Pith. sign in

REVIEW 1 major objections 2 minor 35 references

StrokeTimer estimates ischemic stroke onset time from non-contrast CT using self-supervised disentanglement and energy-guided contrastive learning.

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 · grok-4.3

2026-06-28 07:06 UTC pith:WI2NECJM

load-bearing objection StrokeTimer combines self-supervised disentanglement and energy-guided contrastive learning for three-class NCCT onset estimation and reports solid gains on pooled multi-center data, but the robustness claim rests on unstratified evaluation. the 1 major comments →

arxiv 2606.04722 v1 pith:WI2NECJM submitted 2026-06-03 cs.CV

StrokeTimer: Robust Representation Learning for Ischemic Stroke Onset-Time Estimation from Non-contrast CT

Weiru Wang , Susanne G.H. Olthuis , Elizaveta Lavrova , Robert J. van Oostenbrugge , Charles B.L.M. Majoie , Wim H. van Zwam , Ruisheng Su This is my paper
classification cs.CV
keywords ischemic strokeonset time estimationnon-contrast CTself-supervised learningcontrastive learningdisentanglementmedical imagingstroke classification
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 aims to establish that subtle early ischemic changes on non-contrast CT can be extracted to categorize onset time into three windows even when data is imbalanced and comes from multiple centers with different scanners. StrokeTimer combines self-supervised disentanglement learning with energy-guided contrastive learning to build representations that handle these issues. This matters because treatment eligibility for reperfusion depends on knowing how much time has passed since onset, yet that time is frequently unknown in practice. On a large multi-center dataset the method reaches macro AUC of 0.69 and F1 of 0.57, well above near-chance baselines. Model explanations point to known signs such as gray-white matter blurring and hypodense areas.

Core claim

StrokeTimer integrates self-supervised disentanglement learning with energy-guided contrastive learning to capture subtle ischemic patterns while addressing long-tailed data distributions under acquisition variability. On a large multi-center NCCT dataset from MR CLEAN Registry and MR CLEAN LATE it achieves a macro AUC of 0.69 and a macro F1-score of 0.57 for classifying onset into <4.5 h, 4.5-6 h, and >6 h, improving the strongest baseline by nearly 50 percent.

What carries the argument

Self-supervised disentanglement learning combined with energy-guided contrastive learning, which isolates ischemic features from acquisition variability and class imbalance.

Load-bearing premise

Subtle early ischemic changes on NCCT are consistent enough across centers and scanners to be learned despite pronounced class imbalance and heterogeneity.

What would settle it

An independent multi-center NCCT test set on which the model drops to near-chance macro AUC of approximately 0.33 would falsify the claim that the learned representations are robust.

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

If this is right

  • Onset time can be placed into three clinically actionable windows from routine NCCT scans with usable accuracy.
  • Interpretability maps align with established radiological biomarkers such as gray-white matter blurring.
  • The framework maintains performance across scanner and center differences where prior methods fail.
  • Automatic estimation becomes feasible in settings where patient history is unreliable.

Where Pith is reading between the lines

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

  • The same disentanglement strategy could be tested on other subtle time-dependent imaging features such as tumor growth or cardiac ischemia.
  • If the consistency assumption fails on new scanners, adding explicit domain-adaptation terms might be required.
  • Replacing the three discrete windows with a regression head on the same backbone would test whether continuous onset prediction is also feasible.

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

1 major / 2 minor

Summary. The paper introduces StrokeTimer, a framework that combines self-supervised disentanglement learning with energy-guided contrastive learning to estimate ischemic stroke onset time from non-contrast CT (NCCT) scans. Onset times are binned into three clinically relevant intervals (<4.5 h, 4.5–6 h, >6 h). On a pooled multi-center dataset from the MR CLEAN Registry and MR CLEAN LATE cohorts, the method reports a macro AUC of 0.69 and macro F1-score of 0.57, representing an approximately 50% improvement over the strongest baseline (p < 0.005). Model explanations are said to highlight radiological biomarkers such as gray-white matter blurring. Code is released at https://github.com/BrainVas/StrokeTimer.

Significance. If the reported gains are shown to arise from the proposed disentanglement and contrastive components rather than center-specific acquisition artifacts, the work would be significant for supporting time-critical reperfusion decisions in acute stroke, where NCCT is the most widely available modality and onset-time uncertainty is common. The explicit release of code strengthens reproducibility and allows independent verification of the empirical claims.

major comments (1)
  1. [Experimental evaluation] Experimental evaluation (likely §4): The manuscript evaluates on pooled data from MR CLEAN Registry and MR CLEAN LATE without describing center-stratified splits, leave-one-registry-out validation, or scanner-specific hold-out sets. Because the central claim is robustness to “center-scanner-related heterogeneity,” the absence of such partitioning leaves open the possibility that performance exploits cohort-specific acquisition signatures rather than the subtle ischemic changes asserted in the abstract. A concrete test (e.g., per-center AUC tables or cross-registry evaluation) is required to substantiate the robustness claim.
minor comments (2)
  1. [Methods] Abstract and §3: The loss formulations for the energy-guided contrastive term and the disentanglement objective are referenced but not written out with explicit equations; adding the mathematical definitions would improve clarity without altering the central narrative.
  2. [Results] Table 1 or results section: Baseline descriptions should include the exact architectures and training protocols used for the “strongest baseline” so that the 50% relative improvement can be directly reproduced.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback on the experimental evaluation. The concern regarding potential exploitation of center-specific acquisition signatures is well-taken and directly relevant to our robustness claims. We address this point below and will incorporate the requested analyses in the revision.

read point-by-point responses

  1. Referee: Experimental evaluation (likely §4): The manuscript evaluates on pooled data from MR CLEAN Registry and MR CLEAN LATE without describing center-stratified splits, leave-one-registry-out validation, or scanner-specific hold-out sets. Because the central claim is robustness to “center-scanner-related heterogeneity,” the absence of such partitioning leaves open the possibility that performance exploits cohort-specific acquisition signatures rather than the subtle ischemic changes asserted in the abstract. A concrete test (e.g., per-center AUC tables or cross-registry evaluation) is required to substantiate the robustness claim.

    Authors: We agree that the current pooled evaluation does not sufficiently isolate the contribution of the proposed methods from potential center-specific effects. MR CLEAN Registry and MR CLEAN LATE represent distinct national cohorts with differing inclusion criteria and acquisition protocols, yet we did not report explicit cross-registry or center-stratified results. In the revised manuscript we will add leave-one-registry-out experiments (training on one cohort and testing on the other) together with per-center AUC and macro-F1 tables. These additions will allow direct assessment of whether performance gains persist across acquisition heterogeneity. revision: yes

Circularity Check

0 steps flagged

No circularity in empirical performance claims

full rationale

The paper reports standard empirical metrics (macro AUC 0.69, macro F1 0.57) on held-out multi-center NCCT data from MR CLEAN cohorts. No equations, loss functions, or self-citations are shown that reduce these metrics by construction to fitted parameters, self-defined quantities, or prior author results. The framework (self-supervised disentanglement + energy-guided contrastive learning) is presented as a method whose outputs are evaluated externally; the central claims rest on experimental results rather than any derivation chain that collapses to its inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; the method implicitly relies on standard deep-learning assumptions about representation learning and contrastive objectives.

pith-pipeline@v0.9.1-grok · 5856 in / 1160 out tokens · 21390 ms · 2026-06-28T07:06:41.743424+00:00 · methodology

0 comments
read the original abstract

Ischemic stroke is a major global disease. Treatment decisions are highly time-sensitive, as eligibility for reperfusion therapies relies on the interval between stroke onset and intervention. However, the true onset time is often uncertain in clinical practice, necessitating imaging-based assessment of tissue age as a surrogate marker. Early ischemic changes on routinely acquired non-contrast CT (NCCT) are often subtle, and real-world clinical datasets exhibit pronounced onset-time class imbalance and center-scanner-related heterogeneity. In this work, we propose StrokeTimer, a fully automated framework for onset-time estimation in acute ischemic stroke. StrokeTimer integrates self-supervised disentanglement learning with energy-guided contrastive learning to capture subtle ischemic patterns while addressing long-tailed data distributions under acquisition variability. Onset time is categorized into three clinically relevant windows: <4.5 h, 4.5-6 h, and >6 h. Experimental results on a large multi-center NCCT dataset from two national cohorts, MR CLEAN Registry and MR CLEAN LATE, show that StrokeTimer achieves a macro AUC of 0.69 and a macro F1-score of 0.57, improving the strongest baseline by nearly 50% (p < 0.005). In this realistic, challenging setting, representative baseline approaches exhibit near-chance macro performance. Model explanations further highlight subtle gray-white matter blurring and hypodense regions consistent with established radiological biomarkers. These findings demonstrate the potential of StrokeTimer to support treatment decision-making in acute ischemic stroke. Code is available at https://github.com/BrainVas/StrokeTimer.

Figures

Figures reproduced from arXiv: 2606.04722 by Charles B.L.M. Majoie, Elizaveta Lavrova, Robert J. van Oostenbrugge, Ruisheng Su, Susanne G.H. Olthuis, Weiru Wang, Wim H. van Zwam.

Figure 1
Figure 1. Figure 1: Overview of StrokeTimer: a reconstruction-based self-supervised disen [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Grad-CAM++ heatmap comparison between StrokeTimer and top 4 rep [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison between the disentanglement-only and full StrokeTimer (Dis [PITH_FULL_IMAGE:figures/full_fig_p008_3.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

35 extracted references

  1. [1]

    Neurocomputing392, 189–195 (2020)

    Akkus, Z., Kostandy, P., Philbrick, K.A., Erickson, B.J.: Robust brain extraction tool for ct head images. Neurocomputing392, 189–195 (2020)

    2020

  2. [2]

    New England Journal of Medicine378(8), 708–718 (2018)

    Albers,G.W.,Marks,M.P.,Kemp,S.,Christensen,S.,Tsai,J.P.,Ortega-Gutierrez, S., McTaggart, R.A., Torbey, M.T., Kim-Tenser, M., Leslie-Mazwi, T., et al.: Thrombectomy for stroke at 6 to 16 hours with selection by perfusion imaging. New England Journal of Medicine378(8), 708–718 (2018)

    2018

  3. [3]

    In: 2018 IEEE winter conference on applications of computer vision (WACV)

    Chattopadhyay, A., Sarkar, A., Howlader, P., Balasubramanian, V.N.: Grad- cam++: Generalized gradient-based visual explanations for deep convolutional networks. In: 2018 IEEE winter conference on applications of computer vision (WACV). pp. 839–847. IEEE (2018)

    2018

  4. [4]

    In: International conference on machine learning

    Chen, T., Kornblith, S., Norouzi, M., Hinton, G.: A simple framework for con- trastive learning of visual representations. In: International conference on machine learning. pp. 1597–1607. PMLR (2020)

    2020

  5. [5]

    The Lancet

    Collaborators, G..S., et al.: Global, regional, and national burden of stroke and its risk factors, 1990–2019: a systematic analysis for the global burden of disease study 2019. The Lancet. Neurology20(10), 795 (2021)

    1990

  6. [6]

    Foundations and Trends®in Machine Learning12(4), 307–392 (2019)

    Diederik, P.K., Max, W.: An introduction to variational autoencoders. Foundations and Trends®in Machine Learning12(4), 307–392 (2019)

    2019

  7. [7]

    Feigin, V.L., Brainin, M., Norrving, B., Martins, S.O., Pandian, J., Lindsay, P., F Grupper, M., Rautalin, I.: World stroke organization: global stroke fact sheet

  8. [8]

    International Journal of Stroke20(2), 132–144 (2025)

    2025

  9. [9]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    He, K., Chen, X., Xie, S., Li, Y., Dollár, P., Girshick, R.: Masked autoencoders are scalable vision learners. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 16000–16009 (2022) 10 Wang et al

    2022

  10. [10]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    He, K., Fan, H., Wu, Y., Xie, S., Girshick, R.: Momentum contrast for unsupervised visual representation learning. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 9729–9738 (2020)

    2020

  11. [11]

    BMJ360(2018)

    Jansen, I.G., Mulder, M.J., Goldhoorn, R.J.B.: Endovascular treatment for acute ischaemicstrokeinroutineclinicalpractice:prospective,observationalcohortstudy (mr clean registry). BMJ360(2018)

    2018

  12. [12]

    European Radiology34(10), 6808–6819 (2024)

    Jiang, L., Sun, J., Wang, Y., Yang, H., Chen, Y.C., Peng, M., Zhang, H., Chen, Y., Yin, X.: Diffusion-/perfusion-weighted imaging fusion to automatically identify stroke within 4.5 h. European Radiology34(10), 6808–6819 (2024)

    2024

  13. [13]

    Nature biomedical engineering5(6), 571–585 (2021)

    Kanakasabapathy, M.K., Thirumalaraju, P., Kandula, H., Doshi, F., Sivakumar, A.D., Kartik, D., Gupta, R., Pooniwala, R., Branda, J.A., Tsibris, A.M., et al.: Adaptive adversarial neural networks for the analysis of lossy and domain-shifted datasets of medical images. Nature biomedical engineering5(6), 571–585 (2021)

    2021

  14. [14]

    In: International Conference on Learning Representations (2020)

    Kang, B., Xie, S., Rohrbach, M., Yan, Z., Gordo, A., Feng, J., Kalantidis, Y.: De- coupling representation and classifier for long-tailed recognition. In: International Conference on Learning Representations (2020)

    2020

  15. [15]

    Stroke33(7), 1786–1791 (2002)

    Kucinski, T., Väterlein, O., Glauche, V., Fiehler, J., Klotz, E., Eckert, B., Koch, C., Roöther, J., Zeumer, H.: Correlation of apparent diffusion coefficient and computed tomography density in acute ischemic stroke. Stroke33(7), 1786–1791 (2002)

    2002

  16. [16]

    In: International Conference on Learning Representations (2021)

    Li, J., Zhou, P., Xiong, C., Hoi, S.: Prototypical contrastive learning of unsuper- vised representations. In: International Conference on Learning Representations (2021)

    2021

  17. [17]

    IEEE transactions on pattern analysis and machine intelligence42(2), 318–327 (2020)

    Lin, T.Y., Goyal, P., Girshick, R., He, K., Dollar, P.: Focal loss for dense object detection. IEEE transactions on pattern analysis and machine intelligence42(2), 318–327 (2020)

    2020

  18. [18]

    npj Digital Medicine7(1), 338 (2024)

    Marcus,A.,Mair,G.,Chen,L.,Hallett,C.,Cuervas-Mons,C.G.,Roi,D.,Rueckert, D., Bentley, P.: Deep learning biomarker of chronometric and biological ischemic stroke lesion age from unenhanced ct. npj Digital Medicine7(1), 338 (2024)

    2024

  19. [19]

    In: Proceedings of the AAAI conference on artificial intelligence

    Perez, E., Strub, F., De Vries, H., Dumoulin, V., Courville, A.: Film: Visual rea- soning with a general conditioning layer. In: Proceedings of the AAAI conference on artificial intelligence. vol. 32 (2018)

    2018

  20. [20]

    European Radiol- ogy pp

    van Poppel, L.M., de Vries, L., Mojtahedi, M., Kappelhof, M., Olthuis, S.G., van Oostenbrugge, R., van Zwam, W.H., Jan van Doormaal, P., Beenen, L.F., Roos, Y.B., et al.: Machine learning models for ct-based classification of ischemic stroke onset time within or beyond 4.5 h: a comparison of approaches. European Radiol- ogy pp. 1–11 (2025)

    2025

  21. [21]

    Radiology229(2), 347–359 (2003)

    Provenzale, J.M., Jahan, R., Naidich, T.P., Fox, A.J.: Assessment of the patient with hyperacute stroke: imaging and therapy. Radiology229(2), 347–359 (2003)

    2003

  22. [22]

    Advances in neural information processing systems33, 4175–4186 (2020)

    Ren, J., Yu, C., Ma, X., Zhao, H., Yi, S., et al.: Balanced meta-softmax for long- tailed visual recognition. Advances in neural information processing systems33, 4175–4186 (2020)

    2020

  23. [23]

    In: European Conference on Computer Vision

    Sarhan, M.H., Navab, N., Eslami, A., Albarqouni, S.: Fairness by learning orthog- onal disentangled representations. In: European Conference on Computer Vision. pp. 746–761. Springer (2020)

    2020

  24. [24]

    Ad- vances in neural information processing systems30(2017)

    Snell, J., Swersky, K., Zemel, R.: Prototypical networks for few-shot learning. Ad- vances in neural information processing systems30(2017)

    2017

  25. [25]

    Medical Image Analysis97, 103250 (2024) StrokeTimer: Stroke Onset Time Estimation from NCCT 11

    Sun, J., Liu, Y., Xi, Y., Coatrieux, G., Coatrieux, J.L., Ji, X., Jiang, L., Chen, Y.: Multi-grained contrastive representation learning for label-efficient lesion seg- mentation and onset time classification of acute ischemic stroke. Medical Image Analysis97, 103250 (2024) StrokeTimer: Stroke Onset Time Estimation from NCCT 11

    2024

  26. [26]

    Stroke46(9), 2707–2713 (2015)

    Thomalla, G., Gerloff, C.: Treatment concepts for wake-up stroke and stroke with unknown time of symptom onset. Stroke46(9), 2707–2713 (2015)

    2015

  27. [27]

    In: StrokeWorkshoponImagingandTreatmentChallenges.pp.52–61.Springer(2025)

    Vorberg, L., Rist, L., Ditt, H., Maier, A., Taubmann, O.: Leveraging last-known- well times for radiomics-based stroke onset estimation from non-contrast ct. In: StrokeWorkshoponImagingandTreatmentChallenges.pp.52–61.Springer(2025)

    2025

  28. [28]

    Research8, 1029 (2025)

    Wang,S.,Zhou,X.,Li,C.,Wang,S.,Li,Y.,Tan,T.,Zheng,H.:Generativeartificial intelligence in medical imaging: Foundations, progress, and clinical translation. Research8, 1029 (2025)

    2025

  29. [29]

    IEEE Transactions on Geoscience and Remote Sensing62, 1–16 (2024)

    Wang, Y., Albrecht, C.M., Zhu, X.X.: Multilabel-guided soft contrastive learning for efficient earth observation pretraining. IEEE Transactions on Geoscience and Remote Sensing62, 1–16 (2024)

    2024

  30. [30]

    In: Asian conference on machine learning

    Wang, Y., Zhang, B., Hou, W., Wu, Z., Wang, J., Shinozaki, T.: Margin calibration for long-tailed visual recognition. In: Asian conference on machine learning. pp. 1101–1116. PMLR (2023)

    2023

  31. [31]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Wu,L.,Zhuang,J.,Chen,H.:Voco:Asimple-yet-effectivevolumecontrastivelearn- ing framework for 3d medical image analysis. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 22873–22882 (2024)

    2024

  32. [32]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Yan, W., Wang, Y., Gu, S., Huang, L., Yan, F., Xia, L., Tao, Q.: The domain shift problem of medical image segmentation and vendor-adaptation by unet-gan. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 623–631. Springer (2019)

    2019

  33. [33]

    In: Proceedings of the AAAI conference on artificial intelligence

    Zhang, P., Wu, M.: Multi-label supervised contrastive learning. In: Proceedings of the AAAI conference on artificial intelligence. vol. 38, pp. 16786–16793 (2024)

    2024

  34. [34]

    Advances in neural information processing systems31(2018)

    Zhao,H.,Zhang,S.,Wu,G.,Moura,J.M.,Costeira,J.P.,Gordon,G.J.:Adversarial multiple source domain adaptation. Advances in neural information processing systems31(2018)

    2018

  35. [35]

    NeuroImage243, 118569 (2021)

    Zuo, L., Dewey, B.E., Liu, Y., He, Y., Newsome, S.D., Mowry, E.M., Resnick, S.M., Prince, J.L., Carass, A.: Unsupervised mr harmonization by learning dis- entangled representations using information bottleneck theory. NeuroImage243, 118569 (2021)

    2021