automatic in situ contact angle measurement applied to images of multiphase flow in porous media.Water Resources Research, 60(9):e2023WR036869, 2024

    2024

  17. [17]

    Mahdaviara, M

    M. Mahdaviara, M. J. Shojaei, J. Siavashi, M. Sharifi, and M. J. Blunt. Deep learning for multiphase segmentation of X-ray images of gas diffusion layers.Fuel, 345:128180, 2023

    2023

  18. [18]

    Kirillov, E

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo, P. Dollár, and R. Girshick. Segment anything. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 4015–4026, 2023

    2023

  19. [19]

    J. Ma, Y . He, F. Li, L. Han, C. You, and B. Wang. Segment anything in medical images.Nature Communications, 15(1):654, 2024

    2024

  20. [20]

    J. Wu, W. Ji, Y . Liu, H. Fu, M. Xu, Y . Xu, and Y . Jin. Medical SAM adapter: Adapting segment anything model for medical image segmentation.arXiv preprint arXiv:2304.12620, 2023

    work page arXiv 2023

  21. [21]

    C. Chen, J. Miao, D. Wu, A. Zhong, Z. Yan, S. Kim, J. Hu, Z. Liu, L. Sun, X. Li, T. Liu, P.-A. Heng, and Q. Li. MA-SAM: Modality-agnostic SAM adaptation for 3D medical image segmentation.Medical Image Analysis, 98:103310, 2024

    2024

  22. [22]

    X. Yan, S. Sun, K. Han, T.-T. Le, H. Ma, C. You, and X. Xie. AFTer-SAM: Adapting SAM with axial fusion transformer for medical imaging segmentation. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pages 7975–7984, 2024. 17 Zhang, R., et al

    2024

  23. [23]

    K. Tang, R. T. Armstrong, P. Mostaghimi, Y . Niu, Q. Meyer, C. Zhao, D. P. Finegan, M. Popeil, K. Singh, H. Menke, A. P. Dimou, T. Bultreys, A. Mascini, M. Knackstedt, and Y . D. Wang. Scaling deep learning for material imaging with a pseudo 3D model for domain transfer.Nature Communications, 16:11293, 2025

    2025

  24. [24]

    M. J. Shojaei, M. Blunt, and B. Bijeljic. X-ray tomography dataset of steady-state two-phase flow in bentheimer sandstone.Digital Rocks Portal, 2023

    2023

  25. [25]

    Q. Lin, B. Bijeljic, S. Berg, R. Pini, M. J. Blunt, and S. Krevor. Minimal surfaces in porous media: Pore-scale imaging of multiphase flow in an altered-wettability Bentheimer sandstone.Physical Review E, 99(6):063105, 2019

    2019

  26. [26]

    L. E. Dalton, K. A. Klise, S. Fuchs, D. Crandall, and A. Goodman. Methods to measure contact angles in scco2-brine-sandstone systems.Advances in water resources, 122:278–290, 2018

    2018

  27. [27]

    Jangda, H

    Z. Jangda, H. Menke, A. Busch, S. Geiger, T. Bultreys, and K. Singh. Subsurface hydrogen storage controlled by small-scale rock heterogeneities.International Journal of Hydrogen Energy, 60:1192–1202, 2024

    2024

  28. [28]

    Tawfik, Z

    M. Tawfik, Z. T. Karpyn, and S. X. Huang. Comparative study of traditional and deep-learning denoising approaches for image-based petrophysical characterization of porous media.Frontiers in Water, 3:800369, 2022

    2022

  29. [29]

    Scanziani, K

    A. Scanziani, K. Singh, T. Bultreys, B. Bijeljic, and M. J. Blunt. In situ characterization of immiscible three-phase flow at the pore scale for a water-wet carbonate rock.Advances in Water Resources, 121:446–455, 2018

    2018

  30. [30]

    Singh, H

    K. Singh, H. Menke, M. Andrew, C. Rau, B. Bijeljic, and M. J. Blunt. Time-resolved synchrotron X-ray micro-tomography datasets of drainage and imbibition in carbonate rocks.Scientific Data, 5:180265, 2018

    2018

  31. [31]

    Bultreys

    T. Bultreys. Estaillades carbonate# 2.National Science Foundation: Digital Rocks Portal, 2016

    2016

  32. [32]

    H. M. AlZahrani, B. Bijeljic, R. Chai, and M. J. Blunt. Pore-scale analysis and visualization of tertiary cationic surfactant flooding in a complex carbonate.ACS Omega, 10(43):51383–51395, 2025

    2025

  33. [33]

    A. M. Alhammadi, A. AlRatrout, K. Singh, B. Bijeljic, and M. J. Blunt. In situ characterization of mixed-wettability in a reservoir rock at subsurface conditions.Scientific Reports, 7:10753, 2017

    2017

  34. [34]

    Buades, B

    A. Buades, B. Coll, and J.-M. Morel. A non-local algorithm for image denoising. InProceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), volume 2, pages 60–65, 2005

    2005

  35. [35]

    Mamba: Linear-Time Sequence Modeling with Selective State Spaces

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

    work page internal anchor Pith review arXiv 2023

  36. [36]

    Perez, F

    E. Perez, F. Strub, H. de Vries, V . Dumoulin, and A. Courville. FiLM: Visual reasoning with a general conditioning layer. InProceedings of the AAAI Conference on Artificial Intelligence, volume 32, pages 3942–3951, 2018

    2018

  37. [37]

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, and W. Chen. LoRA: Low-rank adaptation of large language models.arXiv preprint arXiv:2106.09685, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  38. [38]

    S. S. M. Salehi, D. Erdogmus, and A. Gholipour. Tversky loss function for image segmentation using 3D fully convolutional deep networks.arXiv preprint arXiv:1706.05721, 2017

    work page arXiv 2017

  39. [39]

    T.-Y . Lin, P. Goyal, R. Girshick, K. He, and P. Dollár. Focal loss for dense object detection. InProceedings of the IEEE International Conference on Computer Vision (ICCV), pages 2980–2988, 2017

    2017

  40. [40]

    Z. Ma, B. Bijeljic, G. Wen, K. Tang, Y . Wang, and M. J. Blunt. Super-resolution imaging of multiphase fluid distributions in porous media using deep learning.Transport in Porous Media, 152:85, 2025

    2025

  41. [41]

    Isensee, P

    F. Isensee, P. F. Jaeger, S. A. A. Kohl, J. Petersen, and K. H. Maier-Hein. nnU-Net: A self-configuring method for deep learning-based biomedical image segmentation.Nature Methods, 18:203–211, 2021

    2021

  42. [42]

    Myronenko

    A. Myronenko. 3D MRI brain tumor segmentation using autoencoder regularization. InBrainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries, volume 11384 ofLecture Notes in Computer Science, pages 311–320, 2019

    2019

  43. [43]

    Hatamizadeh, V

    A. Hatamizadeh, V . Nath, Y . Tang, D. Yang, H. R. Roth, and D. Xu. Swin UNETR: Swin transformers for semantic segmentation of brain tumors in MRI images. InBrainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries, volume 12962 ofLecture Notes in Computer Science, pages 272–284, 2022

    2022

  44. [44]

    J. Ma, F. Li, and B. Wang. U-Mamba: Enhancing long-range dependency for biomedical image segmentation. arXiv preprint arXiv:2401.04722, 2024

    work page arXiv 2024

  45. [45]

    Y . Gao, A. Q. Raeini, A. M. Selem, I. Bondino, M. J. Blunt, and B. Bijeljic. Pore-scale imaging with measurement of relative permeability and capillary pressure on the same reservoir sandstone sample under water-wet and mixed-wet conditions.Advances in Water Resources, 146:103786, 2020. 18 Zhang, R., et al

    2020

  46. [46]

    Alhosani, B

    A. Alhosani, B. Bijeljic, and M. J. Blunt. Pore-scale imaging and analysis of wettability order, trapping and displacement in three-phase flow in porous media with various wettabilities.Transport in Porous Media, 140:59–84, 2021

    2021

  47. [47]

    Y . Gao, A. Q. Raeini, A. M. Selem, I. Bondino, M. J. Blunt, and B. Bijeljic. Pore-scale imaging with measurement of relative permeability and capillary pressure on the same reservoir sandstone sample under water-wet and mixed-wet conditions.Advances in Water Resources, 146:103786, 2020

    2020

  48. [48]

    Q. Lin, B. Bijeljic, S. Berg, R. Pini, M. J. Blunt, and S. Krevor. Minimal surfaces in porous media: Pore-scale imaging of multiphase flow in an altered-wettability bentheimer sandstone.Physical Review E, 99(6):063105, 2019

    2019

  49. [49]

    R. T. Armstrong, M. L. Porter, and D. Wildenschild. Linking pore-scale interfacial curvature to column-scale capillary pressure.Advances in Water resources, 46:55–62, 2012

    2012

  50. [50]

    L. Zhu, B. Bijeljic, and M. J. Blunt. Generation of pore-space images using improved pyramid wasserstein generative adversarial networks.Advances in Water Resources, 190:104748, 2024

    2024

  51. [51]

    L. Zhu, B. Bijeljic, and M. J. Blunt. Diffusion model-based generation of three-dimensional multiphase pore-scale images.Transport in Porous Media, 152:22, 2025. 19

    2025