pith. sign in

arxiv: 2605.18878 · v1 · pith:5OMOZPXPnew · submitted 2026-05-16 · 📡 eess.SP · cs.CV· cs.LG· eess.IV

Prognostic Value of Lung Ultrasound Biomarkers for Readmission Risk in Congestive Heart Failure: A Pilot Data-Driven Analysis

Jana Armouti , Laura Hutchins , Jacob Duplantis , Thomas Deiss , Thales Nogueira Gomes , Keyur H. Patel , Seema Walvekar , Shane Guillory
show 7 more authors
Thomas H. Fox Amita Krishnan Ricardo Rodriguez Bennett DeBoisblanc Deva Ramanan John Galeotti Gautam Gare
This is my paper

Pith reviewed 2026-05-20 15:37 UTC · model grok-4.3

classification 📡 eess.SP cs.CVcs.LGeess.IV
keywords lung ultrasoundcongestive heart failurereadmission predictionmachine learningtemporal difference featuresmulti-view fusionpoint-of-care ultrasoundpleural line abnormalities
0
0 comments X

The pith

Lung ultrasound from dependent lower regions combined with temporal changes between scans predicts 30-day heart failure readmission with F1 of 0.80.

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

This pilot study tests whether point-of-care lung ultrasound scans acquired during hospitalization can forecast which congestive heart failure patients will return within 30 days. Researchers extract spatiotemporal features from video models and evaluate hand-crafted biomarkers across different lung views, time representations, and fusion strategies. The work addresses a gap where standard risk tools rely on non-imaging data and show only modest accuracy, while lung ultrasound directly visualizes the pulmonary congestion central to decompensation. A sympathetic reader would care because readmissions drive substantial morbidity and cost, and a bedside imaging marker could support more precise discharge decisions if the signals prove reliable.

Core claim

Dependent lower-lung regions carry the strongest prognostic signal, consistent with greater susceptibility to hydrostatic congestion. Temporal difference features between sequential examinations substantially outperform single-timepoint representations. Multi-view feature concatenation yields the best overall performance, with the top multilayer perceptron model achieving an F1 score of 0.80 (95% CI: 0.62-0.96). Biomarker analysis shows pleural-line abnormalities including breaks and indentations are as informative as canonical A-line and B-line markers.

What carries the argument

Quantitative spatiotemporal embeddings extracted from a pretrained Temporal Shift Module ResNet-18 encoder on B-mode lung ultrasound videos, fused via multi-view concatenation and temporal differencing with interpretable biomarker features.

If this is right

  • Dependent lower-lung regions should receive priority in future lung ultrasound protocols for heart failure risk assessment.
  • Capturing temporal differences between serial scans during hospitalization adds substantial predictive value over static single-timepoint images.
  • Multi-view feature concatenation enables clinically usable prediction performance in machine learning models for readmission risk.
  • Pleural-line abnormalities can serve as practical biomarkers alongside or in place of traditional A-lines and B-lines.

Where Pith is reading between the lines

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

  • Focusing ultrasound effort on lower lung zones could shorten examination time while retaining most prognostic information.
  • Tracking congestion trajectory during the hospital stay might allow earlier therapy adjustments before discharge decisions.
  • The same imaging and modeling approach could be tested for predicting fluid overload events in related conditions such as acute kidney injury.

Load-bearing premise

The pilot cohort is representative of the broader congestive heart failure population and readmission labels are not substantially confounded by unmeasured clinical or social factors.

What would settle it

A larger independent validation cohort where lower-lung temporal difference features fail to reach an F1 score above 0.65 for 30-day readmission prediction.

Figures

Figures reproduced from arXiv: 2605.18878 by Amita Krishnan, Bennett DeBoisblanc, Deva Ramanan, Gautam Gare, Jacob Duplantis, Jana Armouti, John Galeotti, Keyur H. Patel, Laura Hutchins, Ricardo Rodriguez, Seema Walvekar, Shane Guillory, Thales Nogueira Gomes, Thomas Deiss, Thomas H. Fox.

Figure 1
Figure 1. Figure 1: CHF 30-day readmission prediction pipeline. Six standardized LUS views (Right/Left 1–3, spanning upper anterior to dependent posterior regions) are acquired at two or more time points during the index hospital￾ization. B-mode video clips are encoded by a pretrained TSM–ResNet-18 to produce 512-dimensional spatiotemporal embeddings. Temporal difference fea￾tures (∆ = Day2 − Day1 , capturing congestion traje… view at source ↗
Figure 2
Figure 2. Figure 2: Imaging biomarkers complement clinical EHR. Left: EHR vs. expert biomarker performance. On a 25-patient held-out test cohort with available clinical EHR records, the clinical EHR model (purple) and expert LUS biomarker model (blue) achieve identical performance. In contrast, the combined model (red) improves the F1-score by +0.127 and correctly identifies all read￾mitted patients, highlighting the compleme… view at source ↗
Figure 3
Figure 3. Figure 3: Multi-view fusion heatmap (MLP, All Views). F1-scores across five fusion strategies (rows) and two temporal representations (columns). Feature concatenation with temporal difference achieves the best performance (0.80). Consistently higher scores in the “Difference” column indicate that temporal representation has greater impact than fusion strategy. 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Left-1 Left-2 Left-3 Right… view at source ↗
Figure 4
Figure 4. Figure 4: Weighted F1 (95% CI) for temporal concatenation (gray) vs. temporal [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Cross-lung training. Comparison of MLP performance with and with￾out left–right pooled augmentation across all LUS views. Fixed conditions: Day 1 vs. Day 2, temporal difference; for “All Views”, embeddings are concatenated across views. Cross-lung pooling improves or maintains performance for most views while preserving the best multi-view result. As shown in [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Grad-CAM visualizations for paired Day 1 (top row) and Day 2 (bot￾tom row) LUS clips. (a) Readmitted patient showing marked Day 2 effusion and corresponding focused model attention. (b) Not-readmitted patient demonstrat￾ing stable Day 2 appearance with diffuse, low-intensity attention [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
read the original abstract

Hospital readmission within 30 days of discharge is a leading driver of morbidity, mortality, and avoidable healthcare expenditure in congestive heart failure (CHF). Current clinical risk stratification tools rely primarily on non-imaging data and exhibit limited predictive performance. Point-of-care lung ultrasound (LUS) offers a sensitive, noninvasive window into the pulmonary congestion that characterizes CHF decompensation, yet its prognostic utility for readmission prediction remains largely unexplored. We present a pilot feasibility study, the first systematic machine learning study using B-mode LUS acquired during hospitalization to predict 30-day CHF readmission. Quantitative spatiotemporal embeddings are extracted from a pretrained Temporal Shift Module (TSM) ResNet-18 encoder, and interpretable biomarker features are separately evaluated. Through structured ablations over lung view, temporal representation, multi-view fusion, and cross-lung augmentation, we identify the key imaging factors driving readmission risk. Our findings reveal that (1) dependent lower-lung regions (Left-3, Right-3) carry the strongest prognostic signal, consistent with their greater susceptibility to hydrostatic congestion; (2) temporal difference features between sequential examinations substantially outperform single-timepoint representations, highlighting the importance of capturing disease trajectory; and (3) multi-view feature concatenation yields the best overall performance, with our top MLP model achieving an F1 score of 0.80 (95% CI: 0.62-0.96). Biomarker analysis further reveals that pleural-line abnormalities, including breaks and indentations, are as informative as the canonical A-line and B-line markers. These results support POCUS-derived biomarkers as practical, interpretable tools for noninvasive CHF risk stratification.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. This pilot feasibility study is the first systematic machine-learning analysis of B-mode lung ultrasound (LUS) acquired during hospitalization for predicting 30-day readmission in congestive heart failure. Quantitative spatiotemporal embeddings are extracted from a pretrained TSM-ResNet-18 encoder; interpretable biomarker features are also evaluated. Structured ablations over lung view, temporal representation (single-timepoint vs. difference), multi-view fusion, and cross-lung augmentation identify dependent lower-lung regions (Left-3, Right-3), temporal-difference features, and multi-view concatenation as superior, with the best MLP reaching F1 = 0.80 (95 % CI 0.62–0.96). Pleural-line abnormalities are reported as comparably informative to classic A- and B-line markers.

Significance. If replicated in larger, externally validated cohorts, the work could establish point-of-care LUS as a practical, interpretable adjunct for noninvasive CHF risk stratification, addressing a gap left by current non-imaging scores. The systematic ablation design and emphasis on clinically meaningful regions and temporal trajectories are positive features that could guide future imaging-biomarker studies.

major comments (3)
  1. [Results] Results (performance table / ablation summary): the reported F1 = 0.80 (CI 0.62–0.96) implies a small held-out test set (likely n < 50–60 independent cases). With multiple ablations performed over views, temporal encodings, and fusion strategies, this width renders the ranking of multi-view concatenation, lower-lung views, and temporal-difference features statistically fragile and vulnerable to a few influential patients or label errors.
  2. [Methods] Methods (cohort and validation description): the manuscript supplies neither the total number of patients/examinations nor the cross-validation scheme (patient-level vs. examination-level splitting). Without explicit leakage controls across temporally or spatially correlated LUS views, the superiority claims for multi-view and temporal-difference representations cannot be confidently separated from data leakage.
  3. [Discussion] Discussion / limitations: the central claim that the identified biomarkers generalize rests on the untested assumption that the small pilot cohort is representative and that 30-day readmission labels are free of substantial unmeasured clinical or social confounding. No sensitivity analysis or comparison against established clinical scores (LACE, HOSPITAL) is presented to quantify incremental value.
minor comments (2)
  1. [Abstract] Abstract: the phrase 'structured ablations' is used without stating the exact number of views, temporal points, or augmentation variants tested; adding these numbers would improve reproducibility.
  2. [Figures] Figure legends: axis labels and color coding for the ablation heat-maps should explicitly indicate whether performance is reported on the same or independent test folds.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and detailed review of our pilot feasibility study. We address each major comment below, indicating where revisions have been made to improve clarity, transparency, and acknowledgment of limitations.

read point-by-point responses

  1. Referee: [Results] Results (performance table / ablation summary): the reported F1 = 0.80 (CI 0.62–0.96) implies a small held-out test set (likely n < 50–60 independent cases). With multiple ablations performed over views, temporal encodings, and fusion strategies, this width renders the ranking of multi-view concatenation, lower-lung views, and temporal-difference features statistically fragile and vulnerable to a few influential patients or label errors.

    Authors: We agree that the wide confidence interval indicates a small test set and that the ablation rankings should be interpreted cautiously in a pilot study. The primary objective was to demonstrate feasibility and highlight promising directions rather than to claim definitive superiority. In the revised manuscript, we have explicitly reported the test set size, emphasized the exploratory character of the ablations in the Results and Discussion, and added a clear statement that larger external validation is required before the observed patterns can be considered robust. revision: partial

  2. Referee: [Methods] Methods (cohort and validation description): the manuscript supplies neither the total number of patients/examinations nor the cross-validation scheme (patient-level vs. examination-level splitting). Without explicit leakage controls across temporally or spatially correlated LUS views, the superiority claims for multi-view and temporal-difference representations cannot be confidently separated from data leakage.

    Authors: We thank the referee for identifying this omission. The revised Methods section now states the total number of patients and examinations, specifies that all splitting and cross-validation were performed at the patient level, and describes the steps taken to prevent leakage from temporally or spatially correlated views. These additions should allow readers to evaluate the validity of the multi-view and temporal-difference comparisons. revision: yes

  3. Referee: [Discussion] Discussion / limitations: the central claim that the identified biomarkers generalize rests on the untested assumption that the small pilot cohort is representative and that 30-day readmission labels are free of substantial unmeasured clinical or social confounding. No sensitivity analysis or comparison against established clinical scores (LACE, HOSPITAL) is presented to quantify incremental value.

    Authors: We acknowledge that generalizability and potential confounding remain untested in this small pilot cohort. The revised Discussion now explicitly lists these assumptions as limitations and notes the absence of sensitivity analyses due to sample size. A head-to-head comparison with LACE or HOSPITAL scores was not possible with the data collected in this pilot; we have therefore framed incremental value as an important objective for future, larger studies rather than a claim supported by the current results. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical ML performance on held-out imaging data

full rationale

The paper conducts a pilot feasibility study extracting spatiotemporal embeddings from a pretrained TSM-ResNet-18 on B-mode lung ultrasound images, then performs structured ablations over views, temporal differences, and multi-view fusion to report F1 scores on readmission prediction. All reported results are data-driven empirical measurements rather than derivations; no equations, fitted parameters renamed as predictions, or self-referential definitions appear. Claims rest on observable performance differences across ablations in a small cohort, with no load-bearing self-citations or uniqueness theorems reducing the central findings to prior author work by construction. The analysis is therefore self-contained against external benchmarks of model evaluation on independent cases.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim rests on transfer learning from a video action-recognition model to lung ultrasound, the representativeness of a small pilot dataset, and the accuracy of readmission outcome labels.

free parameters (2)
  • MLP architecture and hyperparameters
    Model depth, width, and training choices are selected to fit the pilot data.
  • View and temporal ablation thresholds
    Decisions on which lung views and difference features to retain are informed by performance on the same limited cohort.
axioms (2)
  • domain assumption Pretrained TSM ResNet-18 extracts clinically relevant spatiotemporal features from LUS without domain-specific fine-tuning.
    Transfer learning assumption invoked when using the encoder for quantitative embeddings.
  • domain assumption 30-day readmission labels are accurate and free of substantial confounding.
    Standard clinical prediction assumption required for biomarker validity.

pith-pipeline@v0.9.0 · 5912 in / 1512 out tokens · 53674 ms · 2026-05-20T15:37:10.913757+00:00 · methodology

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

296 extracted references · 296 canonical work pages · 22 internal anchors

  1. [1]

    FirstName LastName , title =

    work page

  2. [2]

    FirstName Alpher , title =

    work page

  3. [3]

    Journal of Foo , volume = 13, number = 1, pages =

    FirstName Alpher and FirstName Fotheringham-Smythe , title =. Journal of Foo , volume = 13, number = 1, pages =

    work page

  4. [4]

    Journal of Foo , volume = 14, number = 1, pages =

    FirstName Alpher and FirstName Fotheringham-Smythe and FirstName Gamow , title =. Journal of Foo , volume = 14, number = 1, pages =

    work page

  5. [5]

    FirstName Alpher and FirstName Gamow , title =

    work page

  6. [6]

    Zhou, Luowei and Palangi, Hamid and Zhang, Lei and Hu, Houdong and Corso, Jason and Gao, Jianfeng , booktitle=

    work page

  7. [7]

    Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina , journal=

    work page

  8. [8]

    2021 , url =

    Yinghsan Chang and Mridu Narang and Hisami Suzuki and Guihong Cao and Jianfeng Gao and Yonatan Bisk , journal =. 2021 , url =

    work page 2021

  9. [9]

    Lawrence and Parikh, Devi , journal =

    Antol, Stanislaw and Agrawal, Aishwarya and Lu, Jiasen and Mitchell, Margaret and Batra, Dhruv and Zitnick, C. Lawrence and Parikh, Devi , journal =

    work page

  10. [10]

    Goyal, Yash and Khot, Tejas and Agrawal, Aishwarya and Summers-Stay, Douglas and Batra, Dhruv and Parikh, Devi , journal =

    work page

  11. [11]

    Das, Abhishek and Agrawal, Harsh and Zitnick, C Lawrence and Parikh, Devi and Batra, Dhruv , eprint =

    work page

  12. [12]

    Marino, Kenneth and Rastegari, Mohammad and Farhadi, Ali and Mottaghi, Roozbeh , booktitle=

    work page

  13. [13]

    European Conference on Computer Vision , pages=

    Lin, Tsung-Yi and Maire, Michael and Belongie, Serge and Hays, James and Perona, Pietro and Ramanan, Deva and Doll. European Conference on Computer Vision , pages=. 2014 , organization=

    work page 2014

  14. [14]

    International Journal of Computer Vision , volume=

    Visual genome: Connecting language and vision using crowdsourced dense image annotations , author=. International Journal of Computer Vision , volume=. 2017 , publisher=

    work page 2017

  15. [15]

    Talmor, Alon and Yoran, Ori and Catav, Amnon and Lahav, Dan and Wang, Yizhong and Asai, Akari and Ilharco, Gabriel and Hajishirzi, Hannaneh and Berant, Jonathan , journal=

    work page

  16. [16]

    Advances in neural information processing systems , volume=

    Faster r-cnn: Towards real-time object detection with region proposal networks , author=. Advances in neural information processing systems , volume=

    work page

  17. [17]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Zadeh, Amir and Chan, Michael and Liang, Paul Pu and Tong, Edmund and Morency, Louis-Philippe , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    work page

  18. [18]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Li, Wei-Hong and Liu, Xialei and Bilen, Hakan , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2022 , pages =

    work page 2022

  19. [19]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =

    Fan, Haoqi and Xiong, Bo and Mangalam, Karttikeya and Li, Yanghao and Yan, Zhicheng and Malik, Jitendra and Feichtenhofer, Christoph , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =. 2021 , pages =

    work page 2021

  20. [20]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Video swin transformer , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  21. [21]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Vivit: A video vision transformer , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  22. [22]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Tsm: Temporal shift module for efficient video understanding , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  23. [23]

    arXiv preprint arXiv:2206.08398 , year=

    Learning Generic Lung Ultrasound Biomarkers for Decoupling Feature Extraction from Downstream Tasks , author=. arXiv preprint arXiv:2206.08398 , year=

    work page arXiv

  24. [24]

    Simple, Safe, Same: Lung Ultrasound for COVID-19 - Tabular View - ClinicalTrials.gov , author=

    work page

  25. [25]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Swin transformer: Hierarchical vision transformer using shifted windows , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  26. [26]

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

    An image is worth 16x16 words: Transformers for image recognition at scale , author=. arXiv preprint arXiv:2010.11929 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  27. [27]

    ICML , volume=

    Is space-time attention all you need for video understanding? , author=. ICML , volume=

    work page

  28. [28]

    arXiv preprint arXiv:2203.12602 , year=

    Videomae: Masked autoencoders are data-efficient learners for self-supervised video pre-training , author=. arXiv preprint arXiv:2203.12602 , year=

    work page arXiv

  29. [29]

    arXiv preprint arXiv:2201.07357 , year=

    Weakly Supervised Contrastive Learning for Better Severity Scoring of Lung Ultrasound , author=. arXiv preprint arXiv:2201.07357 , year=

    work page arXiv

  30. [30]

    Journal of Ultrasound in Medicine , volume=

    On the impact of different lung ultrasound imaging protocols in the evaluation of patients affected by coronavirus disease 2019: how many acquisitions are needed? , author=. Journal of Ultrasound in Medicine , volume=. 2021 , publisher=

    work page 2019

  31. [31]

    Cureus , volume=

    Application of lung ultrasound in critical care setting: a review , author=. Cureus , volume=. 2019 , publisher=

    work page 2019

  32. [32]

    Applied Sciences , volume=

    Accelerating detection of lung pathologies with explainable ultrasound image analysis , author=. Applied Sciences , volume=. 2021 , publisher=

    work page 2021

  33. [33]

    IEEE transactions on medical imaging , volume=

    Deep learning for classification and localization of COVID-19 markers in point-of-care lung ultrasound , author=. IEEE transactions on medical imaging , volume=. 2020 , publisher=

    work page 2020

  34. [34]

    IEEE journal of biomedical and health informatics , volume=

    Localizing B-lines in lung ultrasonography by weakly supervised deep learning, in-vivo results , author=. IEEE journal of biomedical and health informatics , volume=. 2019 , publisher=

    work page 2019

  35. [35]

    Medical image analysis , volume=

    Modality alignment contrastive learning for severity assessment of COVID-19 from lung ultrasound and clinical information , author=. Medical image analysis , volume=. 2021 , publisher=

    work page 2021

  36. [36]

    2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI) , pages=

    Dense pixel-labeling for reverse-transfer and diagnostic learning on lung ultrasound for COVID-19 and pneumonia detection , author=. 2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI) , pages=. 2021 , organization=

    work page 2021

  37. [37]

    International conference on machine learning , pages=

    A simple framework for contrastive learning of visual representations , author=. International conference on machine learning , pages=. 2020 , organization=

    work page 2020

  38. [38]

    arXiv preprint arXiv:2010.00747 , year=

    Contrastive learning of medical visual representations from paired images and text , author=. arXiv preprint arXiv:2010.00747 , year=

    work page arXiv 2010

  39. [39]

    IEEE Journal of Biomedical and Health Informatics , volume=

    Contrastive cross-site learning with redesigned net for COVID-19 CT classification , author=. IEEE Journal of Biomedical and Health Informatics , volume=. 2020 , publisher=

    work page 2020

  40. [40]

    and Garofano, Matteo and Ghislandi, Simone and Giannoni, Margherita and Marini, Giorgia and Vainieri, Milena , number =

    Bosa, Iris and Castelli, Adriana and Castelli, Michele and Ciani, Oriana and Compagni, Amelia and Galizzi, Matteo M. and Garofano, Matteo and Ghislandi, Simone and Giannoni, Margherita and Marini, Giorgia and Vainieri, Milena , number =. 2022 , journal =. doi:10.1017/S1744133121000141 , issn =

    work page doi:10.1017/s1744133121000141 2022

  41. [41]

    2021 , journal =

    Roshankhah, Roshan and Karbalaeisadegh, Yasamin and Greer, Hastings and Mento, Federico and Soldati, Gino and Smargiassi, Andrea and Inchingolo, Riccardo and Torri, Elena and Perrone, Tiziano and Aylward, Stephen and Demi, Libertario and Muller, Marie , number =. 2021 , journal =. doi:10.1121/10.0007272 , issn =

    work page doi:10.1121/10.0007272 2021

  42. [42]

    and Aisen, Paul S

    Weiner, Michael W. and Aisen, Paul S. and Jack, Clifford R. and Jagust, William J. and Trojanowski, John Q. and Shaw, Leslie and Saykin, Andrew J. and Morris, John C. and Cairns, Nigel and Beckett, Laurel A. and Toga, Arthur and Green, Robert and Walter, Sarah and Soares, Holly and Snyder, Peter and Siemers, Eric and Potter, William and Cole, Patricia E. ...

    work page doi:10.1016/j.jalz.2010.03.007 2010

  43. [43]

    and Smith, A.B

    Jones, C.D. and Smith, A.B. and Roberts, E.F. Paper Title. Proceedings Title. 2003

    work page 2003

  44. [44]

    Bias in Error Estimation When Using Cross-Validation for Model Selection.” BMC Bioinformatics, 7(1), 91 , volume =

    Varma, Sudhir and Simon, Richard , year =. Bias in Error Estimation When Using Cross-Validation for Model Selection.” BMC Bioinformatics, 7(1), 91 , volume =. BMC bioinformatics , doi =

    work page

  45. [45]

    Statistical Comparisons of Classifiers over Multiple Data Sets , journal =

    Janez Dem. Statistical Comparisons of Classifiers over Multiple Data Sets , journal =. 2006 , volume =

    work page 2006

  46. [46]

    2020 , journal =

    G. 2020 , journal =. doi:10.1016/j.compbiomed.2020.104036 , issn =

    work page doi:10.1016/j.compbiomed.2020.104036 2020

  47. [47]

    2020 , journal =

    Jarosik, Piotr and Klimonda, Ziemowit and Lewandowski, Marcin and Byra, Michal , pages =. 2020 , journal =. doi:10.1016/j.bbe.2020.04.002 , keywords =

    work page doi:10.1016/j.bbe.2020.04.002 2020

  48. [48]

    2019 , journal =

    Byra, Michal and Galperin, Michael and Ojeda-Fournier, Haydee and Olson, Linda and O'Boyle, Mary and Comstock, Christopher and Andre, Michael , number =. 2019 , journal =. doi:10.1002/mp.13361 , issn =

    work page doi:10.1002/mp.13361 2019

  49. [49]

    2020 , journal =

    Biewald, Lukas , url =. 2020 , journal =

    work page 2020

  50. [50]

    2016 , journal =

    Ribeiro, Marco Tulio and Singh, Sameer and Guestrin, Carlos , month =. 2016 , journal =. doi:10.18653/v1/n16-3020 , arxivId =

    work page doi:10.18653/v1/n16-3020 2016

  51. [51]

    Schiffer, Christian and Harmeling, Stefan and Amunts, Katrin and Dickscheid, Timo , url =

    work page

  52. [52]

    2002 , journal =

    Rastle, Kathleen and Harrington, Jonathan and Coltheart, Max , number =. 2002 , journal =. doi:10.1080/02724980244000099 , issn =

    work page doi:10.1080/02724980244000099 2002

  53. [53]

    Liu, Siqi and Xu, Daguang and Zhou, S Kevin and Mertelmeier, Thomas and Wicklein, Julia and Jerebko, Anna and Grbic, Sasa and Pauly, Olivier and Cai, Weidong and Comaniciu, Dorin , arxivId =

    work page

  54. [54]

    Kevin and Mertelmeier, Thomas and Wicklein, Julia and Jerebko, Anna and Grbic, Sasa and Pauly, Olivier and Cai, Weidong and Comaniciu, Dorin , month =

    Liu, Siqi and Xu, Daguang and Zhou, S. Kevin and Mertelmeier, Thomas and Wicklein, Julia and Jerebko, Anna and Grbic, Sasa and Pauly, Olivier and Cai, Weidong and Comaniciu, Dorin , month =. 2017 , journal =

    work page 2017

  55. [55]

    2021 , journal =

    Huang, Huimin and Zhou, Nan and Lin, Lanfen and Hu, Hongjie and Iwamoto, Yutaro and Han, Xian Hua and Chen, Yen Wei and Tong, Ruofeng , pages =. 2021 , journal =. doi:10.1007/978-3-030-87196-3

    work page doi:10.1007/978-3-030-87196-3 2021

  56. [56]

    and De la Torre, Fernando , month =

    Xu, Xiangyu and Chen, Hao and Moreno-Noguer, Francesc and Jeni, Laszlo A. and De la Torre, Fernando , month =. 2021 , journal =

    work page 2021

  57. [57]

    Hendrycks, Dan and Gimpel, Kevin , isbn =

    work page

  58. [58]

    and Karlinsky, Leonid and Codella, James V

    Guo, Yunhui and Codella, Noel C. and Karlinsky, Leonid and Codella, James V. and Smith, John R. and Saenko, Kate and Rosing, Tajana and Feris, Rogerio , month =. 2020 , booktitle =. doi:10.1007/978-3-030-58583-9

    work page doi:10.1007/978-3-030-58583-9 2020

  59. [59]

    A causal framework for discovering and removing direct and indirect discrimination

    2016 , author =. doi:10.48550/arxiv.1611.07509 , arxivId =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1611.07509 2016

  60. [60]

    Li, L. U. and Shangguan, W. E.I. and Deng, Y. I. and Mao, Jiafu and Pan, Jinjing and Wei, N. A.N. and Yuan, H. U.A. and Zhang, Shupeng and Zhang, Yonggen and Dai, Yongjiu , number =. 2020 , journal =. doi:10.1175/JHM-D-19-0209.1 , issn =

    work page doi:10.1175/jhm-d-19-0209.1 2020

  61. [61]

    2022 , journal =

    Schuhmacher, David and Sch. 2022 , journal =. doi:10.1016/J.MEDIA.2022.102594 , issn =

    work page doi:10.1016/j.media.2022.102594 2022

  62. [62]

    Sarkar, Anirban and Vijaykeerthy, Deepak and Sarkar, Anindya and Balasubramanian, Vineeth N , arxivId =

    work page

  63. [63]

    doi:10.48550/arxiv.2209.11142 , arxivId =

    2022 , author =. doi:10.48550/arxiv.2209.11142 , arxivId =

    work page doi:10.48550/arxiv.2209.11142 2022

  64. [64]

    and Malloni, W

    Wein, S. and Malloni, W. M. and Tom. 2021 , journal =. doi:10.1038/s41598-021-87411-8 , issn =

    work page doi:10.1038/s41598-021-87411-8 2021

  65. [65]

    2022 , journal =

    Zhang, Lei and Chu, Ziyue and Liang, Longfei and Yang, Wen Chi , pages =. 2022 , journal =. doi:10.1109/PRAI55851.2022.9904224 , keywords =

    work page doi:10.1109/prai55851.2022.9904224 2022

  66. [66]

    Li, Qing and Huang, Siyuan and Hong, Yining and Zhu, Yixin and Wu, Ying Nian and Zhu, Song-Chun , url =

    work page

  67. [67]

    Tang, Bo and He, Haibo , arxivId =

    work page

  68. [68]

    A Meta-Transfer Objective for Learning to Disentangle Causal Mechanisms

    2019 , author =. doi:10.48550/arxiv.1901.10912 , arxivId =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1901.10912 2019

  69. [69]

    2008 , journal =

    Basarab, Adrian and Liebgott, Hervé and Morestin, Fabrice and Lyshchik, Andrej and Higashi, Tatsuya and Asato, Ryo and Delachartre, Philippe , number =. 2008 , journal =. doi:10.1016/J.MEDIA.2007.10.007 , issn =

    work page doi:10.1016/j.media.2007.10.007 2008

  70. [70]

    2014 , booktitle =

    Wang, Dan and Shang, Yi , month =. 2014 , booktitle =

    work page 2014

  71. [71]

    & Chen, C

    Bafghi, Reza Akbarian and Gurari, Danna , pages =. 2023 , journal =. doi:10.1109/CVPR52729.2023.01560 , issn =

    work page doi:10.1109/cvpr52729.2023.01560 2023

  72. [72]

    and Patel, Chirag J

    Andrews, Bryan and Wongchokprasitti, Chirayu and Visweswaran, Shyam and Lakhani, Chirag M. and Patel, Chirag J. and Cooper, Gregory F. , month =. 2023 , journal =. doi:10.1016/J.ARTMED.2023.102546 , issn =

    work page doi:10.1016/j.artmed.2023.102546 2023

  73. [73]

    Ahn, Donghoon and Kang, Jiwon and Lee, Sanghyun and Min, Jaewon and Kim, Minjae and Jang, Wooseok and Cho, Hyoungwon and Paul, Sayak and Kim, Seonhwa and Hwan, Kyong and †1, Jin and Kim, Seungryong , url =

    work page

  74. [74]

    2013 , journal =

    Hu, Feng and Li, Hang , volume =. 2013 , journal =. doi:10.1155/2013/694809 , issn =

    work page doi:10.1155/2013/694809 2013

  75. [75]

    Wu, Wei and Lu, Fan and Wang, Yunnan and Yang, Shuai and Liu, Shi and Wang, Fangjing and Zhu, Qian and Sun, He and Wang, Yong and Ma, Shuailei and Ren, Yiyu and Zhang, Kejia and Yu, Hui and Zhao, Jingmei and Zhou, Shuai and Qiu, Zhenqi and Xiong, Houlong and Wang, Ziyu and Wang, Zechen and Cheng, Ran and Li, Yong-Lu and Huang, Yongtao and Zhu, Xing and Sh...

    work page

  76. [76]

    Arbel, Julyan and Pitas, Konstantinos and Vladimirova, Mariia and Fortuin, Vincent , arxivId =

    work page

  77. [77]

    Kohl, Simon A A and Romera-Paredes, Bernardino and Meyer, Clemens and De Fauw, Jeffrey and Ledsam, Joseph R and Maier-Hein, Klaus H and Eslami, S M Ali and Jimenez Rezende, Danilo and Ronneberger, Olaf , arxivId =

    work page

  78. [78]

    doi:10.48550/arxiv.2206.05498 , arxivId =

    2022 , author =. doi:10.48550/arxiv.2206.05498 , arxivId =

    work page doi:10.48550/arxiv.2206.05498 2022

  79. [79]

    2019 , journal =

    Remeseiro, Beatriz and Bolon-Canedo, Veronica , pages =. 2019 , journal =. doi:10.1016/J.COMPBIOMED.2019.103375 , issn =

    work page doi:10.1016/j.compbiomed.2019.103375 2019

  80. [80]

    2020 , booktitle =

    Chen, Ting and Kornblith, Simon and Norouzi, Mohammad and Hinton, Geoffrey , month =. 2020 , booktitle =

    work page 2020

Showing first 80 references.