Pith. sign in

REVIEW 4 major objections 6 minor 75 references

Unraveling Normal Anatomy via Fluid-Driven Anomaly Randomization

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read UNA is a single modality-agnostic deep network that reconstructs healthy brain anatomy from CT and T1w/T2w/FLAIR MRI scans containing stroke lesions, without needing fine-tuning on the input image.

desk verdict UNA's fluid-driven anomaly randomization and contralateral contrastive loss are genuinely new and the evaluation is thorough, but the central claim about real-pathology reconstruction is only supported indirectly, so the paper deserves a serious referee with requests for direct validation. read the letter →

arxiv 2501.13370 v1 pith:Q57XV4GC submitted 2025-01-23 eess.IV cs.CV

Peirong Liu , Ana Lawry Aguila , Juan E. Iglesias This is my paper · ORCID
classification eess.IVcs.CV
keywords normalbrainanatomyreconstructionfluid-drivenanomalyrandomizationadvection-diffusionPDEmodality-agnosticlearningcontralateralsymmetrypriorstrokelesionsynthesisdetectionsyntheticmedicalimagegeneration
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

UNA is a single deep-learning network that takes a brain scan, CT or any of the major MRI contrasts, and reconstructs the same brain as it would appear without pathology, working on both healthy scans and scans with stroke lesions and requiring no fine-tuning on the image being analyzed. This matters because the standard toolkit of brain analysis, segmentation, parcellation, registration, and super-resolution, is built and trained for healthy, high-resolution T1w MRI, and degrades in clinical settings where both the acquisition and the anatomy differ. To train such a model without expensive pathology annotations, the paper introduces fluid-driven anomaly randomization, which evolves a handful of real stroke segmentations through an advection-diffusion equation inside the brain boundary to produce an unlimited supply of realistic lesion shapes and intensities on the fly. Exploiting the brain's near-mirror symmetry, the network borrows healthy tissue from the contralateral hemisphere through a self-contrastive loss, and the authors report state-of-the-art reconstruction across CT and T1w/T2w/FLAIR MRI on both simulated and real stroke data, with direct application to anomaly detection.

What carries the argument

The load-bearing mechanism is fluid-driven anomaly randomization: a pathology probability map $P(x,t)$ is treated as mass evolving under the advection-diffusion PDE $\frac{\partial P}{\partial t} = -\nabla\times\Psi(x)\cdot\nabla P + \nabla\cdot(\Phi^2(x)\,\nabla P)$, where the curl potential $\Psi$ guarantees incompressible transport and the squared field $\Phi^2$ guarantees non-negative diffusion, with a zero-Neumann boundary condition keeping the anomaly inside the brain. This turns a few gold-standard stroke segmentations into an unlimited stream of training anomalies whose shapes and intensities obey realistic constraints by construction. The second mechanism is the contralateral-paired input: each image is fed together with its axial-flipped version, registered back to the original frame, so the model can borrow healthy anatomy from the mirror hemisphere; an intra-subject contrastive loss (Eq. 8) enforces similarity to that healthy mirrored tissue and distinctiveness from the diseased input. Together these let a standard 3D U-Net, trained half on synthetic and half on real data, map any input modality to clean anatomy.

What would settle it

Apply UNA to stroke cases where the lesion substantially involves both hemispheres, so the mirrored counterpart of the diseased tissue is itself diseased; if reconstruction error in the affected regions does not clearly exceed that for unilateral lesions, the symmetry assumption is not in fact load-bearing, whereas a sharp degradation would confirm the mechanism. A complementary check the paper does not report is to compare the shape and intensity distribution of the fluid-generated anomaly profiles against held-out real lesion segmentations with a distributional metric, which would test whether the simulated pathologies are quantitatively realistic rather than merely visually plausible.

Watch

Extended reading notes

Core claim

The paper's central claim is that healthy brain anatomy can be reconstructed from pathological scans of any common modality by a single model, without fine-tuning, provided the training set is enriched with synthetic pathologies generated by a fluid-dynamics process. Starting from a real stroke segmentation or a random noise shape, the anomaly map is transported forward in time under an incompressible velocity field and a non-negative diffusion field confined by the brain boundary, yielding endless variations of lesion geometry and intensity; these are encoded onto synthetic healthy scans of randomized contrast and resolution. The reconstruction network is trained with a voxel-wise loss whose weighting focuses on the diseased regions, plus an intra-subject contrastive loss that pulls the output toward the registered mirror-image healthy anatomy and pushes it away from the abnormal input. Evaluated on simulated lesions with known ground truth, on two real stroke datasets, and in a downstream anomaly-detection task, the authors report that UNA outperforms the closest contrast-agnostic synthesis and anomaly-detection methods in full-brain, healthy-region, and diseased-region metrics.

Load-bearing premise

The load-bearing premise is that the brain is close enough to mirror-symmetric that the flipped healthy hemisphere supplies a valid reference for the diseased side; the paper's own failure case shows what happens when that premise runs out, when a lesion heavily occludes the underlying anatomy, UNA over-corrects because no healthy mirror tissue remains to borrow from.

Editorial extensions

If this is right

  • General-purpose analysis tools that expect healthy, high-resolution T1w MRI, including segmentation, parcellation, registration, and super-resolution, can be applied to clinical CT and MRI scans with stroke lesions because UNA first restores the underlying healthy anatomy.
  • One trained UNA model spans CT, T1w, T2w, and FLAIR MRI at variable resolution without retraining, which would let a single pipeline handle heterogeneous clinical archives.
  • Anomaly detection follows directly from the reconstruction: the voxel-wise absolute difference between the diseased input and UNA's healthy output locates lesions, and the paper reports state-of-the-art Dice scores against dedicated unsupervised anomaly detectors.
  • Training requires only a few gold-standard pathology segmentations alongside healthy anatomy labels, since the fluid-driven generator multiplies those few examples into an unlimited set of anomaly profiles.
  • On healthy inputs, the same network performs anatomy reconstruction and super-resolution, so it can serve as a standard normalizer upstream of existing tools rather than only as a pathology-specific fix.

Reading between the lines

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

  • Because the anomaly generator is initialized from stroke segmentations but also from pure noise, the same training recipe probably transfers to other lesion types such as tumors, white-matter disease, or hemorrhage, provided the intensity-encoding rule (Eqs. 5-6) is adapted; the paper does not test this.
  • The over-correction failure mode implies a ceiling on any contralateral-borrowing method: pathologies that are bilateral or that nearly replace an entire hemisphere lack a healthy mirror reference, so clinical deployment would need to detect and flag such cases rather than trust the reconstruction.
  • A cheap, decisive experiment is to ablate the PDE-based generator against pure noise augmentation while matching lesion volume and intensity statistics; the paper's UNA-(b) ablation is close but does not match statistics, so it does not isolate the value of fluid realism.
  • UNA's reconstructed healthy anatomy is itself a candidate covariate for stroke outcome prediction or for pathology-aware registration, directions the paper lists as future work but does not quantify.
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

4 major / 6 minor

Summary. UNA, the manuscript under review, proposes a contrast- and resolution-agnostic deep learning framework to reconstruct healthy brain anatomy from images that may contain pathology. The method has three main components: (i) fluid-driven anomaly randomization, which evolves a small set of real lesion segmentations through an advection-diffusion PDE to synthesize diverse pathology profiles; (ii) a 3D UNet trained on a combination of synthetic and real data with a reconstruction loss and a contralateral-paired input that exploits brain symmetry; and (iii) an intra-subject self-contrastive loss that encourages the reconstruction to match the healthy contralateral tissue while remaining distinct from the diseased region. The authors report state-of-the-art results on simulated and real data from CT, T1w, T2w, and FLAIR MRI, and demonstrate a direct application to anomaly detection.

Significance. If the central claim is correct, UNA would be a useful tool for enabling general-purpose healthy-brain analysis models (e.g., SynthSeg, FreeSurfer) to be applied to clinical scans with pathology, without requiring paired annotations or fine-tuning. The paper's strengths include a clearly specified PDE-based generative model with well-defined boundary conditions, a transparent implementation with code release, extensive experiments across eight public datasets and multiple modalities, and an honest discussion of the method's limitations, notably the failure case in Fig. 4. However, the evidence for the load-bearing part of the claim—accurate reconstruction of anatomy occluded by real pathology—is currently indirect: the only quantitative lesion-region scores come from the authors' own synthetic generator, and real stroke data are evaluated only in healthy regions, where the loss trivially encourages copying the input. The contribution is therefore real but not yet fully demonstrated.

major comments (4)
  1. [Sec. 5.2 and Table 2] The only quantitative evaluation on real stroke data is limited to healthy regions, where the training loss for real images (Eq. 7) supervises the reconstruction toward the input itself; these scores are thus largely a copy-the-input measure and do not attest to the central claim of restoring anatomy occluded by real pathology. The failure case in Fig. 4 (bottom row) shows exactly the lesion-region error that remains unmeasured. Please provide a quantitative assessment of lesion-region reconstruction on real data, for instance by evaluating on a dataset with pre-stroke or post-resolution healthy ground truth, or by a manual expert scoring study, and if none is feasible, state this restriction explicitly and temper the corresponding claims.
  2. [Sec. 5.1 and Table C.2] The synthetic test set used in Table 1 is produced by the same fluid-driven anomaly randomization used for training, and the hyperparameters (λ_p, λ_contrast) are selected on the same synthetic test protocol described in Table C.2 as 'same as first-row group in Tab. 1'. Consequently, the diseased-region (D) scores in Table 1 are vulnerable to generator/test overlap and do not provide independent evidence of generalization to real pathology. Please validate on a synthetic test set generated with a different generator or with substantially different parameters, and ensure that hyperparameter selection is performed on a validation set disjoint from the test set.
  3. [Sec. 4, Eq. (8)] The intra-subject contrastive loss relies on the assumption that the contralateral hemisphere is a reliable healthy reference. This assumption can fail in stroke patients with bilateral pathology, mid-line shift, or mass effect, and the paper's own failure case (Sec. 5.2, Fig. 4 bottom row) is a direct instance of the resulting over-correction. Please provide an analysis of the prevalence and severity of this failure mode on the real datasets, and discuss possible mitigations, such as restricting the contralateral reference to confidently healthy tissue.
  4. [Sec. 4, Eq. (7)] The definition of the voxel weight k(x) is inconsistent with the accompanying text. As printed, for d=1 (real images) the weight is 1−p(x) inside the lesion region and 0 outside it, which contradicts the statement that the loss is computed 'exclusively for the healthy regions'; for d=0 (synthetic images) the weight is 1 inside the lesion and 0 outside, again contradicting the statement that the loss spans the whole brain. Please correct the equation and provide the intended weighting.
minor comments (6)
  1. [Abstract and Sec. 5] The phrase 'state-of-the-art performance in all tasks and modalities' is broader than the evidence: the real-data evaluation on stroke datasets covers only healthy regions, and ISLES is evaluated only qualitatively. Please make the claim more precise, e.g., 'on simulated lesion regions and healthy regions of real scans'.
  2. [Introduction vs. Related Work] The novelty relative to PEPSI is not fully clear because PEPSI also handles pathology in a contrast-agnostic manner; the Introduction lists several differences, but a direct empirical comparison on the same evaluation protocol would strengthen the positioning.
  3. [References] References [44] and [45] are the same paper (Modat et al., NiftyReg), and references [16] and [17] are the same paper (Fischl et al.); please merge the duplicate entries.
  4. [Table 2] The claim of 'approximately 10%' improvement over the next best method on ATLAS is not tied to a specific metric; please specify the metric and report confidence intervals or statistical significance.
  5. [Table 3] The anomaly detection Dice scores for UNA are in the range 0.31–0.37; a brief discussion of the operating point and the clinical or practical utility of these scores would help the reader interpret the results.
  6. [Appendix A] The Perlin noise code seeds the random number generator with the current time, which hurts reproducibility; please fix the seed using a constant or a configurable parameter.

Circularity Check

2 steps flagged · score 6.0 of 10

Diseased-region scores (Tab. 1, columns D) are measured only on UNA's own fluid-driven generator, seeded from the same DStroke pool as training and inverting the corruption trained in Eq. (7); hyperparameters are tuned on that same protocol (Tab. C.2), and real stroke data are scored only in healthy regions (Tab. 2).

  1. fitted input called prediction [Sec. 5 Synthetic Data Generation; Sec. 5.1 / Tab. 1 (D columns); App. C / Tab. C.2; generator Eqs. (1)-(6); training loss Eq. (7)]
    "we employ our synthetic generator to create 1,000 testing samples from DHealthy, encoded with random anomaly profiles from DStroke ... Remarkably, UNA outperforms competing models by a large margin in anatomy reconstruction within diseased tissue ... Hyperparameter search of UNA. Testing images are real T1w MRI encoded with simulated pathology (same as the first-row group in Tab. 1)."

    The Tab. 1 test images are corrupted by UNA's own fluid-driven randomizer (Alg. 1; Eqs. 1-6) from the same DStroke profile pool used to seed training anomalies, and for synthetic samples Eq. (7) supervises reconstruction across the whole brain, so the model is trained to invert exactly this corruption model. The D-region scores thus measure in-distribution inversion of the paper's own pathology generator rather than restoration of real pathology. The reduction is completed by App. C: the loss weights λp and λcontrast are selected on this same protocol ('same as the first-row group in Tab. 1'), so the reported disease-region margins are fitted to the evaluation distribution, and Tab. 2 never measures lesion regions on real stroke data.

  2. other [Sec. 5.2 / Tab. 2 caption; Eq. (7) real-data loss]
    "Since we do not have ground truth anatomy for the stroke datasets, we only report the reconstruction performance within healthy regions."

    Flagged per the reviewing rule as missing support rather than a reduction: on ATLAS/ISLES the paper states that no ground-truth healthy anatomy exists and scores only healthy regions, while for real images Eq. (7) masks lesions and supervises healthy regions toward the input. Tab. 2 therefore mostly quantifies input fidelity in unaffected tissue, and the central claim that UNA reconstructs normal anatomy inside real lesions is never directly evaluated. Its only quantitative lesion-region support is the in-distribution synthetic result of Step 1, whose documented failure mode (Fig. 4, bottom row; Sec. 6 'over-corrects') is itself a lesion-region error.

full rationale

The core method is not circular: UNA is evaluated against ground-truth healthy anatomy in synthetic settings and against healthy regions in real data, and the training loss for real images supervises only healthy voxels, which is not equivalent to the central claim. However, the paper's strongest evidence for diseased-region reconstruction comes from Tab. 1, whose test images are generated using the paper's own fluid-driven anomaly generator (Eqs. 1-6) seeded from the same DStroke profile pool as training, with the whole-brain reconstruction loss of Eq. (7) applied to synthetic data. The hyperparameters in Tab. C.2 are tuned on this same synthetic test protocol, and real stroke evaluation in Tab. 2 explicitly excludes lesions due to lack of ground truth. The documented failure case (Fig. 4, bottom) shows a lesion-region error. Thus the central claim that UNA restores healthy anatomy in real diseased regions is either untested or supported only by in-distribution generator fit, making the reported lesion-region numbers vulnerable to circularity. Healthy-region results remain largely independent evidence, so the score is 6 rather than higher.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The method introduces no new physical entities. The central claim rests on hand-chosen loss weights and domain priors about brain symmetry and lesion appearance. The main free parameters were tuned on the same synthetic evaluation protocol, which is the primary circularity concern.

free parameters (5)
  • lambda_p (anomaly attention weight, Eq. 7) = 1.0
    Selected through hyperparameter search on the same synthetic test protocol (Tab. C.2); controls the weighting of reconstruction loss inside and outside pathology regions.
  • lambda_contrast (self-contrastive weight, Eq. 9) = 2.0
    Selected through hyperparameter search on the same synthetic test protocol (Tab. C.2); controls the strength of the contralateral contrastive loss.
  • lambda_grad (gradient L1 weight, Eq. 7) = not specified
    Appears in the reconstruction loss but no value is reported in the paper or appendix.
  • alpha, beta, gamma (temperature scaling, Eq. 8) = not specified
    Temperature parameters of the contrastive loss are not reported.
  • V_multiplier (velocity field scale, Appendix A.2) = not specified
    Scales the random incompressible velocity field in the anomaly generator; set in code but not reported.
assumptions (5)
  • domain assumption Advection-diffusion PDE with incompressible flow and non-negative diffusion (Eq. 3) models realistic anomaly evolution.
    The paper assumes that transporting a pathology probability map via this PDE yields realistic and diverse anomaly profiles, but this realism is not quantitatively validated.
  • domain assumption Zero-Neumann boundary condition (Eq. 2) confines anomalies to the brain region.
    Used to ensure generated anomalies stay inside the brain, but it does not enforce tissue-specific constraints.
  • domain assumption Brain anatomy is approximately symmetric, so the contralateral hemisphere can serve as a healthy reference.
    Entering via the contralateral-paired input and Eq. (8). The paper's own failure case (Fig. 4) demonstrates the risk of this assumption.
  • domain assumption Pathology appearance follows the intensity prior of Eq. (6): darker than white matter in T1w-like and brighter in T2w/FLAIR-like contrasts, with 20% sign randomization.
    Used to encode anomaly profiles into healthy images; a strong assumption for CT and heterogeneous lesions.
  • domain assumption The random modality synthesis of Eq. (4) covers the real acquisition variability of the test datasets.
    This is the same assumption used by SynthSR and Brain-ID; the paper does not independently verify it for CT and FLAIR.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unraveling Normal Anatomy via Fluid-Driven Anomaly Randomization." pith.science (2026). https://pith.science/paper/Q57XV4GC

@misc{pith2026250113370,
  author       = {Pith},
  title        = {Pith review of: Unraveling Normal Anatomy via Fluid-Driven Anomaly Randomization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Q57XV4GC}},
  note         = {Machine review of arXiv:2501.13370}
}
read the original abstract

Data-driven machine learning has made significant strides in medical image analysis. However, most existing methods are tailored to specific modalities and assume a particular resolution (often isotropic). This limits their generalizability in clinical settings, where variations in scan appearance arise from differences in sequence parameters, resolution, and orientation. Furthermore, most general-purpose models are designed for healthy subjects and suffer from performance degradation when pathology is present. We introduce UNA (Unraveling Normal Anatomy), the first modality-agnostic learning approach for normal brain anatomy reconstruction that can handle both healthy scans and cases with pathology. We propose a fluid-driven anomaly randomization method that generates an unlimited number of realistic pathology profiles on-the-fly. UNA is trained on a combination of synthetic and real data, and can be applied directly to real images with potential pathology without the need for fine-tuning. We demonstrate UNA's effectiveness in reconstructing healthy brain anatomy and showcase its direct application to anomaly detection, using both simulated and real images from 3D healthy and stroke datasets, including CT and MRI scans. By bridging the gap between healthy and diseased images, UNA enables the use of general-purpose models on diseased images, opening up new opportunities for large-scale analysis of uncurated clinical images in the presence of pathology. Code is available at https://github.com/peirong26/UNA.

Figures

Figures reproduced from arXiv: 2501.13370 by the authors.

Figure 1
Figure 1. Powered by the proposed fluid-driven anomaly randomization, [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. UNA’s framework overview for modality-agnostic learning of healthy anatomy, supported by fluid-driven anomaly randomization. real datasets containing pathology (DReal) and synthetic im￾ages (DSynth) generated from fluid-driven anomaly random￾ization (Sec. 3), featuring varying simulated modalities and abnormality conditions. During training, we define the fol￾lowing healthy anatomy reconstruction loss, which takes i… view at source ↗
Figure 3
Figure 3. Qualitative comparisons on healthy anatomy recon [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Qualitative comparisons on healthy anatomy reconstruc [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Visualizations of directly applying UNA’s healthy anatomy reconstruction for anomaly detection. The estimated anomaly is computed as the absolute difference between diseased T1w MRI scans and UNA’s reconstructed healthy anatomy. Image Source Dataset SynthSR [26] Brain-…
Figure 6
Figure 6. Figure 6: Ablations on UNA’s healthy anatomy reconstruction. Method L1 (↓) PSNR (↑) SSIM (↑) F H D F H D F H D UNA-(a) 0.0229 0.0193 0.0008 23.71 25.09 38.92 0.859 0.890 0.904 UNA-(b) 0.0195 0.0182 0.0005 25.79 27.30 42.35 0.903 0.925 0.950 UNA-(c) 0.0155 0.0163 0.0004 30.00 31.…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

75 extracted references · 73 canonical work pages

  1. [1]

    https://monai.io/model-zoo.html

    MONAI model zoo. https://monai.io/model-zoo.html. 2 8

    work page

  2. [2]

    Vector potentials in three-dimensional non- smooth domains

    Ch ´erif Amrouche, Christine Bernardi, Monique Dauge, and Vivette Girault. Vector potentials in three-dimensional non- smooth domains. Mathematical Methods in the Applied Sci- ences, 1998

    work page 1998

  3. [3]

    Lp-theory for vector potentials and sobolev’s inequalities for vector fields: Application to the Stokes equations with pressure boundary conditions

    Ch ´erif Amrouche and Nour El Houda Seloula. Lp-theory for vector potentials and sobolev’s inequalities for vector fields: Application to the Stokes equations with pressure boundary conditions. Mathematical Models and Methods in Applied Sciences, 2013

    work page 2013

  4. [4]

    Advanced normalization tools (ANTs)

    Brian B Avants, Nick Tustison, Gang Song, et al. Advanced normalization tools (ANTs). Insight j, 2009. 5

    work page 2009

  5. [5]

    Guttag, and Adrian V

    Guha Balakrishnan, Amy Zhao, Mert Rory Sabuncu, John V . Guttag, and Adrian V . Dalca. V oxelMorph: A learning framework for deformable medical image registration. IEEE Transactions on Medical Imaging, 2018. 1

    work page 2018

  6. [6]

    Autoencoders for unsuper- vised anomaly segmentation in brain MR images: a compar- ative study

    Christoph Baur, Stefan Denner, Benedikt Wiestler, Nassir Navab, and Shadi Albarqouni. Autoencoders for unsuper- vised anomaly segmentation in brain MR images: a compar- ative study. Medical Image Analysis, 2021. 6, 7, 8

    work page 2021

  7. [7]

    Greve, Oula Puonti, Axel Thielscher, Koen Van Leemput, Bruce R

    Benjamin Billot, Douglas N. Greve, Oula Puonti, Axel Thielscher, Koen Van Leemput, Bruce R. Fischl, et al. Synth- Seg: Segmentation of brain MRI scans of any contrast and resolution without retraining. Medical Image Analysis, 2021. 2, 4, 15

    work page 2021

  8. [8]

    Robust machine learning segmentation for large-scale analysis of heterogeneous clinical brain MRI datasets

    Benjamin Billot, Colin Magdamo, You Cheng, Steven E Arnold, Sudeshna Das, and Juan Eugenio Iglesias. Robust machine learning segmentation for large-scale analysis of heterogeneous clinical brain MRI datasets. Proceedings of the National Academy of Sciences, 2023. 2

    work page 2023

Show all 75 references
  1. [9]

    Matthew R. G. Brown, Gagan Preet Singh Sidhu, Russell Greiner, Nasimeh Asgarian, Meysam Bastani, Peter H. Sil- verstone, et al. ADHD-200 global competition: diagnos- ing ADHD using personal characteristic data can outperform resting state fMRI measurements. Frontiers in System...

    2012

  2. [10]

    Deep learning for physical processes: Incorporating prior scientific knowledge

    Emmanuel de B ´ezenac, Arthur Pajot, and Patrick Gallinari. Deep learning for physical processes: Incorporating prior scientific knowledge. In ICLR, 2018. 2, 3

    2018

  3. [11]

    de V os, Floris F

    Bob D. de V os, Floris F. Berendsen, Max A. Viergever, Hes- sam Sokooti, Marius Staring, and Ivana Iˇsgum. A deep learn- ing framework for unsupervised affine and deformable image registration. Medical Image Analysis, 2019. 1

    2019

  4. [12]

    Learning general- purpose biomedical volume representations using random- ized synthesis

    Neel Dey, Benjamin Billot, Hallee E Wong, Clinton J Wang, Mengwei Ren, P Ellen Grant, et al. Learning general- purpose biomedical volume representations using random- ized synthesis. arXiv, abs/2411.02372, 2024. 2

    2024 arXiv

  5. [13]

    Local temperature scaling for probability calibration

    Zhipeng Ding, Xu Han, Peirong Liu, and Marc Nietham- mer. Local temperature scaling for probability calibration. In ICCV, 2021. 1

    2021

  6. [14]

    Discrete vector potential representation of a divergence-free vector field in three-dimensional domains: Numerical analysis of a model problem

    Francois Dubois. Discrete vector potential representation of a divergence-free vector field in three-dimensional domains: Numerical analysis of a model problem. SIAM Journal on Numerical Analysis, 1990

    1990

  7. [15]

    Van Essen, K ˆamil U ˘gurbil, Edward J

    David C. Van Essen, K ˆamil U ˘gurbil, Edward J. Auer- bach, Deanna M. Barch, Timothy Edward John Behrens, Richard D. Bucholz, et al. The human connectome project: A data acquisition perspective. NeuroImage, 2012. 5, 6, 7, 8, 14, 15

    2012

  8. [16]

    Whole brain segmentation: automated labeling of neuroanatomical struc- tures in the human brain

    Bruce Fischl, David H Salat, Evelina Busa, Marilyn Albert, Megan Dieterich, Christian Haselgrove, et al. Whole brain segmentation: automated labeling of neuroanatomical struc- tures in the human brain. Neuron, 2002. 5, 15

    2002

  9. [17]

    Fischl, David H

    Bruce R. Fischl, David H. Salat, Evelina Busa, Marilyn S. Albert, Megan Dieterich, Christian Haselgrove, et al. Whole brain segmentation automated labeling of neuroanatomical structures in the human brain. Neuron, 2002. 2

    2002

  10. [18]

    Rainey-Smith, Sabine M

    Christopher Fowler, Stephanie R. Rainey-Smith, Sabine M. Bird, Julia Bomke, Pierrick T. Bourgeat, et al. Fifteen years of the australian imaging, biomarkers and lifestyle (AIBL) study: Progress and observations from 2,359 older adults spanning the spectrum from cognitive norma...

  11. [19]

    Learning to estimate single-view volumetric flow motions without 3d supervision

    Erik Franz, Barbara Solenthaler, and Nils Thuerey. Learning to estimate single-view volumetric flow motions without 3d supervision. In ICLR, 2023. 3

    2023

  12. [20]

    Gottlieb

    Sigal Gottlieb and Lee-Ad J. Gottlieb. Strong stability pre- serving properties of Runge-Kutta time discretization meth- ods for linear constant coefficient operators. Journal of Sci- entific Computing, 2003. 4

    2003

  13. [21]

    Un- supervised 3d out-of-distribution detection with latent diffu- sion models

    Mark S Graham, Walter Hugo Lopez Pinaya, Paul Wright, Petru-Daniel Tudosiu, Yee H Mah, James T Teo, et al. Un- supervised 3d out-of-distribution detection with latent diffu- sion models. In MICCAI, 2023. 6, 7, 8

    2023

  14. [22]

    ISLES 2022: A multi-center magnetic resonance imag- ing stroke lesion segmentation dataset

    Moritz R Hernandez Petzsche, Ezequiel de la Rosa, Uta Hanning, Roland Wiest, Waldo Valenzuela, Mauricio Reyes, et al. ISLES 2022: A multi-center magnetic resonance imag- ing stroke lesion segmentation dataset. Scientific data, 2022. 4, 6, 7, 14

    2022

  15. [23]

    Greve, Juan Eugenio Iglesias, Bruce R

    Malte Hoffmann, Benjamin Billot, Douglas N. Greve, Juan Eugenio Iglesias, Bruce R. Fischl, and Adrian V . Dalca. SynthMorph: Learning contrast-invariant registration with- out acquired images. IEEE Transactions on Medical Imag- ing, 2020. 2

    2020

  16. [24]

    Mora, Adrian V

    Andrew Hoopes, Jocelyn S. Mora, Adrian V . Dalca, Bruce R. Fischl, and Malte Hoffmann. SynthStrip: skull-stripping for any brain image. NeuroImage, 2022. 2, 14

    2022

  17. [25]

    Modern diagnostic imaging technique applica- tions and risk factors in the medical field: a review

    Shah Hussain, Iqra Mubeen, Niamat Ullah, Syed Shahab Ud Din Shah, Bakhtawar Abduljalil Khan, Muhammad Za- hoor, et al. Modern diagnostic imaging technique applica- tions and risk factors in the medical field: a review. BioMed research international, 2022. 1

    2022

  18. [26]

    Magdamo, Steve Arnold, Sudeshna Das, et al

    Juan Eugenio Iglesias, Benjamin Billot, Yael Balbastre, Colin G. Magdamo, Steve Arnold, Sudeshna Das, et al. Synthsr: A public AI tool to turn heterogeneous clinical brain scans into high-resolution T1-weighted images for 3D morphometry. Science Advances, 2023. 2, 4, 6, 7, 8, 15, 16

    2023

  19. [27]

    Alexander, et al

    Juan Eugenio Iglesias, Benjamin Billot, Yael Balbastre, Azadeh Tabari, John Conklin, Daniel C. Alexander, et al. Joint super-resolution and synthesis of 1 mm isotropic MP- RAGE volumes from clinical MRI exams with scans of dif- ferent orientation, resolution and contrast. NeuroImage,

  20. [28]

    Jack, Matt A

    Clifford R. Jack, Matt A. Bernstein, Nick C Fox, Paul M. Thompson, Gene E. Alexander, Danielle J. Harvey, et al. The 9 Alzheimer’s disease neuroimaging initiative (ADNI): MRI methods. Journal of Magnetic Resonance Imaging, 2008. 5, 6, 7, 8, 14, 15

    2008

  21. [29]

    Konstantinos Kamnitsas, Christian Ledig, Virginia F. J. Newcombe, Joanna P. Simpson, Andrew D. Kane, David K. Menon, et al. Efficient multi-scale 3D CNN with fully con- nected CRF for accurate brain lesion segmentation. Medical Image Analysis, 2016. 1

    2016

  22. [30]

    LaMontagne, Sarah J

    Pamela J. LaMontagne, Sarah J. Keefe, Wallace Lauren, Chengjie Xiong, Elizabeth A. Grant, Krista L. Moulder, et al. OASIS-3: Longitudinal neuroimaging, clinical, and cognitive dataset for normal aging and alzheimer’s disease. Alzheimer’s & Dementia, 2018. 5, 6, 7, 14, 15

    2018

  23. [31]

    Quantifying white matter hyperintensity and brain volumes in heterogeneous clinical and low-field portable MRI

    Pablo Laso, Stefano Cerri, Annabel Sorby-Adams, Jennifer Guo, Farrah Mateen, Philipp Goebl, et al. Quantifying white matter hyperintensity and brain volumes in heterogeneous clinical and low-field portable MRI. In ISBI, 2024. 4

    2024

  24. [32]

    A unifying framework for partial volume segmentation of brain MR images

    Koenraad Van Leemput, Frederik Maes, Dirk Vandermeulen, and Paul Suetens. A unifying framework for partial volume segmentation of brain MR images. IEEE Transactions on Medical Imaging, 2003. 2

    2003

  25. [33]

    Randall J. LeVeque. Finite Volume Methods for Hyperbolic Problems. Cambridge University Press, 2002. 4

    2002

  26. [34]

    Trivializations for gradient-based optimization on manifolds

    Mario Lezcano-Casado. Trivializations for gradient-based optimization on manifolds. In NeurIPS, 2019

    2019

  27. [35]

    Cheap orthogonal constraints in neural networks: A simple parametrization of the orthogonal and unitary group

    Mario Lezcano-Casado and David Mart ´ınez-Rubio. Cheap orthogonal constraints in neural networks: A simple parametrization of the orthogonal and unitary group. In ICML, 2019

    2019

  28. [36]

    A large, open source dataset of stroke anatomical brain images and manual lesion segmentations

    Sook-Lei Liew, Julia M Anglin, Nick W Banks, Matt Sondag, Kaori L Ito, Hosung Kim, et al. A large, open source dataset of stroke anatomical brain images and manual lesion segmentations. Scientific data, 2018. 4, 6, 7, 8, 14

    2018

  29. [37]

    Lee, Stephen R

    Peirong Liu, Yueh Z. Lee, Stephen R. Aylward, and Marc Niethammer. Perfusion imaging: An advection diffusion ap- proach. IEEE Transactions on Medical Imaging , 2021. 2, 3

    2021

  30. [38]

    Alexander, and Juan E

    Peirong Liu, Oula Puonti, Xiaoling Hu, Daniel C. Alexander, and Juan E. Iglesias. Brain-ID: Learning contrast-agnostic anatomical representations for brain imaging. In ECCV,

  31. [39]

    PEPSI: Pathology-enhanced pulse-sequence-invariant representations for brain MRI

    Peirong Liu, Oula Puonti, Annabel Sorby-Adams, William T Kimberly, and Juan E Iglesias. PEPSI: Pathology-enhanced pulse-sequence-invariant representations for brain MRI. In MICCAI, 2024. 2, 4, 6, 7, 16

    2024

  32. [40]

    Discovering hidden physics be- hind transport dynamics

    Peirong Liu, Lin Tian, Yubo Zhang, Stephen Aylward, Yueh Lee, and Marc Niethammer. Discovering hidden physics be- hind transport dynamics. In CVPR, 2021. 3, 4

    2021

  33. [41]

    On the application of the Helmholtz- Hodge decomposition in projection methods for incompress- ible flows with general boundary conditions

    Filippo Maria Denaro. On the application of the Helmholtz- Hodge decomposition in projection methods for incompress- ible flows with general boundary conditions. International Journal for Numerical Methods in Fluids, 2003

    2003

  34. [42]

    The multimodal brain tumor image segmentation benchmark (BRATS)

    Bjoern H Menze, Andras Jakab, Stefan Bauer, Jayashree Kalpathy-Cramer, Keyvan Farahani, Justin Kirby, et al. The multimodal brain tumor image segmentation benchmark (BRATS). IEEE Transactions on Medical Imaging, 2014. 3

    2014

  35. [43]

    V-Net: Fully convolutional neural networks for volumetric medical image segmentation

    Fausto Milletar `ı, Nassir Navab, and Seyed-Ahmad Ahmadi. V-Net: Fully convolutional neural networks for volumetric medical image segmentation. 3DV, 2016. 1

    2016

  36. [44]

    Fast free-form deformation using graphics processing units

    Marc Modat, Gerard R Ridgway, Zeike A Taylor, Manja Lehmann, Josephine Barnes, David J Hawkes, et al. Fast free-form deformation using graphics processing units. Computer methods and programs in biomedicine, 2010. 4, 5

    2010

  37. [45]

    Ridgway, Zeike A

    Marc Modat, Gerard R. Ridgway, Zeike A. Taylor, Manja Lehmann, Josephine Barnes, David John Hawkes, et al. Fast free-form deformation using graphics processing units. Computer Methods and Programs in Biomedicine, 2010. 14, 15

    2010

  38. [46]

    Krumholz, Jure Leskovec, Eric J

    Michael Moor, Oishi Banerjee, Zahra F H Abad, Harlan M. Krumholz, Jure Leskovec, Eric J. Topol, and Pranav Ra- jpurkar. Foundation models for generalist medical artificial intelligence. Nature, 2023. 2

    2023

  39. [47]

    Improving noise

    Ken Perlin. Improving noise. In PACMCGIT, 2002. 12

    2002

  40. [48]

    Fast and sequence-adaptive whole-brain segmen- tation using parametric bayesian modeling

    Oula Puonti, Juan Eugenio Iglesias, and Koenraad Van Leemput. Fast and sequence-adaptive whole-brain segmen- tation using parametric bayesian modeling. NeuroImage,

  41. [49]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. In MICCAI, 2015. 1, 6, 16

    2015

  42. [50]

    Nonrigid reg- istration using free-form deformations: application to breast MR images

    Daniel Rueckert, Luke I Sonoda, Carmel Hayes, Derek LG Hill, Martin O Leach, and David J Hawkes. Nonrigid reg- istration using free-form deformations: application to breast MR images. IEEE Transactions on Medical Imaging, 1999. 4, 5

    1999

  43. [51]

    Accurate point cloud registration with robust optimal transport

    Zhengyang Shen, Jean Feydy, Peirong Liu, Ariel H Curiale, Ruben San Jose Estepar, Raul San Jose Estepar, and Marc Niethammer. Accurate point cloud registration with robust optimal transport. NeurIPS, 2021. 2

    2021

  44. [52]

    Large language models encode clinical knowledge

    Karan Singhal, Shekoofeh Azizi, Tao Tu, Said Mahdavi, Ja- son Wei, Hyung Won Chung, et al. Large language models encode clinical knowledge. Nature, 2022. 2

    2022

  45. [53]

    PWC-net: CNNs for optical flow using pyramid, warping, and cost volume

    Deqing Sun, Xiaodong Yang, Ming-Yu Liu, and Jan Kautz. PWC-net: CNNs for optical flow using pyramid, warping, and cost volume. In CVPR, 2018. 2

    2018

  46. [54]

    Worrall, Enrico Kaden, and Daniel C

    Ryutaro Tanno, Daniel E. Worrall, Enrico Kaden, and Daniel C. Alexander. Uncertainty modelling in deep learning for safer neuroimage enhancement: Demonstration in diffu- sion MRI. NeuroImage, 2020. 1

    2020

  47. [55]

    Raft: Recurrent all-pairs field transforms for optical flow

    Zachary Teed and Jia Deng. Raft: Recurrent all-pairs field transforms for optical flow. In ECCV, 2020. 2

    2020

  48. [56]

    Fluid registration between lung CT and stationary chest to- mosynthesis images

    Lin Tian, Connor Puett, Peirong Liu, Zhengyang Shen, Stephen R Aylward, Yueh Z Lee, and Marc Niethammer. Fluid registration between lung CT and stationary chest to- mosynthesis images. In MICCAI, 2020. 2

    2020

  49. [57]

    Fultz, et al

    Qiyuan Tian, Berkin Bilgic ¸, Qiuyun Fan, Chanon Ngam- sombat, Natalia Zaretskaya, Nina E. Fultz, et al. Improving in vivo human cerebral cortical surface reconstruction using data-driven super-resolution. Cerebral Cortex, 2020. 1

    2020

  50. [58]

    Towards generalist biomedical ai

    Tao Tu, Shekoofeh Azizi, Danny Driess, Mike Schaeker- mann, Mohamed Amin, Pi-Chuan Chang, et al. Towards generalist biomedical ai. arXiv, abs/2307.14334, 2023. 2 10

    2023 arXiv

  51. [59]

    SAM-Med3D-MoE: Towards a non- forgetting segment anything model via mixture of experts for 3D medical image segmentation

    Guoan Wang, Jin Ye, Junlong Cheng, Tianbin Li, Zhaolin Chen, Jianfei Cai, et al. SAM-Med3D-MoE: Towards a non- forgetting segment anything model via mixture of experts for 3D medical image segmentation. In MICCAI, 2024. 2

    2024

  52. [60]

    Deep visual domain adapta- tion: A survey

    Mei Wang and Weihong Deng. Deep visual domain adapta- tion: A survey. Neurocomputing, 2018. 1

    2018

  53. [61]

    Weiner, Dallas P Veitch, Paul S

    Michael W. Weiner, Dallas P Veitch, Paul S. Aisen, Lau- rel A Beckett, Nigel J. Cairns, Robert C. Green, et al. The Alzheimer’s disease neuroimaging initiative 3: Continued innovation for clinical trial improvement.Alzheimer’s & De- mentia, 2017. 5, 6, 7, 8, 14, 15

    2017

  54. [62]

    HelmFluid: Learning helmholtz dynamics for interpretable fluid prediction

    Lanxiang Xing, Haixu Wu, Yuezhou Ma, Jianmin Wang, and Mingsheng Long. HelmFluid: Learning helmholtz dynamics for interpretable fluid prediction. In ICML, 2024. 3

    2024

  55. [63]

    Quick- silver: Fast predictive image registration – a deep learning approach

    Xiao Yang, Roland Kwitt, and Marc Niethammer. Quick- silver: Fast predictive image registration – a deep learning approach. NeuroImage, 2017. 1

    2017

  56. [64]

    Quicksilver: Fast predictive image registration–a deep learning approach

    Xiao Yang, Roland Kwitt, Martin Styner, and Marc Nietham- mer. Quicksilver: Fast predictive image registration–a deep learning approach. NeuroImage, 2017. 2

    2017

  57. [65]

    Ayhan, Dominic J Williamson, Robbert R

    Yukun Zhou, Mark A Chia, Siegfried Karl Wagner, Murat S. Ayhan, Dominic J Williamson, Robbert R. Struyven, et al. A foundation model for generalizable disease detection from retinal images. Nature, 2023. 2 11 Unraveling Normal Anatomy via Fluid-Driven Anomaly Randomization (Ap...

    2023

  58. [67]

    Generate a grid of random gradients at lattice points (Line 25-30)

  59. [68]

    Compute the relative position of the point inside the grid cell (Line 31-44)

  60. [69]

    Calculate the dot product of the gradients and the relative position vectors (45-52)

  61. [70]

    Apply the fade function to smooth the interpolation (Line 54-61)

  62. [71]

    Interpolate between dot products to get a smooth value (Line 62-69)

  63. [72]

    generateperlin noise 3d

    Threshold to get a random shape of anomaly profile (Line 71-78). 1 import os, time 2 import numpy as np 3 4 def interpolant(t): 5 return t *t*t*(t*(t*6 - 15) + 10) 6 7 def generate_perlin_noise_3d(shape, res, tileable=(False, False, False), interpolant=interpolant, percentile=...

  64. [73]

    Generate three individual Perlin noise maps for the potential ( Ψ) construction (Line 6-9)

  65. [74]

    Reshape the noise map to match the current subject sample’s patch size (Line 11-23)

  66. [75]

    Surjectively map the random potential to its corresponding incompressible flow space via ?? (Line 25-28). 1 import torch 2 3 def generate_velocity_3d(shape, perlin_res, V_multiplier, device, save_orig_for_visualize = False): 4 pad_shape = [ 200, 200, 200 ] 5 13 6 # Generate ra...

    2022

  67. [2024]

    2, 4, 6, 7, 8, 15, 16

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.