Pith. sign in

REVIEW 4 major objections 4 minor 65 references

Advancing Stroke Risk Prediction Using a Multi-modal Foundation Model

T0 review · 4 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims that jointly aligning brain MRI and clinical tabular encoders with a contrastive loss and an image-tabular matching loss predicts pre-stroke risk better than any uni-modal or supervised multimodal baseline on the UK…

desk verdict New application, old building blocks: the multimodal setup and cohort work are real, but the statistical basis for the performance claims doesn't hold. read the letter →

arxiv 2411.09822 v2 pith:XC6OMK54 submitted 2024-11-14 cs.CV cs.AI

Camille Delgrange , Olga Demler , Samia Mora , Bjoern Menze , Ezequiel de la Rosa , Neda Davoudi This is my paper · ORCID
classification cs.CVcs.AI
keywords strokeriskpredictionself-supervisedlearningcontrastivemultimodalfoundationmodelbrainMRIT2-FLAIRclinicaltabulardataUKBiobank
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper's aim is to show that stroke risk can be predicted before onset by jointly learning from a patient's pre-stroke brain MRI and clinical records, without requiring annotated training data. It proposes a self-supervised framework that aligns image and tabular embeddings in a shared latent space using a contrastive objective plus an image-tabular matching task. On the UK Biobank test set, the combined model reaches a ROC-AUC of 74.42 and a balanced accuracy of 71.11, ahead of all uni-modal baselines and of the best supervised multimodal baseline (DAFT at 73.82). The authors present this as the first pre-onset stroke risk model built from multimodal clinical data, with heatmaps showing that the model focuses on white-matter hyperintensity regions tied to aging and stroke.

What carries the argument

The load-bearing mechanism is the joint alignment of two encoders in a shared latent space: a CLIP-style contrastive loss defined between image and tabular projections, combined with an image-tabular matching (ITM) loss over hard negative pairs and a cross-attention transformer module that produces a joint [CLS] representation. The CLIP loss maximizes cosine similarity between matching image-tabular pairs while minimizing similarity to in-batch negatives, and the ITM loss adds a binary match/no-match prediction on mined hard negatives; together they force the two modalities to share a common patient representation. The shared space is what allows the downstream classifier to draw on both modalities, and the paper's UMAP analysis shows that uni-modal pretraining leaves image and tabular embeddings separated while the multimodal pretraining overlaps them.

What would settle it

Retrain the model with healthy controls sampled from stroke-free UK Biobank participants who did not return for a second imaging visit, matched for age and sex; if the ROC-AUC advantage over DAFT disappears or falls below significance, the central claim is confounded.

Watch

Extended reading notes

Core claim

The central discovery is that aligning 3D T2-FLAIR brain MRIs with structured clinical data in a shared embedding space produces representations that transfer better to pre-stroke risk classification than either modality alone or than supervised fusion. The method pre-trains a ResNet-50 image encoder and an MLP tabular encoder with a CLIP-style contrastive loss that pulls matching image-tabular pairs together and pushes mismatched pairs apart, together with an image-tabular matching (ITM) loss that uses hard negative pairs and a cross-attention transformer to force genuine cross-modal interaction. After fine-tuning on 278 labeled examples, the model reaches ROC-AUC 74.42 and balanced accuracy 71.11 on a 93-patient test set, surpassing the best self-supervised tabular baseline by 2.1% in ROC-AUC and 10.6% in balanced accuracy, the best self-supervised image baseline by 2.8% and 12.8%, and the best supervised multimodal baseline (DAFT) by 7.6% in balanced accuracy. The authors report that the improvement over all self-supervised baselines is statistically significant when evaluated at the real 0.52% stroke prevalence with bootstrapping, and that removing the ITM loss lowers performance.

Load-bearing premise

The load-bearing premise is the definition of healthy controls as participants who were still stroke-free at a second imaging visit; if returning for follow-up is itself a marker of different health or health-seeking behavior, the reported gains could partly reflect that difference rather than stroke risk.

Editorial extensions

If this is right

  • If correct, pre-stroke imaging and routine clinical data contain complementary signals that a contrastively aligned model can exploit without expert annotations.
  • Self-supervised pretraining on unlabeled biobank-scale data can substitute for scarce pre-onset stroke labels; frozen representations nearly match trainable ones on the small fine-tuning set.
  • The ITM loss contributes measurable gains over pure CLIP alignment, implying that hard-negative cross-modal matching adds signal beyond simple embedding alignment.
  • The model's GradCAM activations concentrate in periventricular and deep white-matter hyperintensities, a pattern consistent with known vascular aging and stroke pathology, and could guide imaging biomarker discovery.
  • Evaluated at the real 0.52% stroke prevalence, the model's balance-accuracy lead over self-supervised baselines persists under bootstrap reweighting, supporting its potential as a screening aid rather than a diagnostic.

Reading between the lines

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

  • If the healthy-control definition (stroke-free at a second imaging visit) is replaced by stroke-free participants who did not return for follow-up, the model may lose part of its advantage, because the current controls could let the model learn follow-up attendance or health-seeking behavior rather than pure stroke risk.
  • The CLIP-plus-ITM recipe is modality-agnostic; the same shared-latent alignment with hard-negative matching could be transferred to other large biobanks that pair imaging with structured health records for different disease outcomes.
  • The finding that accuracy peaks for strokes occurring 2-3 years after the scan suggests a possible imaging-detectable prodromal window; a testable extension would stratify by time-to-onset in a larger cohort and check whether the signal remains stable.
  • The 93-patient test set is small, so a practical follow-up beyond ranking metrics would be to examine calibration and threshold choice at realistic prevalence before any clinical screening use.
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 / 4 minor

Summary. The paper proposes a self-supervised multimodal framework that aligns 3D T2-FLAIR brain MRI and tabular clinical/imaging-derived features in a shared latent space using a CLIP-style contrastive loss together with an image-tabular matching (ITM) loss. The method is pre-trained on 5,000 UK Biobank samples and fine-tuned on 278 samples, then evaluated on a 93-patient hold-out set for pre-stroke risk prediction. The authors compare against supervised and self-supervised unimodal and multimodal baselines, reporting higher ROC-AUC and balanced accuracy for the proposed model, and support the result with interpretability analyses (UMAP embeddings, integrated gradients, and GradCAM heatmaps). The paper also includes a bootstrap-based significance analysis in Table 5 and a discussion of clinical plausibility.

Significance. If the reported gains are statistically reliable, the paper would provide a useful demonstration that self-supervised multimodal pretraining can combine imaging and tabular data for a low-prevalence clinical outcome such as pre-stroke risk prediction. The work has practical strengths: a released code repository, a reasonably comprehensive set of baselines, and interpretability analyses that connect model attention to white-matter hyperintensities, which are clinically relevant. However, the central comparative claim is not yet convincingly supported: the primary benchmarking table lacks confidence intervals, the bootstrap significance analysis is internally inconsistent with its own stated protocol, and the reported effect sizes in the abstract and conclusion do not match Table 4. The contribution is promising but needs substantial statistical and reporting revisions before the headline claims can be accepted.

major comments (4)
  1. [Section 4.1, Table 4] The key comparison of the proposed model against baselines is presented as point estimates on a single 93-patient test set. No confidence intervals, standard errors, or repeated-seed results are reported. Given the small test set and the number of models compared, the claimed improvements (e.g., ROC-AUC 74.42 vs 73.82 for DAFT; balanced accuracy 71.11 vs 67.85 for SCARF T T) are not shown to be statistically reliable. The authors should report confidence intervals or resampling-based intervals for all models in Table 4, and should specify how many random seeds were used and how the Youden-index operating point selected on the validation set affects the reported test metrics.
  2. [Section 2.2.2, Table 5] The bootstrap procedure is underdescribed and internally inconsistent. With n=93 and the stated 0.52% stroke prevalence, each bootstrap resample would contain about 0.48 stroke patients, so sensitivity and F1 would be undefined for most resamples under binomial sampling; yet Table 5 reports tight intervals such as sensitivity 0.846 [0.770–0.910] and F1 0.744 [0.723–0.765]. The authors must specify the exact resampling protocol: whether whole patients are resampled, what prevalence is used, how undefined metrics are handled, and how paired predictions are constructed for the Wilcoxon signed-rank test. As written, Table 5 cannot serve as evidence of significance, and the overlapping AUC confidence intervals (Ours 0.748 [0.696–0.798]; CLIP 0.734 [0.680–0.787]; SCARF 0.727 [0.672–0.780]) further undermine the claimed p<0.01 for the primary metric.
  3. [Abstract, Section 5, Table 4] The reported improvement magnitudes are inconsistent across the manuscript. The abstract states gains of 2.6% (2.6%) in ROC-AUC and 3.3% (5.6%) in balanced accuracy over self-supervised tabular (image) methods, while Section 5 states 2.1% (2.8%) in ROC-AUC and 10.6% (12.8%) in balanced accuracy, and 7.6% accuracy over the best multimodal supervised model. Neither set of numbers is fully reproducible from Table 4: for example, Ours T T has AUC 74.42 versus best SCARF 72.16 (difference 2.26) and best SimCLR 72.11 (difference 2.31), while balanced accuracy differences versus SCARF T T and SimCLR T T are 3.26 and 5.55, respectively. The authors should recompute and harmonize all reported effect sizes under a clearly stated comparison convention.
  4. [Section 2.1.1, Table 1] The healthy control definition—participants who were still stroke-free at a second imaging visit—may introduce selection bias: controls who return for follow-up may differ from the general population in health-seeking behavior and unmeasured risk, and the model may partly learn to distinguish 'returned for follow-up' rather than stroke risk. This concern should be analyzed, for example by comparing the control group's risk-factor distribution with the broader UKB non-stroke population or by using all stroke-free participants as controls, or at least discussed explicitly as a limitation affecting generalization.
minor comments (4)
  1. [Section 3.3, Table 4] The column header 'Acc' is used interchangeably with 'balanced accuracy' in the text; please use a single term consistently throughout the manuscript.
  2. [Section 4.2.2, Figure 6] The Mann-Whitney U test on GradCAM activations is a single unadjusted comparison on 93 test scans; please report the effect size and clarify that this analysis is exploratory.
  3. [References] Several references have formatting errors or incomplete bibliographic details (e.g., [63] contains a duplicate DOI/URL pattern and placeholder text); please check the reference list against the publisher's style.
  4. [Table 5] The confidence intervals in Table 5 use inconsistent decimal formatting (F1 intervals have four digits while other metrics have three); please unify the formatting.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's claims are empirical benchmarking results on a held-out split, not derivations that assume their own conclusions.

full rationale

The paper's central claim is that a multi-modal self-supervised framework (CLIP loss plus ITM loss) outperforms unimodal and supervised multimodal baselines on stroke risk prediction. This is established through a standard empirical pipeline: pre-training on 5000 UK Biobank samples, fine-tuning on 278 labeled samples, and evaluating on a held-out test set of 93 samples that were not used during pre-training or fine-tuning. The reported metrics (ROC-AUC, balanced accuracy, F1, sensitivity) are computed from model predictions on that held-out split, and no metric is obtained by construction from a fitted parameter, a definitional identity, or a self-citation. The paper does not invoke a uniqueness theorem, does not import a modeling ansatz exclusively from the authors' own prior work, and does not rename a known empirical pattern as a new organizational principle. The cited Transformer and CLIP/ITM components are attributed to external prior work and are used as architectural building blocks, not as evidence that the target result is forced. The acknowledged limitations (UK Biobank demographics, small test set, pre-stroke imaging availability, and variability with onset time) are stated explicitly in the Limitations section and concern generalizability and statistical reliability rather than circular reasoning. The most serious concern raised by the skeptic is the internal consistency of Table 5's bootstrap at 0.52% prevalence with n=93, but that is a statistical soundness issue, not a circularity issue: even if the bootstrap protocol is flawed, the headline comparison is not reduced to its own inputs by definition. No circular step meeting the required evidentiary standard (a quoted reduction of the claimed result to its inputs) was found.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim depends on a set of standard ML hyperparameters (temperature, loss weights, augmentation rates, batch size, learning rate) and on domain assumptions about label accuracy, control group selection, transferability of self-supervised features, and the predictive value of FLAIR white matter hyperintensities. No new physical entities or external priors are introduced.

free parameters (6)
  • CLIP temperature tau = 0.1
    Chosen following SimCLR [22]; not tuned to the dataset. Affects the sharpness of the contrastive distribution.
  • loss weight lambda = 0.5
    Balances image-to-tabular and tabular-to-image CLIP losses; set by hand without reported ablation.
  • tabular corruption rate = 0.3
    Taken from SCARF [26], the baseline method; used for tabular augmentations during pre-training.
  • image augmentation rate = 0.95
    Chosen to let the model occasionally see unaltered images; stated in Section 2.2.
  • batch size = 6
    Small batch size used for all models; affects the number of contrastive negatives and was not systematically varied.
  • learning rate
    Swept through a range per method; exact values are only reported in supplement Tables S4/S5, not in the main text.
assumptions (4)
  • domain assumption ICD9/10 codes correctly identify incident stroke events in UK Biobank
    Used to label stroke cases in Section 2.1.1; miscoding would mislabel outcomes and affect all performance estimates.
  • domain assumption Healthy controls defined by the second imaging visit constitute an unbiased comparison group
    Controls are stroke-free at the second scan; this may select a healthier subgroup, as discussed in Section 2.1.1.
  • domain assumption Self-supervised contrastive pre-training on 5,000 unannotated subjects transfers to the downstream fine-tuning labels
    The premise that CLIP/ITM pretraining learns useful representations for the stroke classification task; tested only indirectly via downstream metrics.
  • domain assumption T2-FLAIR white matter hyperintensities are valid precursors of future stroke
    Motivates the imaging modality selection based on literature [33-36]; if FLAIR is not informative for early stroke, the imaging branch adds little.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Advancing Stroke Risk Prediction Using a Multi-modal Foundation Model." pith.science (2026). https://pith.science/paper/XC6OMK54

@misc{pith2026241109822,
  author       = {Pith},
  title        = {Pith review of: Advancing Stroke Risk Prediction Using a Multi-modal Foundation Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XC6OMK54}},
  note         = {Machine review of arXiv:2411.09822}
}
read the original abstract

Predicting stroke risk is a complex challenge that can be enhanced by integrating diverse clinically available data modalities. This study introduces a self-supervised multimodal framework that combines 3D brain imaging, clinical data, and image-derived features to improve stroke risk prediction prior to onset. By leveraging large unannotated clinical datasets, the framework captures complementary and synergistic information across image and tabular data modalities. Our approach is based on a contrastive learning framework that couples contrastive language-image pretraining with an image-tabular matching module, to better align multimodal data representations in a shared latent space. The model is trained on the UK Biobank, which includes structural brain MRI and clinical data. We benchmark its performance against state-of-the-art unimodal and multimodal methods using tabular, image, and image-tabular combinations under diverse frozen and trainable model settings. The proposed model outperformed self-supervised tabular (image) methods by 2.6% (2.6%) in ROC-AUC and by 3.3% (5.6%) in balanced accuracy. Additionally, it showed a 7.6% increase in balanced accuracy compared to the best multimodal supervised model. Through interpretable tools, our approach demonstrated better integration of tabular and image data, providing richer and more aligned embeddings. Gradient-weighted Class Activation Mapping heatmaps further revealed activated brain regions commonly associated in the literature with brain aging, stroke risk, and clinical outcomes. This robust self-supervised multimodal framework surpasses state-of-the-art methods for stroke risk prediction and offers a strong foundation for future studies integrating diverse data modalities to advance clinical predictive modelling.

Figures

Figures reproduced from arXiv: 2411.09822 by the authors.

Figure 1
Figure 1. Pipeline for joint imaging and tabular data pre-training (A) and supervised fine tuning (B). [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 3
Figure 3. Distribution of metrics with respect to group partition. [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. UMAP projections of tabular and imaging embeddings from validation set, using (A) [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: Feature importance analysis for model interpretability with anatomical T2-FLAIR scans [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: Boxplot of normalized GradCAM activations of white matter hyperintensities masks over [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

65 extracted references · 38 canonical work pages

  1. [1]

    Projected Global Trends in Ischemic Stroke Incidence, Deaths and Disability- Adjusted Life Years From 2020 to 2030

    Liyuan Pu et al. “Projected Global Trends in Ischemic Stroke Incidence, Deaths and Disability- Adjusted Life Years From 2020 to 2030”. In: Stroke 54.5 (May 2023), pp. 1330–1339. ISSN : 15244628. DOI: 10.1161/STROKEAHA.122.040073

    work page doi:10.1161/strokeaha.122.040073 2020

  2. [2]

    Global, regional, and national burden of stroke and its risk factors, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019

    Valery L. Feigin and et. al. Stark. “Global, regional, and national burden of stroke and its risk factors, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019”. In: The Lancet. Neurology 20.10 (2021), pp. 1–26. ISSN : 1474-4465. DOI: 10.1016/S1474- 4422(21)00252-0

    doi:10.1016/s1474- 2021

  3. [3]

    A Brief Review of Cardiovascular Diseases, Associated Risk Factors and Current Treatment Regimes

    Gagan D. Flora and Manasa K. Nayak. “A Brief Review of Cardiovascular Diseases, Associated Risk Factors and Current Treatment Regimes”. In:Current Pharmaceutical Design 25.38 (Sept. 2019), pp. 4063–4084. ISSN : 13816128. DOI: 10.2174/1381612825666190925163827

    work page doi:10.2174/1381612825666190925163827 2019

  4. [4]

    Probabilistic Temporal Prediction of Continuous Disease Tra- jectories and Treatment Effects Using Neural SDEs

    Joshua Durso-Finley et al. “Probabilistic Temporal Prediction of Continuous Disease Tra- jectories and Treatment Effects Using Neural SDEs”. In: Medical Image Computing and Computer Assisted Intervention - MICCAI 2024 . Ed. by Marius George Linguraru et al. Cham: Springer Nature Switzerland, 2024, pp. 400–410. ISBN : 978-3-031-72384-1. DOI: https://doi.or...

    work page doi:10.1007/978-3-031-72384-1_38 2024

  5. [5]

    Biological effects and safety in magnetic resonance imaging: A review

    Valentina Hartwig et al. “Biological effects and safety in magnetic resonance imaging: A review”. In: International Journal of Environmental Research and Public Health6.6 (2009), pp. 1778–1798. ISSN : 16604601. DOI: 10.3390/ijerph6061778

    work page doi:10.3390/ijerph6061778 2009

  6. [6]

    Applying Deep Learning to Medical Imaging: A Review

    Huanhuan Zhang and Yufei Qie. “Applying Deep Learning to Medical Imaging: A Review”. In: MDPI 13.18 (Sept. 2023), p. 10521. ISSN : 20763417

    work page 2023

  7. [7]

    Use of deep learning to predict final ischemic stroke lesions from initial magnetic resonance imaging

    Yannan Yu et al. “Use of deep learning to predict final ischemic stroke lesions from initial magnetic resonance imaging”. In: JAMA network open 3.3 (2020), e200772–e200772. DOI: 10.1001/jamanetworkopen.2020.0772

    work page arXiv 2020

  8. [8]

    Development and validation of a new algorithm for improved cardiovascular risk prediction

    Julia Hippisley-Cox et al. “Development and validation of a new algorithm for improved cardiovascular risk prediction”. In: Nature Medicine 30.5 (Apr. 2024), pp. 1440–1447. ISSN : 1546170X. DOI: https://doi.org/10.1038/s41591-024-02905-y

    doi:10.1038/s41591-024-02905-y 2024

Show all 65 references
  1. [9]

    Recurrent atherosclerotic cardiovascular event rates differ among patients meeting the very high risk definition according to age, sex, race/ethnicity, and socioeconomic status

    Jaejin An et al. “Recurrent atherosclerotic cardiovascular event rates differ among patients meeting the very high risk definition according to age, sex, race/ethnicity, and socioeconomic status”. In: Journal of the American Heart Association 9.23 (Dec. 2020). ISSN : 20479980....

    2020 doi

  2. [10]

    Development of machine learning-based models to predict 10-year risk of cardiovascular disease: a prospective cohort study

    Jia You et al. “Development of machine learning-based models to predict 10-year risk of cardiovascular disease: a prospective cohort study”. In: Stroke and vascular neurology 8.6 (Dec. 2023), pp. 475–485. ISSN : 2059-8696. DOI: 10.1136/svn-2023-002332

    2023 doi

  3. [11]

    Cardiovascular disease risk prediction using automated machine learning: A prospective study of 423,604 UK Biobank participants

    Ahmed M. Alaa et al. “Cardiovascular disease risk prediction using automated machine learning: A prospective study of 423,604 UK Biobank participants”. In: PLoS ONE 14.5 (May 2019). ISSN : 19326203. DOI: 10.1371/journal.pone.0213653

    2019 doi

  4. [12]

    The UK Biobank imaging enhancement of 100,000 participants: rationale, data collection, management and future directions

    Thomas J. Littlejohns et al. “The UK Biobank imaging enhancement of 100,000 participants: rationale, data collection, management and future directions”. In: Nature Communications 11.1 (May 2020), pp. 1–12. ISSN : 20411723. DOI: https://doi.org/10.1038/s41467- 020-15948-9

    2020 doi

  5. [13]

    Online at https://www.ukbiobank.ac.uk/

    UK Biobank Website. Online at https://www.ukbiobank.ac.uk/ . Website accessed on 2024-03-05. 2021. URL: https://www.ukbiobank.ac.uk/. 19

    2024

  6. [14]

    Multimodal neuroimaging computing: a review of the applications in neuropsychiatric disorders

    Sidong Liu et al. “Multimodal neuroimaging computing: a review of the applications in neuropsychiatric disorders”. In: Springer 2.3 (Sept. 2015), pp. 167–180. ISSN : 21984026. DOI: https://doi.org/10.1007/s40708-015-0019-x

    2015 doi

  7. [15]

    Improvement in the Prediction of Cerebrovascular Events With White Matter Hyperintensity

    Adam de Havenon et al. “Improvement in the Prediction of Cerebrovascular Events With White Matter Hyperintensity”. In: Journal of the American Heart Association 12.13 (2023). DOI: 10.1161/JAHA.123.029374. URL: https://www.ahajournals.org/doi/abs/10. 1161/JAHA.123.029374

    2023 doi

  8. [16]

    Long-term cancer survival prediction using multimodal deep learning

    Luís A. Vale-Silva and Karl Rohr. “Long-term cancer survival prediction using multimodal deep learning”. In: Scientific Reports 11.1 (June 2021), pp. 1–12. ISSN : 20452322. DOI: https://doi.org/10.1038/s41598-021-92799-4

    2021 doi

  9. [17]

    Predicting Stroke through Retinal Graphs and Multimodal Self-supervised Learning

    Yuqing Huang et al. “Predicting Stroke through Retinal Graphs and Multimodal Self-supervised Learning”. In: 2024. arXiv: 2411.05597 [cs.CV]. URL: https://arxiv.org/abs/2411. 05597

    2024 arXiv

  10. [18]

    Predicting recovery following stroke: Deep learning, multimodal data and feature selection using explainable AI

    Adam White et al. “Predicting recovery following stroke: Deep learning, multimodal data and feature selection using explainable AI”. In: NeuroImage: Clinical 43 (2024), p. 103638. ISSN : 2213-1582. DOI: https : / / doi . org / 10 . 1016 / j . nicl . 2024 . 103638. URL: https:/...

    2024

  11. [19]

    Functional Outcome Prediction in Acute Ischemic Stroke Using a Fused Imaging and Clinical Deep Learning Model

    Yongkai Liu et al. “Functional Outcome Prediction in Acute Ischemic Stroke Using a Fused Imaging and Clinical Deep Learning Model”. In: Stroke 54.9 (Sept. 2023), pp. 2316–2327. ISSN : 15244628

    2023

  12. [20]

    DAFT: A universal module to interweave tabular data and 3D images in CNNs

    Tom Nuno Wolf, Sebastian Pölsterl, and Christian Wachinger. “DAFT: A universal module to interweave tabular data and 3D images in CNNs”. In: NeuroImage 260 (Oct. 2022), p. 119505. ISSN : 10959572. DOI: https://doi.org/10.1016/j.neuroimage.2022.119505

    2022

  13. [21]

    A Cookbook of Self-Supervised Learning

    Randall Balestriero et al. “A Cookbook of Self-Supervised Learning”. In: 2023

    2023

  14. [22]

    A simple framework for contrastive learning of visual representations

    Ting Chen et al. “A simple framework for contrastive learning of visual representations”. In: 37th International Conference on Machine Learning, ICML 2020 PartF168147-3.Figure 1 (2020), pp. 1575–1585

    2020

  15. [23]

    Bootstrap your own latent a new approach to self-supervised learning

    Jean Bastien Grill et al. “Bootstrap your own latent a new approach to self-supervised learning”. In: Advances in Neural Information Processing Systems. V ol. 2020-Decem. Neural information processing systems foundation, June 2020. ISBN : 2006.07733v3. DOI: https://doi.org/ 10...

    2020 doi arXiv

  16. [24]

    Momentum Contrast for Unsupervised Visual Representation Learning

    Kaiming He et al. “Momentum Contrast for Unsupervised Visual Representation Learning”. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. IEEE Computer Society, Nov. 2020, pp. 9726–9735.DOI: 10.1109/CVPR42600. 2020.00975

    2020

  17. [25]

    VIME: Variational information maximizing exploration

    Rein Houthooft et al. “VIME: Variational information maximizing exploration”. In: Advances in Neural Information Processing Systems. V ol. 0. Neural information processing systems foundation, May 2016, pp. 1117–1125. DOI: https://doi.org/10.48550/arXiv.1509. 08731

    2016 doi

  18. [26]

    Scarf: Self-Supervised Contrastive Learning Using Random Feature Cor- ruption

    Dara Bahri et al. “Scarf: Self-Supervised Contrastive Learning Using Random Feature Cor- ruption”. In: ICLR 2022 - 10th International Conference on Learning Representations (2022), pp. 1–24. DOI: https://doi.org/10.48550/arXiv.2106.15147

    2022 doi arXiv

  19. [27]

    Cross-modal autoencoder framework learns holistic representations of cardiovascular state

    Adityanarayanan Radhakrishnan et al. “Cross-modal autoencoder framework learns holistic representations of cardiovascular state”. In: Nature Communications 14.1 (Dec. 2023). ISSN : 20411723. DOI: https://doi.org/10.1038/s41467-023-38125-0 . 20

    2023 doi

  20. [28]

    TIP: Tabular-Image Pre-training for Multimodal Classification with Incomplete Data

    Siyi Du et al. “TIP: Tabular-Image Pre-training for Multimodal Classification with Incomplete Data”. In: arxiv (July 2024). DOI: https://doi.org/10.48550/arXiv.2407.07582

    2024 doi arXiv

  21. [29]

    Best of Both Worlds: Multimodal Contrastive Learning with Tabular and Imaging Data

    Paul Hager, Martin J. Menten, and Daniel Rueckert. “Best of Both Worlds: Multimodal Contrastive Learning with Tabular and Imaging Data”. In: arXiv (2023), pp. 23924–23935. DOI: https://doi.org/10.48550/arXiv.2303.14080

    2023 doi arXiv

  22. [30]

    Enhancing Cardiovascular Disease Prediction through Multi-Modal Self-Supervised Learning

    Francesco Girlanda et al. “Enhancing Cardiovascular Disease Prediction through Multi-Modal Self-Supervised Learning”. In: BMVC. 2024

    2024

  23. [31]

    Learning Transferable Visual Models From Natural Language Super- vision

    Alec Radford et al. “Learning Transferable Visual Models From Natural Language Super- vision”. In: Proceedings of Machine Learning Research. V ol. 139. ML Research Press, Feb. 2021, pp. 8748–8763. ISBN : 9781713845065. DOI: https://doi.org/10.48550/arXiv. 2103.00020

    2021 doi arXiv

  24. [32]

    Align before Fuse: Vision and Language Representation Learning with Mo- mentum Distillation

    Junnan Li et al. “Align before Fuse: Vision and Language Representation Learning with Mo- mentum Distillation”. In: Advances in Neural Information Processing Systems. V ol. 12. Neural information processing systems foundation, July 2021, pp. 9694–9705. ISBN : 9781713845393. DO...

    2021 doi arXiv

  25. [33]

    Pathological correlates of white matter hyperintensities on magnetic resonance imaging

    Yong Soo Shim et al. “Pathological correlates of white matter hyperintensities on magnetic resonance imaging”. In: Dementia and Geriatric Cognitive Disorders 39 (Feb. 2015), pp. 92–

    2015

  26. [34]

    The clinical importance of white matter hyperintensities on brain magnetic resonance imaging: Systematic review and meta-analysis

    Stéphanie Debette and H. S. Markus. “The clinical importance of white matter hyperintensities on brain magnetic resonance imaging: Systematic review and meta-analysis”. In: 341.7767 (Aug. 2010), p. 288. ISSN : 17561833

    2010

  27. [35]

    Cerebral white matter hyperintensities predict functional stroke outcome

    Li Min Liou et al. “Cerebral white matter hyperintensities predict functional stroke outcome”. In: Cerebrovascular Diseases 29.1 (Dec. 2009), pp. 22–27. ISSN : 10159770. DOI: 10.1159/ 000255970

    2009

  28. [36]

    Imaging of Ischemic Stroke

    Carlos Leiva-Salinas and Max Wintermark. “Imaging of Ischemic Stroke”. In: Neuroimaging clinics of North America 20.4 (Nov. 2010), p. 455. ISSN : 10525149

    2010

  29. [37]

    Accuracy of electronic health record data for identifying stroke cases in large-scale epidemiological studies: A systematic review from the UK Biobank Stroke Outcomes Group

    Rebecca Woodfield, Ian Grant, and Cathie L.M. Sudlow. “Accuracy of electronic health record data for identifying stroke cases in large-scale epidemiological studies: A systematic review from the UK Biobank Stroke Outcomes Group”. In: PLoS ONE 10.10 (Oct. 2015). ISSN : 19326203...

    doi

  30. [38]

    Online at https://biobank.ndph.ox.ac

    UK Biobank Stroke outcomes : Category 43 . Online at https://biobank.ndph.ox.ac. uk/crystal/label.cgi?id=43 . Website accessed on 2024-07-15. 2024. URL: https: //biobank.ndph.ox.ac.uk/crystal/label.cgi?id=43

    2024

  31. [39]

    mice: Multivariate imputation by chained equations in R

    Stef van Buuren and Karin Groothuis-Oudshoorn. “mice: Multivariate imputation by chained equations in R”. In: Journal of Statistical Software45.3 (Dec. 2011), pp. 1–67. ISSN : 15487660. DOI: https://doi.org/10.18637/jss.v045.i03

    2011 doi

  32. [40]

    Montreal Neurological Institute

    MNI. Montreal Neurological Institute . Online at https : / / www . mcgill . ca / neuro / article/open-science/neuro-launches-new-visual-identity . Website accessed on 2024-06-23. 2024. URL: https://www.mcgill.ca/neuro/article/open-science/ neuro-launches-new-visual-identity

    2024

  33. [41]

    Image processing and Quality Control for the first 10,000 brain imaging datasets from UK Biobank

    Fidel Alfaro-Almagro et al. “Image processing and Quality Control for the first 10,000 brain imaging datasets from UK Biobank”. In: NeuroImage 166 (Feb. 2018), pp. 400–424. ISSN : 1053-8119. DOI: https://doi.org/10.1016/j.neuroimage.2017.10.034

    2018 doi

  34. [42]

    Computational radiomics system to decode the radiographic phenotype

    Joost JM Van Griethuysen et al. “Computational radiomics system to decode the radiographic phenotype”. In: Cancer research 77.21 (2017), e104–e107. DOI: 10.1158/0008-5472.CAN- 17-0339. 21

    2017 doi

  35. [43]

    Attention is all you need

    Ashish Vaswani et al. “Attention is all you need”. In:Advances in Neural Information Process- ing Systems. V ol. 2017-Decem. Neural information processing systems foundation, June 2017, pp. 5999–6009. ISBN : 1706.03762v7. DOI: https://doi.org/10.48550/arXiv.1706. 03762

    2017 doi arXiv

  36. [44]

    CT-BERT: Learning Better Tabular Representations Through Cross-Table Pre-training

    Chao Ye et al. “CT-BERT: Learning Better Tabular Representations Through Cross-Table Pre-training”. In: Proceedings of the ACM Web Conference 2024 Singapore1 (July 2023). DOI: https://doi.org/10.48550/arXiv.2307.04308

    2024 doi arXiv

  37. [45]

    Densely connected convolutional networks

    Gao Huang et al. “Densely connected convolutional networks”. In: Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017 . V ol. 2017-Janua. Institute of Electrical and Electronics Engineers Inc., Aug. 2017, pp. 2261–2269. ISBN : 9781538604571....

    2017 doi

  38. [46]

    A parameter-efficient deep learning approach to predict conversion from mild cognitive impairment to Alzheimer’s disease

    Simeon Spasov et al. “A parameter-efficient deep learning approach to predict conversion from mild cognitive impairment to Alzheimer’s disease”. In: NeuroImage 189 (Apr. 2019), pp. 276–

    2019

  39. [47]

    Adam: A method for stochastic optimization

    Diederik P. Kingma and Jimmy Lei Ba. “Adam: A method for stochastic optimization”. In: 3rd International Conference on Learning Representations, ICLR 2015 - Conference Track Proceedings. International Conference on Learning Representations, ICLR, Dec. 2015. DOI: https://doi.or...

    2015 doi arXiv

  40. [48]

    UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

    Leland McInnes, John Healy, and James Melville. “UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction.” In: (2020). arXiv: 1802.03426 [stat.ML]. URL: https://arxiv.org/abs/1802.03426

    2020 arXiv

  41. [49]

    Captum: A unified and generic model interpretability library for PyTorch

    Narine Kokhlikyan et al. “Captum: A unified and generic model interpretability library for PyTorch”. In: (2020). arXiv: 2009.07896 [cs.LG]

    2020 arXiv

  42. [50]

    Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization

    Ramprasaath R. Selvaraju et al. “Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization”. In: Proceedings of the IEEE International Conference on Computer Vision. V ol. 2017-Octob. Institute of Electrical and Electronics Engineers Inc., Dec. 2017, pp....

    2017 doi

  43. [51]

    MONAI: An open-source framework for deep learning in healthcare

    M. Jorge Cardoso and Li et. al. “MONAI: An open-source framework for deep learning in healthcare”. In: (2022). URL: https : / / arxiv . org / abs / 2211 . 02701v1 % 20http : //arxiv.org/abs/2211.02701

    2022 arXiv

  44. [52]

    Deep residual learning for image recognition

    Kaiming He et al. “Deep residual learning for image recognition”. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. V ol. 2016-Decem. IEEE Computer Society, Dec. 2016, pp. 770–778. ISBN : 9781467388504

    2016

  45. [53]

    Classification of Subtypes of Ischemic Stroke

    Harold P. Adams and José Biller. “Classification of Subtypes of Ischemic Stroke”. In:Stroke 46.5 (2015), e114–e117. DOI: 10.1161/STROKEAHA.114.007773 . URL: https://www. ahajournals.org/doi/abs/10.1161/STROKEAHA.114.007773

    2015 doi

  46. [54]

    An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies

    Peter C. Austin. “An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies”. In: Multivariate Behavioral Research 46.3 (2011), pp. 399–424. DOI: 10.1080/00273171.2011.568786 . URL: https://www.ncbi.nlm. nih.gov/pmc/articles/P...

    2011

  47. [56]

    White matter hyperintensities and recurrent stroke risk in patients with stroke with small-vessel disease

    J-H Park et al. “White matter hyperintensities and recurrent stroke risk in patients with stroke with small-vessel disease”. In: European Journal of Neurology 26.6 (2019), pp. 911–918. DOI: https://doi.org/10.1111/ene.13908. 22

    2019 doi

  48. [57]

    MRI assessment of whole-brain structural changes in aging

    Hui Guo et al. “MRI assessment of whole-brain structural changes in aging”. In: Clinical Interventions in Aging 12 (Aug. 2017), pp. 1251–1270. ISSN : 11781998. DOI: https://doi. org/10.1016/j.arr.2018.11.004

    2017 doi

  49. [58]

    Multimodal brain-age prediction and cardiovascular risk: The Whitehall II MRI sub-study

    Ann-Marie G De Lange et al. “Multimodal brain-age prediction and cardiovascular risk: The Whitehall II MRI sub-study”. In: NeuroImage 222 (2020), p. 117292. DOI: 10 . 1016 / j . neuroimage.2020.117292

    2020

  50. [59]

    Association of brain age, lesion volume, and functional outcome in patients with stroke

    Sook-Lei Liew et al. “Association of brain age, lesion volume, and functional outcome in patients with stroke”. In: Neurology 100.20 (2023), e2103–e2113. DOI: https://doi.org/ 10.1212/WNL.0000000000207219

    2023 doi

  51. [60]

    Brain age predicts mortality

    James H Cole et al. “Brain age predicts mortality”. In: Molecular psychiatry 23.5 (2018), pp. 1385–1392. DOI: https://doi.org/10.1038/mp.2017.62

    2018 doi

  52. [61]

    Towards performance-maximizing neural network pruning via global channel attention

    Yingchun Wang et al. “Towards performance-maximizing neural network pruning via global channel attention”. In: Neural Networks 171 (2024), pp. 104–113. ISSN : 0893-6080. DOI: https://doi.org/10.1016/j.neunet.2023.11.065

    2024 doi

  53. [62]

    Score -CAM: Score-Weighted Visual Explanations for Convolutional Neural Networks

    Haofan Wang et al. “Score -CAM: Score-Weighted Visual Explanations for Convolutional Neural Networks”. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. 2020, pp. 24–25. DOI: 10.1109/CVPRW50498.2020.00124. URL: https://arxiv.org/a...

    2020

  54. [63]

    Opti-CAM: Optimizing Saliency Maps for Interpretability

    Hanwei Zhang et al. “Opti-CAM: Optimizing Saliency Maps for Interpretability”. In:Computer Vision and Image Understanding xxxx.yy (2024). Preprint available at arXiv:2301.07002. DOI: 10.1016/j.cviu.2024.104101 . URL: https://doi.org/10.1016/j.cviu.2024. 104101

    2024 arXiv

  55. [64]

    Reliable Saliency Maps for Weakly -Supervised Localization of Disease Patterns

    Maximilian Möller et al. “Reliable Saliency Maps for Weakly -Supervised Localization of Disease Patterns”. In: Interpretable and Annotation-Efficient Learning for Medical Image Computing. V ol. xxxx. Lecture Notes in Computer Science. Springer, 2020, pp. 63–72.DOI: 10.1007/978...

    2020 doi

  56. [104]

    DOI: 10.1159/000366411

    ISSN : 14219824. DOI: 10.1159/000366411

    doi

  57. [287]

    DOI: https://doi.org/10.1016/j.neuroimage.2019.01.031

    ISSN : 10959572. DOI: https://doi.org/10.1016/j.neuroimage.2019.01.031

    2019 doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.