Pith. sign in

REVIEW 5 major objections 5 minor 31 references

Structure and Smoothness Constrained Dual Networks for MR Bias Field Correction

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

Pith's one-line read S2DNets is a self-supervised dual-network framework that corrects MRI bias fields by enforcing piece-wise structure and smoothness constraints, with no clean reference images, and it reports better fidelity and segmentation Dice than…

desk verdict A promising self-supervised bias-field correction pipeline, but the TV loss and λ schedule as printed cancel each other, so the central ablation is not reproducible from the text. read the letter →

arxiv 2507.01326 v1 pith:C7J5ZRJ4 submitted 2025-07-02 eess.IV cs.CV

classification eess.IVcs.CV
keywords biasfieldcorrectionMRIintensityinhomogeneityself-supervisedlearningdualnetworksfuzzyclusteringsmoothnessconstraintstructuraldownstreamsegmentation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper is trying to establish that MRI bias field correction can be solved self-supervised, without paired or unpaired clean images, by making two physical properties explicit: the corrected image should be piece-wise constant within tissue classes, and the bias field should be smooth. To that end it proposes S2DNets, two encoder-decoder networks that alternately reconstruct each other's outputs: one estimates a fuzzy clustering map of tissue classes, the other estimates the multiplicative bias field. The losses come from closed-form solutions of a unified objective that combines local Gaussian smoothness, a global smoothness term, and the clustering structure constraint. On HCP and BrainWeb the paper reports best or near-best fidelity and coefficient-of-variation scores among conventional and deep-learning comparators, and improved Dice for downstream segmentation with a learned U-Net and with classical segmentation software. A reader should care because if the claim holds, non-uniform intensity can be removed from clinical MRI without any clean reference data, which is often unavailable.

What carries the argument

The load-bearing object is the closed-form solution loop in Eqs. (5)--(7), which turns a single variational objective into two reconstruction losses. For fixed bias field and cluster centers, Eq. (6) gives the optimal membership probability map; for fixed memberships and centers, Eq. (7) gives the optimal bias field. Each subnetwork is trained to output the quantity whose closed-form reconstruction matches its prediction, so the two networks supervise each other. Local smoothness enters through the masked Gaussian kernel $K(r,s)$ in the objective, global smoothness through the added regularization term on the bias field, and the structure constraint through the clustering assumption that clean tissue intensities are piece-wise constant.

What would settle it

Run the released code with $\lambda$ set exactly as Section 3.2 prescribes and with the smoothness loss removed; if the two networks produce the same outputs and the reported ablation gap (roughly 10--18% PSNR) does not appear, then the claimed smoothness mechanism is not what drives the improvement.

Watch

Extended reading notes

Core claim

The central discovery claimed is that the bias field and the piece-wise constant structure map can be learned as mutually constraining dual tasks, and that the fixed-point equations of a single objective suffice to supervise both networks. Given a corrupted image, the model alternates between Eq. (6), which reconstructs the clustering probability map from the current bias field, and Eq. (7), which reconstructs the bias field from the current clustering map; training each network to reproduce the other's reconstruction enforces consistency without clean references. The paper argues that this dual feedback, strengthened by local Gaussian smoothness and a global smoothness term on the bias field, removes non-uniform intensity while retaining edges and tissue boundaries. It reports that the complete model reaches PSNR 28.40 / SSIM 0.979 on HCP, 31.78 / 0.979 on BrainWeb T1, and 30.98 / 0.969 on BrainWeb T2, and that the global smoothness loss contributes PSNR gains of roughly 9.8%, 17.6%, and 17.3% across the three datasets.

Load-bearing premise

The results depend on the implemented training loss differing from Eqs. (10)--(11) as written, because the printed 'total variation' term is a sum of second derivatives and the printed adaptive weight $\lambda = \mathrm{Loss}_{\mathrm{bias}}/\mathrm{Loss}_{\mathrm{tv}}$ would cancel the smoothness term, so the claimed smoothness mechanism must come from a corrected loss that the paper does not specify.

Editorial extensions

If this is right

  • If S2DNets is correct, bias field correction can be done entirely without clean reference scans, removing a major data bottleneck for clinical MRI pipelines.
  • The reported downstream Dice improvements imply that correcting the bias field with structural and smoothness constraints before segmentation helps both learned and classical segmentation tools.
  • The ablation claims that the global smoothness term alone raises PSNR by roughly 10--18%, which would mean smoothness is the component doing much of the correction work rather than the clustering structure term.
  • Because the same closed-form dual-loop losses need no labels, the framework could be retrained cheaply on new scanners or protocols when intensity distributions change.

Reading between the lines

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

  • A testable extension the authors do not report is cross-dataset transfer: train S2DNets on HCP and apply it to BrainWeb without retraining, which would show whether the learned smoothness and structure priors generalize across scanners rather than fitting dataset-specific statistics.
  • The closed-form dual-loss design is not obviously limited to MRI; the same multiplicative model with a piece-wise constant prior describes illumination fields in natural images, so the method could be retargeted to Retinex-style corrections, but the paper gives no evidence on that domain.
  • Since evaluation is on 2D slices, an implicit claim is that slicewise bias fields are consistent with the 3D volume; a volumetric implementation would need to verify that neighboring slices do not produce discontinuous bias estimates.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 paper proposes S2DNets, a self-supervised dual-network method for MR bias field correction. Two encoder–decoder networks estimate a fuzzy clustering membership map and a bias field, and are trained with losses constructed from the closed-form fixed-point updates of a unified energy functional that combines piecewise-constant structural constraints with local and global smoothness terms. Experiments on HCP and BrainWeb report improved PSNR/SSIM, lower coefficient of variation, and better downstream segmentation Dice compared with several classical and deep-learning baselines. The authors state that source code is publicly available.

Significance. If the method were reproducible as described, the core idea—using closed-form updates of a fuzzy-clustering bias model as self-supervised dual-network losses—would be a useful contribution, and the downstream segmentation evaluation is a strength. The paper also provides a public code link and uses a clearly stated multiplicative image formation model. However, the manuscript contains load-bearing inconsistencies in the definition and weighting of the smoothness loss, and in the selection of the cluster number, so the reported gains are not currently supported by the text.

major comments (5)
  1. [§2.2, Eq. (10)] Equation (10) defines the 'total variation loss' as ∫Ω [∂²xb(r) + ∂²yb(r)] dr, which is the integral of the signed Laplacian, not a total variation seminorm. Such an integral can be zero or negative, and it does not penalize oscillations of b in the way a true TV term would. The authors should replace Eq. (10) with a correct total variation expression (e.g., ∫Ω |∇b| dr or ∫Ω (|∂xb| + |∂yb|) dr) and ensure that all experimental results are described consistently with that corrected definition.
  2. [§3.2, λ setting] The text states λ = Loss_bias / Loss_tv. If λ is computed from the same loss tensors without stopping gradients, the total loss becomes Loss_bias + (Loss_bias / Loss_tv)·Loss_tv = 2·Loss_bias, so the TV term cancels identically and the ablations S2DNets* and S2DNets in Tables 1–3 should coincide. If λ is detached or computed from a different schedule, that schedule is not stated. The manuscript must specify exactly how λ is computed, whether gradients flow through it, and how the reported differences between S2DNets* and S2DNets arise.
  3. [§3.3 and Fig. 2] The cluster number N is selected by 'gradually increas[ing] it by 1 to find appropriate value 4 as Fig 2 shown', and Fig. 2 reports performance on HCP, BrainWeb T1, and BrainWeb T2. If these are the same test sets used for the comparisons in Tables 1–3, then N is tuned on the test data, which invalidates a fair comparison with other methods. An independent validation split must be used for choosing N, or the authors must demonstrate that no test information influenced the choice.
  4. [§3.1–§3.2] For HCP, the paper reports PSNR/SSIM and downstream segmentation Dice against ground truth, but it never states how clean HCP images and corrupted inputs were obtained. The sentence 'training sets are further randomly splitted into 10000/5000 in HCP ... for corrupted input data and unpaired clean data' implies that synthetic bias fields were applied to HCP slices, but the protocol (e.g., bias field model, strength range, whether the same slices appear in corrupted and clean partitions) is not described. This is essential for reproducibility and for interpreting the 'Original' row in Tables 1–2.
  5. [§2.2, Eq. (6)] Equation (6) is not the exact closed-form minimizer of Eq. (4) with respect to the membership variables. Minimizing Eq. (4) for fixed b and c gives u_i(s) proportional to [∫ K(r,s)|I(s) − b(r)c_i|² dr]^{-1/(p-1)}, which depends on a kernel-smoothed version of the intensity. Equation (6) instead uses the pointwise term |I(r) − c_i ∫ K(r,s)b(s)ds|². If this replacement is an approximation, it should be stated explicitly; otherwise the claimed derivation of the reconstruction loss is incorrect.
minor comments (5)
  1. [Abstract and code link] The GitHub URL in the abstract contains duplicated braces ('.../S2DNets}{https://.../S2DNets'); it should be a single clean link.
  2. [Terminology] The paper alternates between 'unsupervised' and 'self-supervised'; since the method trains without any clean reference, the term self-supervised should be defined and used consistently.
  3. [Fig. 2] The figure caption in the text does not state which metric is plotted on the vertical axis or how the curves are aggregated; please specify this in the caption.
  4. [§3.4, downstream segmentation] Table 3 reports Dice for GM and WM using a pre-trained 2D U-Net and FSL, but the text does not specify whether the U-Net is a single multi-class model or separate binary models, nor which FSL tool and parameters were used; these details are needed for reproducibility.
  5. [References] Reference [28] appears not to be cited in the text, and reference [29] has garbled author formatting; please check the reference list against the citations.

Circularity Check

1 steps flagged · score 6.0 of 10

Smoothness-constraint claim cancels by construction under the published λ schedule.

  1. self definitional [Section 2.2, Eq. (10); Section 3.2, Implementation Details; Tables 1-3]
    "LossT V=∫Ω[∂2 xb(r)+∂2 yb(r)]dr ... The total loss for training bias estimation network is: Lossbias + λLossT V ... To balance the weight between local and global smoothness constraints, λ = Lossbias / Losstv is adaptively adjusted during training process."

    As written, substituting λ into the total loss yields Lossbias + (Lossbias/LossTV)·LossTV = 2·Lossbias, so the global smoothness term cancels algebraically and the objective is simply a constant multiple of the bias-field reconstruction loss. The ablation between S2DNets* (without TV) and S2DNets (with TV) could not therefore produce the reported PSNR gains (9.8%, 17.6%, 17.26%) under the published equations; the claimed contribution of the TV term reduces to the same loss by construction. Moreover, Eq. (10) is the integral of the signed Laplacian, not a total-variation penalty, so even with a fixed λ it would not enforce the stated smoothness constraint.

full rationale

The self-supervised closed-form dual losses are not themselves circular: they train the networks to satisfy consistency/fixed-point equations derived from the fuzzy-clustering objective, and the headline PSNR/SSIM/CV and Dice evaluations against BrainWeb ground truth and HCP downstream segmentation are external. No load-bearing self-citation appears. However, the paper's distinctive smoothness constraint is rendered vacuous by its own equations: with λ = Lossbias/LossTV, the combined loss equals 2·Lossbias, so the central ablation result cannot be reproduced from the text, and Eq. (10) is not a total-variation term. This is a self-definitional reduction of the paper's main claimed improvement rather than a merely missing implementation detail.

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

The paper does not introduce new physical entities. Its free parameters are mostly standard (p, d, sigma from prior work) plus the cluster count N chosen on the test set and an ambiguous TV weight lambda. The axioms are standard assumptions in the bias-field literature: multiplicative corruption model, piece-wise constant tissues, smooth bias field, and the validity of alternating-minimization fixed points as training signals.

free parameters (6)
  • cluster number N = 4
    Selected by increasing from 2 and observing test-set performance in Fig. 2; model selection on test data.
  • lambda (TV weight) = stated as Loss_bias/Loss_tv, ambiguous
    As written this makes the total loss 2*Loss_bias, contradicting the ablation; an implicit scalar is missing.
  • fuzziness factor p = 2
    Set to 2 per reference [5]; standard in fuzzy clustering.
  • kernel size d = 17
    Adopted from reference [9] with relationship d <= 4*sigma + 1.
  • sigma (Gaussian shape) = 4
    Adopted from reference [9].
  • MOTSU threshold count M = 3
    Default; affects the foreground mask in Eq. (2).
assumptions (6)
  • domain assumption Multiplicative bias field model I = i*b + n with noise estimated by quasi-Gaussian functions
    Standard model in MRI bias field literature (Eq. 1).
  • domain assumption Clean MR image intensity is piece-wise constant
    Foundation of the fuzzy clustering constraint (Eq. 4).
  • domain assumption Bias field is locally smooth, modeled by masked Gaussian kernel K
    Eq. 3; standard assumption in this literature.
  • domain assumption Closed-form updates u', b', c_i (Eqs. 5-7) are valid fixed points of the objective
    Used directly as training losses; assumes alternating minimization converges to a useful minimum.
  • domain assumption U-Net architectures can represent the clustering and bias functions well enough
    No theoretical guarantee; empirical only.
  • ad hoc to paper Reconstructed probability/bias losses (Eqs. 8-9) guide the networks to a non-degenerate solution
    The objective admits trivial fixed points (e.g., constant bias); the paper does not analyze or avoid them.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Structure and Smoothness Constrained Dual Networks for MR Bias Field Correction." pith.science (2026). https://pith.science/paper/C7J5ZRJ4

@misc{pith2026250701326,
  author       = {Pith},
  title        = {Pith review of: Structure and Smoothness Constrained Dual Networks for MR Bias Field Correction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/C7J5ZRJ4}},
  note         = {Machine review of arXiv:2507.01326}
}
read the original abstract

MR imaging techniques are of great benefit to disease diagnosis. However, due to the limitation of MR devices, significant intensity inhomogeneity often exists in imaging results, which impedes both qualitative and quantitative medical analysis. Recently, several unsupervised deep learning-based models have been proposed for MR image improvement. However, these models merely concentrate on global appearance learning, and neglect constraints from image structures and smoothness of bias field, leading to distorted corrected results. In this paper, novel structure and smoothness constrained dual networks, named S2DNets, are proposed aiming to self-supervised bias field correction. S2DNets introduce piece-wise structural constraints and smoothness of bias field for network training to effectively remove non-uniform intensity and retain much more structural details. Extensive experiments executed on both clinical and simulated MR datasets show that the proposed model outperforms other conventional and deep learning-based models. In addition to comparison on visual metrics, downstream MR image segmentation tasks are also used to evaluate the impact of the proposed model. The source code is available at: https://github.com/LeongDong/S2DNets}{https://github.com/LeongDong/S2DNets.

Figures

Figures reproduced from arXiv: 2507.01326 by the authors.

Figure 1
Figure 1. Contributions and framework of proposed model. smoothly varying bias field. However, it is limited in describing the bias field with complex distribution. Ma et al. [30] bypassed explicit modeling of bias field smoothness but instead employed a global hard threshold on image gradients to preserve structural edges for sparse reconstruction. However, due to the lack of smoothness constraints on the bias field, it suff… view at source ↗
Figure 2
Figure 2. The performance of proposed model on HCP, BrainWeb T1, T2 datasets with different N values. structural details, while ensuring that the intensity distribution aligns with cor￾responding clean image, as measured by Structural Similarity (SSIM) and Peak Signal Noise Ratio (PSNR); 2) the non-uniform intensity within tissues or or￾gans should be effectively corrected, which can be evaluated by Coefficient of Variation (… view at source ↗
Figure 3
Figure 3. The performance of different methods. Yellow rectangle denotes remained bias field or distorted structure. In contrast, S2DNets effectively remove bias fields while preserving structural details with high fidelity. The ablation study on lossT V reveals that while CV values increase slightly (no more than 0.011), PSNR improves significantly by 9.8%, 17.6% and 17.26% on HCP, Brainweb T1 and T2 datasets, respectively. … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 31 canonical work pages

  1. [1]

    K., Agrawal, S., Panda, R., Dora, L., Abraham, A.: Intensity inhomo- geneity correction in brain MRI: a systematic review of techniques, current trends and future challenges

    Mishro, P. K., Agrawal, S., Panda, R., Dora, L., Abraham, A.: Intensity inhomo- geneity correction in brain MRI: a systematic review of techniques, current trends and future challenges. Neural Computing and Applications37, 1821–1838 (2025)

  2. [2]

    Physics in Medicine & Biology69, 23TR01 (2024)

    Singh, R., Singh, N., Kaur, L.: Deep learning methods for 3D magnetic resonance image denoising, bias field and motion artifact correction: a comprehensive review. Physics in Medicine & Biology69, 23TR01 (2024)

  3. [3]

    G., Zijdenbos, A

    Sled, J. G., Zijdenbos, A. P., Evans, A. C.: A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Transactions on Medical Imaging 17(1), 87–97 (1998)

  4. [4]

    J., Avants, B

    Tustison, N. J., Avants, B. B., Cook, P. A., Zheng, Y., Egan, A., Yushkevich, P.-A., Gee, J. C.: N4ITK: improved N3 bias correction. IEEE Transactions on Medical Imaging, 29(6), 1310–1320 (2010)

  5. [5]

    K., Agrawal, S., Panda, R., Abraham, A.: Novel fuzzy clustering-based bias field correction technique for brain magnetic resonance images

    Mishro, P. K., Agrawal, S., Panda, R., Abraham, A.: Novel fuzzy clustering-based bias field correction technique for brain magnetic resonance images. IET Image Processing, 14(9), 1929-1936 (2020)

  6. [6]

    S., Zeng, Y., Chai

    Li, Z., Tang. S., Zeng, Y., Chai. S., Ye, W., Yang, F., Huang, K.: A level-set method with a multiplicative-additive constraint model for image segmentation and bias correction. Knowledge-Based Systems,297(3), 111972, (2024)

  7. [7]

    B.: AMSLS: adaptive multi-scale level set method based on local entropy for image segmentation

    Feng, C., Gao, W., Wang, R., Yang, Y., Wu. B.: AMSLS: adaptive multi-scale level set method based on local entropy for image segmentation. Applied Mathematical Modelling, 134, 500–519 (2024)

  8. [8]

    C., Davatzikos, C.: Multiplicative intrinsic component optimiza- tion (MICO) for MRI bias field estimation and tissue segmentation

    Li, C., Gore, J. C., Davatzikos, C.: Multiplicative intrinsic component optimiza- tion (MICO) for MRI bias field estimation and tissue segmentation. Magnetic Resonance Imaging, 32(7), 913–923 (2014)

Show all 31 references
  1. [9]

    IEEE Transactions on Imaging Processing,24(11), 3927–3938 (2015)

    Duan, Y., Chang, H., Huang, W., Zhou, J., Lu, Z., Wu, C.: TheL0 regularized Mumford-Shahmodelsforbiasfieldcorrectionandsegmentationofmedicalimages. IEEE Transactions on Imaging Processing,24(11), 3927–3938 (2015)

  2. [10]

    R.: Deep semi-supervised bias field correction of Mr images

    Goldfryd, T., Gordon, S., Raviv, T. R.: Deep semi-supervised bias field correction of Mr images. In: IEEE 18th International Symposium on Biomedical Imaging. pp. 1836-1840. IEEE, Nice (2021) 10 D. Liang et al

  3. [11]

    K., Lin, W., Wang, L., Shen, D., Li, G.: ABCnet: Adversarial bias correction network for infant brain MR images

    Chen, L., Wu, Z., Hu, D., Wang, F., Smith, J. K., Lin, W., Wang, L., Shen, D., Li, G.: ABCnet: Adversarial bias correction network for infant brain MR images. Medical Image Analysis,72, 102133 (2021)

  4. [12]

    J., Patel, P., Liu, T., Yang, X.: Intensity non-uniformity correction in MR imaging using residual cycle generative adversarial network

    Dai, X., Lei, Y., Liu, Y., Wang, T., Ren, L., Curran, W. J., Patel, P., Liu, T., Yang, X.: Intensity non-uniformity correction in MR imaging using residual cycle generative adversarial network. Physics in Medicine & Biology,65, 215025 (2020)

  5. [13]

    M., Jing, B., Zhang, G., Chen, G

    Chen, J, Lu, R., Ye, S., Guang, M., Tassew, T. M., Jing, B., Zhang, G., Chen, G. Shen, D.: Image Recovery matters: a recovery-extraction framework for robust fetal brain extraction from MR images. IEEE Journal of Biomedical and Health Informatics, 28(2), 823-834 (2024)

  6. [14]

    In: International Conference on Medical Imaging With Deep Learning

    Simko, A., Lofstedt, T., Garpebring, A., Nyholm, T., Jonsson, J.: MRI bias field correction with an implicitly trained CNN. In: International Conference on Medical Imaging With Deep Learning. pp. 1125-1138 (2022)

  7. [15]

    In: Interna- tional Conference on Bioinformatics and Biomedicine

    Chen, S., Feng, C., Li, W., Yang, J., Zhao, D.: BECNN: bias field estimation CNN trained with a dual route implicit supervised learning strategy. In: Interna- tional Conference on Bioinformatics and Biomedicine. pp. 3119-3122. IEEE, Lisbon (2024)

  8. [16]

    arXiv: 2501.12244 (2025)

    Yang, H., Timko, E., Fernandez, B.: Zero-shot bias correction: efficient MR image inhomogeneity reduction without any data. arXiv: 2501.12244 (2025)

  9. [17]

    Y., et al.: DeepN4: learning N4ITK bias field correction for T1-weighted images

    Kanakaraj, P., Yao, T., Cai, L. Y., et al.: DeepN4: learning N4ITK bias field correction for T1-weighted images. Neuroinformatics,22, 193–205 (2024)

  10. [18]

    R., Coupe, P., Manjon, J

    Perez-Caballero, M., Morell-Ortega, S., Perez, M. R., Coupe, P., Manjon, J. V.: Unsupervised deep learning method for bias correction. In: Medical Imaging with Deep Learning. pp.1–9. Paris (2024)

  11. [19]

    C.: Learning to enhance low-light image via zero-reference deep curve estimation

    Li, C., Guo, C., Loy, C. C.: Learning to enhance low-light image via zero-reference deep curve estimation. IEEE Transactions on Pattern Analysis and Machine Intel- ligence, 44(8), 4225–4238 (2022)

  12. [20]

    IEEE Transactions on Image Processing, 33, 1432-1447 (2024)

    Chen, Y., Jiang, R., Zheng, Y., Sheng, B., Yang, Z.-X., Wu, E.: Dual branch multi- level semantic learning for few-shot segmentation. IEEE Transactions on Image Processing, 33, 1432-1447 (2024)

  13. [21]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Guo, Y., Chen, J., Wang, J., Chen, Q., Cao, J., Deng, Z., Xu, Y., Tan, M.: Close-loop matters: dual regression networks for single image super-resolution. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 5407-

  14. [22]

    N., Li, Y

    Berahmand, K., Bahadori, S., Abadeh, M. N., Li, Y. F., Xu, Y.: SDAC-DA: semi- supervised deep attributed clustering using dual autoencoder. IEEE Transactions on Knowledge and Data Engineering,36(11), 6989–7002 (2024)

  15. [23]

    In: 18th International Conference on Medical Image Com- puting and Computer-Assisted Intervention

    Ronneberger, O., Fischer, P., Brox, T.: U-Net: convolutional networks for biomedi- cal image segmentation. In: 18th International Conference on Medical Image Com- puting and Computer-Assisted Intervention. pp. 234-241. Springer, Munich (2015)

  16. [24]

    S., Glasser, M

    Elam, J. S., Glasser, M. F., Harms, M. P., et al.: The human connectome project: a retrospective. Neuroimage,244, 118543 (2021)

  17. [25]

    L., Zijdenbos, A

    Collins, D. L., Zijdenbos, A. P., Kollokian, V., Sled, J. G., Kabani, N. J., Holmes, C. J., Evans, A. C.: Design and construction of a realistic digital brain phantom. IEEE Transactions on Medical Imaging17(3), 463-468 (1998)

  18. [26]

    Multimedia Tools and Applications32, 15007- 15026 (2023)

    Ning, G.: Two-dimensinal Otsu multi-threshold image segmentation based on hy- brid whale optimization algorithm. Multimedia Tools and Applications32, 15007- 15026 (2023)

  19. [27]

    IEEE Transactions on Image Processing,32, 2132–2146 (2023) S2DNets: MR Bias Field Correction 11

    Chen, Y., Wang, Z., Bai, X.: Fuzzy sparse subspace clustering for infrared image segmentation. IEEE Transactions on Image Processing,32, 2132–2146 (2023) S2DNets: MR Bias Field Correction 11

  20. [28]

    C.: A generic approach towards image manipulation pa- rameter estimation using convolutional neural networks

    Bayar, B., Stamm, M. C.: A generic approach towards image manipulation pa- rameter estimation using convolutional neural networks. In: Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security. pp. 147-157. ACM, New York (2017)

  21. [29]

    K.: Reproducibility of brain MRI segmentation algorithons: empirical comparison of local MAP PSTA- PLE, FreeSurfer, and FSL-FIRST

    Velasco-Annis, C., Akhondi-Asi A., Stamm, A., Warfield, S. K.: Reproducibility of brain MRI segmentation algorithons: empirical comparison of local MAP PSTA- PLE, FreeSurfer, and FSL-FIRST. Journal of Neuroimaging28(2), 162-172 (2018)

  22. [30]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Ma, W., Morel, J.-M., Osher, S., Chien, A.: An L1-based variational model for Retinex theory and its application to medical images. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 153-160. IEEE, Colorado (2011)

  23. [5416]

    IEEE, Seattle (2020)

Pith tools

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