Pith. sign in

REVIEW 2 major objections 5 minor 32 references

Predicting future brain MRI directly in voxel space with a coarse-to-fine transformer outperforms latent-space and single-pathway baselines, the paper claims.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 06:02 UTC pith:SSZFZZOL

load-bearing objection Novel voxel-space hierarchy, but the SOTA claim is undercut by undisclosed loss weights and missing significance tests. the 2 major comments →

arxiv 2607.27537 v1 pith:SSZFZZOL submitted 2026-07-30 cs.CV

ProgFormer: Hierarchical Voxel Diffusion Transformer for Longitudinal Brain MRI Prediction

Dexuan Ding , Yuankai Qi , Luping Zhou , Jian Yang , Quan Z. Sheng , Ming-Hsuan Yang This is my paper
classification cs.CV
keywords longitudinal brain MRIdiffusion transformerflow matchingvoxel-space generationAlzheimer's diseasemedical image synthesiscoarse-to-finetrajectory prediction
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper attempts to establish that longitudinal brain MRI prediction is better done directly in voxel space, avoiding the information loss of latent autoencoders, but with a two-pathway design that separates global structural modelling from local refinement. It introduces ProgFormer, where a coarse pathway reasons over 3D patch tokens with spatial and causal temporal attention to set the whole-brain structure, and a fine pathway refines individual voxels within each patch using the coarse representation as spatio-temporal grounding. The two pathways jointly predict a velocity field via conditional flow matching, and the future scan is generated by Euler integration from Gaussian noise. On ADNI, AIBL, and OASIS, the paper reports the best pairwise PSNR on all three cohorts and the best regional accuracy for the interleaved variant, along with trajectory improvements from observed history. A sympathetic reader would care because the method addresses a real trade-off: preserving stable anatomy while capturing subtle, localized disease-driven changes.

Core claim

The central claim is that hierarchical voxel-space diffusion modelling can outperform both latent-space and unified voxel-space predictors for longitudinal brain MRI. ProgFormer's coarse pathway operates on non-overlapping p×p×p patch tokens, applying spatial attention within each scan and causal temporal attention across observed scans to encode whole-brain structure and longitudinal context. The fine pathway treats each voxel in the flow state as a token, applies voxel-level self-attention within each patch, and uses coarse-to-fine cross-attention to condition voxel refinement on the patch-level representation. The predicted velocity is the sum of the coarse and fine velocities, and traini

What carries the argument

The central mechanism is the hierarchical velocity-field decomposition: a single network predicts a voxel-space velocity νθ = νcoarse + νfine through conditional flow matching, where the coarse pathway produces a patch-level velocity from spatio-temporal token attention and the fine pathway produces a voxel-level velocity guided by cross-attention to the coarse features. This avoids a separately learned autoencoder and lets global structure and local change be modelled by separate pathways within one end-to-end network. The weighting of the loss by SynthSeg regional labels is a second load-bearing component that directs training toward progression-relevant tissue.

Load-bearing premise

The method's advantage relies on hand-assigned per-region loss weights λ_r, derived from SynthSeg labels, whose specific values are not reported; if these weights are not reproducible or transferable to other cohorts, the reported performance gains may not hold.

What would settle it

Run ProgFormer-S on the ADNI pairwise split with all regional weights λ_r set to 1 and report PSNR/SSIM; if the resulting drop is substantially smaller than the 1.64 dB and 5.96 SSIM points reported in Table 2, the strong dependence on the manual weighting—and hence the claim that the hierarchical architecture alone delivers the gain—is falsified.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • The reported pairwise PSNR gains (0.20–0.28 dB over the strongest baseline) on ADNI, AIBL, and OASIS would establish ProgFormer-S as the best single-scan predictor among the compared methods.
  • Using the full observed history in the trajectory setting improves PSNR by up to 1.28 dB on ADNI over single-scan input, showing that the causal temporal attention is extracting useful longitudinal signal.
  • The ablation study attributes the bulk of the performance to the coarse pathway: removing it collapses PSNR from roughly 26–28 dB to about 7 dB, confirming that patch-level spatio-temporal reasoning is the backbone of the method.
  • The fine pathway and coarse-to-fine cross-attention contribute 1.4–1.8 dB and 2–8 SSIM points, and their contribution grows under trajectory input, indicating that coarse grounding matters more when history is available.
  • The interleaved variant achieves the lowest regional MAE on all three cohorts, suggesting a trade-off between image-level fidelity (sequential) and regional structure (interleaved).

Where Pith is reading between the lines

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

  • The unreported values of the regional loss weights λ_r are a reproducibility gap: if they were tuned on one cohort and then applied to others, part of the reported advantage could be dataset-specific; a uniform-weight ablation is the natural check.
  • The causal temporal attention receives scans in chronological order with zero-padding to a common length; a natural next experiment is varying the number of observed scans to see whether prediction error saturates or the model learns to down-weight old scans.
  • The same two-pathway recipe could transfer to other longitudinal volumetric imaging tasks—such as cardiac MRI or tumour progression—where the majority of the volume is static and changes are localized; the paper does not test this, but the mechanism is architecture-agnostic.
  • Because the fine pathway operates on individual patches with coarse-to-fine attention but no cross-patch voxel attention, the model may be limited in refining structures that span patch boundaries; increasing patch size or adding fine-level cross-patch attention is a testable modification.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 5 minor

Summary. The paper proposes ProgFormer, a hierarchical voxel-space diffusion transformer for longitudinal brain MRI prediction. The model combines a coarse patch-level pathway that performs spatial and causal temporal attention across observed scans with a fine voxel-level pathway that refines local structure under coarse-to-fine cross-attention. Both pathways jointly estimate a velocity field via conditional flow matching, and the future scan is obtained by integrating this field with Euler steps, directly in voxel space and without a separately learned autoencoder. The method is evaluated on ADNI, AIBL, and OASIS under pairwise and trajectory settings, reporting the best PSNR on all three cohorts for the sequential variant, with competitive SSIM and regional MAE against BrLP, CounterSynth, TADM-3D, and SADM.

Significance. If the results hold, the paper makes a useful contribution to longitudinal brain MRI prediction by demonstrating that direct voxel-space diffusion modeling can be made effective through a hierarchical coarse-to-fine architecture, avoiding the information loss of latent autoencoders. The flow-matching formulation is standard and the ablations (Table 2) support the importance of the coarse pathway and coarse-to-fine cross-attention. The evaluation across three cohorts and two settings is thorough and includes clear preprocessing and implementation details. However, the central state-of-the-art claim is currently fragile: the reported advantage over baselines is small (0.20–0.28 dB PSNR), no significance tests are provided, and the method's performance depends strongly on an unreported, manually chosen regional weighting in the training loss. These issues must be resolved before the claim can be accepted.

major comments (2)
  1. [§3.7, Eq. (21) and §4.3, Table 2] The regional weights λ_r in the weighted flow-matching objective are never reported. The ablation in Table 2 shows that removing them degrades PSNR by 1.64 dB (pairwise) and 3.45 dB (trajectory) on ADNI, while the claimed margins over the strongest baselines are only 0.20–0.28 dB. This implies that the state-of-the-art result is not an inherent property of the hierarchical architecture but is contingent on a hand-chosen weighting scheme whose values, and tuning procedure, are undisclosed. The authors must report the λ_r values, provide code or a sensitivity analysis, and present results with uniform weights. Additionally, because the same anatomical regions emphasized in the loss (hippocampus, ventricles, etc.) are used to compute R-MAE, the regional metric is partially aligned with the training objective; this potential circularity should be explicitly discussed.
  2. [§4.1, Table 1] No statistical significance testing is reported. The PSNR differences between ProgFormer-S and the strongest baseline are within one standard deviation on all three cohorts (e.g., 26.39±1.68 vs. 26.19±2.64 on ADNI). Without paired significance tests (e.g., Wilcoxon signed-rank or bootstrap confidence intervals over test subjects) for PSNR, SSIM, and R-MAE, the claim that ProgFormer achieves state-of-the-art performance is not supported. The authors should add such tests for all pairwise and trajectory comparisons.
minor comments (5)
  1. [Eq. (22)] The condition vector in Equation (22) is written as c_{T+1}, but the covariate vector is defined as c_T in Eq. (1) and the text says c_T remains fixed during inference. This is likely a typo and should be corrected.
  2. [§4.3, Table 2] The ablation 'w/o Coarse Pathway' is difficult to interpret because the fine pathway in the full model receives coarse representations via cross-attention. With the coarse pathway removed, the fine pathway has no coarse context; the resulting collapse (PSNR 7.23) is expected and not informative. Please clarify what variant is used in this ablation.
  3. [Table 1] The row label 'T SADM' is not defined in the table caption or text. It is later explained that SADM is evaluated in single-input (I=1) and multiple-input (I=T) configurations. Please label this row clearly (e.g., 'SADM (I=T)') and, for completeness, report the single-input SADM trajectory result as well.
  4. [§4.2, Trajectory prediction] The sentence 'Compared with the strongest I=1 result for each metric, using the observed history improves PSNR by 1.28 on ADNI, 0.17 on AIBL, and 0.79 on OASIS' is ambiguous about the comparison basis. From Table 1, the difference between ProgFormer-S pairwise and trajectory on ADNI is 1.51 dB, not 1.28. Please reconcile these numbers and specify exactly which baseline and protocol (pairwise vs I=1 trajectory) are being compared.
  5. [Figure 6] The colorbar for the absolute error maps is not shown, making it difficult to gauge the magnitude of the differences. Consider adding a color scale or normalizing the maps consistently.

Circularity Check

0 steps flagged

No circular derivation: predictions are evaluated against held-out external baselines; the unreported λ_r weights are a reproducibility concern, not circularity.

full rationale

ProgFormer's derivation chain is self-contained. Conditional flow matching defines the supervision target as ν* = y_1 − y_0 (Eq. 4), the model learns ν_θ, and inference integrates this velocity field with Euler steps (Eqs. 22–23). No fitted parameter or predefined quantity is renamed as a prediction. The training loss (Eq. 21) weights voxels according to SynthSeg labels via manually predefined λ_r; these weights are inputs to the loss, not outputs of the model, and the paper reports results on held-out test partitions of ADNI, AIBL, and OASIS against external methods. The overlap between the weighted anatomical regions and those used in the R-MAE metric means the training objective is aligned with one evaluation metric, but that is a standard objective design choice rather than a circular reduction: the main reported claims (PSNR, SSIM, R-MAE) are not forced by construction, and the loss itself does not contain the evaluation outcome. The ablation (Table 2) shows the method is sensitive to the region weights, and the failure to report the λ_r values is a genuine reproducibility and robustness limitation, but it is not an instance of circular reasoning under the criteria of this pass. No load-bearing argument reduces to a self-citation, an imported uniqueness theorem, or a renamed known result.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

The central claim rests on standard flow matching, the availability of accurate SynthSeg segmentations for loss weighting, and several architectural assumptions (additive velocity composition, zero-padded causal attention) that are not derived from first principles.

free parameters (4)
  • λ_r regional loss weights
    Hand-assigned weights for SynthSeg labels in the flow-matching loss; ablation shows they contribute 1.64–3.45 dB PSNR, but values are not reported.
  • Patch size p = 4
    Hand-chosen patch size controlling coarse/fine token granularity; no sensitivity analysis.
  • Euler steps K = 100
    Inference steps; trade-off between compute and quality, no ablation.
  • Number of coarse/fine blocks and hidden dims = L_c=8, L_f=4, d_c=384, d_f=256
    Architecture hyperparameters chosen without reported tuning or sensitivity.
axioms (5)
  • standard math Conditional flow matching with linear interpolation (Eq. 3-4) transports Gaussian noise to the target MRI distribution.
    Background from Lipman et al. 2023; used as the generative framework.
  • domain assumption SynthSeg accurately segments brain structures in the ground-truth future scans and produces reliable labels for loss weighting.
    The entire loss-weighting scheme (Eq. 21) depends on SynthSeg label maps; errors would misweight the objective. Section 3.7.
  • domain assumption Affine registration to MNI152 and intensity normalization make longitudinal scans comparable and remove non-biological variation.
    All methods use this preprocessing; if misalignment remains, local changes such as hippocampal atrophy may be confounded. Section 4.1.
  • ad hoc to paper The additive composition of coarse and fine velocities (Eq. 20) is sufficient to represent the true velocity field.
    The architecture assumes the target velocity factorizes as a sum of patch-level and voxel-level components; no theoretical guarantee. Eq. 20.
  • ad hoc to paper Causal temporal attention with zero-padded shorter sequences preserves longitudinal context.
    The model assumes that attending to all past scans in order is beneficial and that padding does not introduce bias. Section 3.4.

pith-pipeline@v1.3.0-daily-deepseek · 13358 in / 11122 out tokens · 101037 ms · 2026-08-01T06:02:02.045357+00:00 · methodology

0 comments
read the original abstract

Predicting future structural MRI of a brain is challenging because longitudinal changes are often subtle and confined to specific anatomical regions, while most subject-specific brain structure remains stable over time. An effective model should therefore preserve global brain structural consistency while remaining sensitive to fine-grained disease progression. Existing latent-space-based methods improve computational efficiency, but suffer from information loss during their compression-reconstruction procedure. In contrast, direct voxel-space methods avoid latent reconstruction but commonly use a unified prediction pathway to model brain structure and progression-related changes. Subtle local changes may therefore be overshadowed by the dominant stable brain structure. To address these challenges, we propose ProgFormer, a hierarchical voxel-space Diffusion Transformer for longitudinal brain MRI prediction. ProgFormer uses a coarse pathway to perform the primary volumetric prediction from 3D patch tokens. This pathway models overall brain structure and longitudinal context. The fine pathway then uses the coarse representations as spatio-temporal grounding for voxel-level refinement within individual patches. The two pathways jointly estimate a velocity field directly in voxel space through conditional flow matching, enabling end-to-end prediction without a separately learned image autoencoder. The predicted future scan is then generated from Gaussian noise by integrating the estimated velocity field over a sequence of Euler steps. Extensive experimental results on three widely used benchmarks, ADNI, AIBL, and OASIS, under both pairwise and trajectory settings demonstrate favourable performance compared against several state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2607.27537 by Dexuan Ding, Jian Yang, Luping Zhou, Ming-Hsuan Yang, Quan Z. Sheng, Yuankai Qi.

Figure 1
Figure 1. Figure 1: Longitudinal brain MRI prediction paradigms. (a) [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of ProgFormer. (a) ProgFormer processes the observed scans and current flow state through a coarse patch [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: AdaLN-gated attention and Euler inference. Left: [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Sequential and interleaved pathway composi￾tion. to fine without feeding fine features back into subsequent coarse stages. The sequential variant transfers the final coarse representation throughout voxel refinement. The interleaved variant instead dis￾tributes coarse-to-fine transfer across network depth. We use the sequential variant as the primary model and compare both compo￾sitions experimentally. 3.7… view at source ↗
Figure 6
Figure 6. Figure 6: Coarse-to-fine refinement. The top row shows the [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗

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

32 extracted references · 3 canonical work pages · 2 internal anchors

  1. [1]

    Hypothetical model of dynamic biomarkers of the Alzheimer's pathological cascade

    Jack, Clifford R Jr and Knopman, David S and Jagust, William J and Shaw, Leslie M and Aisen, Paul S and Weiner, Michael W and Petersen, Ronald C and Trojanowski, John Q , journal =. Hypothetical model of dynamic biomarkers of the Alzheimer's pathological cascade. , volume =

  2. [2]

    The clinical use of structural MRI in Alzheimer disease

    Frisoni, Giovanni B and Fox, Nick C and Jack, Clifford R Jr and Scheltens, Philip and Thompson, Paul M , journal =. The clinical use of structural MRI in Alzheimer disease. , volume =

  3. [3]

    The boundary shift integral: an accurate and robust measure of cerebral volume changes from registered repeat MRI

    Freeborough, P A and Fox, N C , journal =. The boundary shift integral: an accurate and robust measure of cerebral volume changes from registered repeat MRI. , volume =

  4. [4]

    and Rueckert, Daniel , journal =

    Ledig, Christian and Schuh, Andreas and Guerrero, Ricardo and Heckemann, Rolf A. and Rueckert, Daniel , journal =. Structural brain imaging in Alzheimer's disease and mild cognitive impairment: biomarker analysis and shared morphometry database , volume =

  5. [5]

    and Ravì, Daniele , title =

    Puglisi, Lemuel and Alexander, Daniel C. and Ravì, Daniele , title =. proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2024 , year =

    2024

  6. [6]

    Generative Modeling of Neurodegenerative Brain Anatomy with 4D Longitudinal Diffusion Model

    Nivetha Jayakumar and Swakshar Deb and Bahram Jafrasteh and Qingyu Zhao and Miaomiao Zhang , title =. CoRR , volume =. 2026 , url =. doi:10.48550/ARXIV.2604.22700 , eprinttype =. 2604.22700 , timestamp =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2604.22700 2026

  7. [7]

    CoRR , volume =

    Hao Chen and Rui Yin and Yifan Chen and Qi Chen and Chao Li , title =. CoRR , volume =. 2025 , url =. doi:10.48550/ARXIV.2512.09185 , eprinttype =. 2512.09185 , timestamp =

    Pith/arXiv arXiv doi:10.48550/arxiv.2512.09185 2025

  8. [8]

    Temporally-aware diffusion model for brain progression modelling with bidirectional temporal regularisation , journal =

    Mattia Litrico and Francesco Guarnera and Mario Valerio Giuffrida and Daniele Rav. Temporally-aware diffusion model for brain progression modelling with bidirectional temporal regularisation , journal =. 2026 , url =. doi:10.1016/J.COMPMEDIMAG.2025.102688 , timestamp =

    arXiv 2026

  9. [9]

    Information Processing in Medical Imaging - 28th International Conference,

    Jee Seok Yoon and Chenghao Zhang and Heung. Information Processing in Medical Imaging - 28th International Conference,. 2023 , url =. doi:10.1007/978-3-031-34048-2\_30 , timestamp =

    doi:10.1007/978-3-031-34048-2 2023

  10. [10]

    CoRR , volume =

    Yongsheng Yu and Wei Xiong and Weili Nie and Yichen Sheng and Shiqiu Liu and Jiebo Luo , title =. CoRR , volume =. 2025 , url =. doi:10.48550/ARXIV.2511.20645 , eprinttype =. 2511.20645 , timestamp =

    Pith/arXiv arXiv doi:10.48550/arxiv.2511.20645 2025

  11. [11]

    2026 , eprint=

    PointDiT: Pixel-Space Diffusion for Monocular Geometry Estimation , author=. 2026 , eprint=

    2026

  12. [12]

    CoRR , volume =

    Tianhong Li and Kaiming He , title =. CoRR , volume =. 2025 , url =. doi:10.48550/ARXIV.2511.13720 , eprinttype =. 2511.13720 , timestamp =

    Pith/arXiv arXiv doi:10.48550/arxiv.2511.13720 2025

  13. [13]

    DiT-3D: Exploring Plain Diffusion Transformers for 3D Shape Generation , booktitle =

    Shentong Mo and Enze Xie and Ruihang Chu and Lanqing Hong and Matthias Nie. DiT-3D: Exploring Plain Diffusion Transformers for 3D Shape Generation , booktitle =. 2023 , url =

    2023

  14. [14]

    Gomez and Lukasz Kaiser and Illia Polosukhin , title =

    Ashish Vaswani and Noam Shazeer and Niki Parmar and Jakob Uszkoreit and Llion Jones and Aidan N. Gomez and Lukasz Kaiser and Illia Polosukhin , title =. CoRR , volume =. 2017 , url =. 1706.03762 , timestamp =

    Pith/arXiv arXiv 2017

  15. [15]

    Latte: Latent Diffusion Transformer for Video Generation , journal =

    Xin Ma and Yaohui Wang and Xinyuan Chen and Gengyun Jia and Ziwei Liu and Yuan. Latte: Latent Diffusion Transformer for Video Generation , journal =. 2025 , url =

    2025

  16. [16]

    CoRR , volume =

    William Peebles and Saining Xie , title =. CoRR , volume =. 2022 , url =. doi:10.48550/ARXIV.2212.09748 , eprinttype =. 2212.09748 , timestamp =

    Pith/arXiv arXiv doi:10.48550/arxiv.2212.09748 2022

  17. [17]

    Proceedings of the 38th International Conference on Machine Learning , pages =

    Is Space-Time Attention All You Need for Video Understanding? , author =. Proceedings of the 38th International Conference on Machine Learning , pages =. 2021 , editor =

    2021

  18. [18]

    Alzheimer's Disease Neuroimaging Initiative (ADNI): clinical characterization

    Petersen, R C and Aisen, P S and Beckett, L A and Donohue, M C and Gamst, A C and Harvey, D J and Jack, C R Jr and Jagust, W J and Shaw, L M and Toga, A W and Trojanowski, J Q and Weiner, M W , journal =. Alzheimer's Disease Neuroimaging Initiative (ADNI): clinical characterization. , volume =

  19. [19]

    Ellis, Kathryn A and Bush, Ashley I and Darby, David and De Fazio, Daniela and Foster, Jonathan and Hudson, Peter and Lautenschlager, Nicola T and Lenzo, Nat and Martins, Ralph N and Maruff, Paul and Masters, Colin and Milner, Andrew and Pike, Kerryn and Rowe, Christopher and Savage, Greg and Szoeke, Cassandra and Taddei, Kevin and Villemagne, Victor and ...

  20. [20]

    and Benzinger, Tammie LS

    LaMontagne, Pamela J. and Benzinger, Tammie LS. and Morris, John C. and Keefe, Sarah and Hornbeck, Russ and Xiong, Chengjie and Grant, Elizabeth and Hassenstab, Jason and Moulder, Krista and Vlassenko, Andrei G. and Raichle, Marcus E. and Cruchaga, Carlos and Marcus, Daniel , journal =. OASIS-3: Longitudinal Neuroimaging, Clinical, and Cognitive Dataset f...

  21. [21]

    N4ITK: improved N3 bias correction

    Tustison, Nicholas J and Avants, Brian B and Cook, Philip A and Zheng, Yuanjie and Egan, Alexander and Yushkevich, Paul A and Gee, James C , journal =. N4ITK: improved N3 bias correction. , volume =

  22. [22]

    Mora and Adrian V

    Andrew Hoopes and Jocelyn S. Mora and Adrian V. Dalca and Bruce Fischl and Malte Hoffmann , title =. NeuroImage , volume =. 2022 , url =. doi:10.1016/J.NEUROIMAGE.2022.119474 , timestamp =

    arXiv 2022

  23. [23]

    Evans and Kelly Botteron and C

    Vladimir Fonov and Alan C. Evans and Kelly Botteron and C. Robert Almli and Robert C. McKinstry and D. Louis Collins , journal =. Unbiased average age-appropriate atlases for pediatric studies , volume =

  24. [24]

    Jorge Cardoso and Sebastien Ourselin and Geraint Rees and John Ashburner and Parashkev Nachev , journal =

    Guilherme Pombo and Robert Gray and M. Jorge Cardoso and Sebastien Ourselin and Geraint Rees and John Ashburner and Parashkev Nachev , journal =. Equitable modelling of brain imaging by counterfactual augmentation with morphologically constrained 3D deep generative models , volume =

  25. [25]

    Yaron Lipman and Ricky T. Q. Chen and Heli Ben. Flow Matching for Generative Modeling , booktitle =. 2023 , url =

    2023

  26. [26]

    Chambon and Christian Bluethgen and Jip Prince and Curtis P

    Maya Varma and Ashwin Kumar and Rogier van der Sluijs and Sophie Ostmeier and Louis Blankemeier and Pierre J. Chambon and Christian Bluethgen and Jip Prince and Curtis P. Langlotz and Akshay Chaudhari , title =. CoRR , volume =. 2025 , url =. doi:10.48550/ARXIV.2502.14753 , eprinttype =. 2502.14753 , timestamp =

    Pith/arXiv arXiv doi:10.48550/arxiv.2502.14753 2025

  27. [27]

    Wavelet-Fusion Diffusion Model for Multimodal Brain MRI Synthesis with Modality and Metadata Conditioning

    Muhammad Nabi Yasinzai and Remika Mito and Mangor Pedersen , title =. CoRR , volume =. 2026 , url =. doi:10.48550/ARXIV.2606.00689 , eprinttype =. 2606.00689 , timestamp =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2606.00689 2026

  28. [28]

    and Marquand, Andre F

    Verdi, Serena and Rutherford, Saige and Fraza, Charlotte and Tosun, Duygu and Altmann, Andre and Raket, Lars Lau and Schott, Jonathan M. and Marquand, Andre F. and Cole, James H. and for the Alzheimer's Disease Neuroimaging Initiative , title =. Alzheimer's & Dementia , volume =. doi:https://doi.org/10.1002/alz.14174 , url =. https://alz-journals.onlineli...

    work page doi:10.1002/alz.14174

  29. [29]

    and Wang, Jiancong and Wisse, Laura E.M

    Dong, Mengjin and Xie, Long and Das, Sandhitsu R. and Wang, Jiancong and Wisse, Laura E.M. and deFlores, Robin and Wolk, David A. and Yushkevich, Paul A. , journal =. Regional deep atrophy: Using temporal information to automatically identify regions associated with Alzheimer's disease progression from longitudinal MRI , volume =

  30. [30]

    Greve and Oula Puonti and Axel Thielscher and Koen

    Benjamin Billot and Douglas N. Greve and Oula Puonti and Axel Thielscher and Koen. SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining , volume =. Medical Image Analysis , pages =

  31. [31]

    Longitudinal structural MRI-based deep learning and radiomics features for predicting Alzheimer's disease progression , volume =

    Aghajanian, Sepehr and Mohammadifard, Fateme and Mohammadi, Ida and Rajai Firouzabadi, Shahryar and Baradaran Bagheri, Ali and Moases Ghaffary, Elham and Mirmosayyeb, Omid , journal =. Longitudinal structural MRI-based deep learning and radiomics features for predicting Alzheimer's disease progression , volume =

  32. [32]

    A longitudinal analysis of T1-weighted MRI features associated with progression from mild cognitive impairment to Alzheimer's disease , volume =

    Wang, Yanxia and Song, Wangchen and Yang, Xinyu and Meng, Weijing and Ma, Yonghua and Wang, Aimin and Guo, Guiya and Zhang, Zhaoxue and Li, Zihui and Han, Hairui and Wang, Suzhen and Shi, Fuyan , journal =. A longitudinal analysis of T1-weighted MRI features associated with progression from mild cognitive impairment to Alzheimer's disease , volume =