pith. sign in

arxiv: 2605.26483 · v1 · pith:VCUVSVBKnew · submitted 2026-05-26 · 💻 cs.CV

Clinically-Grounded Counterfactual Reasoning for Medical Video Diagnosis

Jianzhe Gao , Churan Wang , Weiyi Zhang , Jianghua Li , Li-An Li , Wenguan Wang , Yixin Zhu , Yizhou Wang This is my paper

Pith reviewed 2026-06-29 18:33 UTC · model grok-4.3

classification 💻 cs.CV
keywords counterfactual reasoningmedical video diagnosisdiffusion modelsclinical rulescolposcopycolonoscopyweakly supervised learningtemporal consistency
0
0 comments X

The pith

MedVCR uses a diffusion model to generate counterfactual tissue changes and clinical rules to improve medical video diagnosis accuracy.

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

The paper presents MedVCR as a framework that shifts medical video diagnosis from end-to-end appearance-based learning to explicit counterfactual reasoning that mimics how clinicians compare observed tissue behavior against expected pathological alternatives. It builds a diffusion-based generator to produce synthetic evolutions of tissue under targeted disease states, then trains representations under three explicit clinical rules that enforce temporal consistency, separate healthy from diseased patterns, and align real and generated sequences. A dual prediction head combines whole-video assessment with per-frame counterfactual comparisons, producing measurable gains on both fully labeled colposcopy videos and weakly labeled colonoscopy videos.

Core claim

MedVCR establishes that diagnostic performance in dynamic medical videos rises when a diffusion generator creates tissue sequences conditioned on specific pathological states and the resulting pairs are constrained by temporal consistency, pathological separability, and counterfactual alignment; the dual prediction strategy then fuses video-level and frame-level signals to reach higher accuracy than standard supervised or weakly supervised baselines.

What carries the argument

The Counterfactual Generator, a diffusion model that synthesizes tissue evolution under specified pathological states, together with the three clinical rules that structure representation learning.

If this is right

  • Performance rises 2.6 to 10.2 percent over leading baselines on both fully supervised colposcopy and weakly supervised colonoscopy tasks.
  • The three clinical rules encode diagnostic knowledge directly into the learned representations without requiring extra manual annotations.
  • Dual prediction at video and frame levels allows the model to use both global sequence context and localized counterfactual contrasts.
  • Ablation results isolate the contribution of the generator, the rule-based representation module, and the dual prediction head.

Where Pith is reading between the lines

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

  • The same generator-plus-rules structure could be tested on other time-varying medical imaging modalities such as ultrasound or endoscopic fluorescence sequences.
  • Generated counterfactuals might serve as visual explanations that clinicians can inspect to verify whether the model attends to the same tissue features they use.
  • If the clinical rules prove sufficient, the approach could lower the amount of labeled video needed for training in new diagnostic domains.

Load-bearing premise

The generated counterfactual videos accurately reflect real tissue changes that would occur under the stated pathological conditions without introducing misleading visual artifacts.

What would settle it

Compare the diffusion-generated sequences against actual longitudinal patient videos that document the same tissue progressing from one pathological state to another; large visual or diagnostic mismatch would falsify the premise that the generator supplies clinically faithful counterfactuals.

Figures

Figures reproduced from arXiv: 2605.26483 by Churan Wang, Jianghua Li, Jianzhe Gao, Li-An Li, Weiyi Zhang, Wenguan Wang, Yixin Zhu, Yizhou Wang.

Figure 1
Figure 1. Figure 1 [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of MEDVCR. Given a medical video sequence, the CG (Sec. 3.1) synthesizes alternative tissue transitions under benign/malignant hypotheses. The CRL module (Sec. 3.2) encodes factual and counterfactual frames, enforcing clinical rules for temporal consistency, pathological separability, and counterfactual alignment. The DDP strategy (Sec. 3.3) then integrates temporal context with frame-level counte… view at source ↗
Figure 3
Figure 3. Figure 3: Visualization of CG (Sec. 3.1) across two medical scenarios. (a) Colposcopy: synthesizing reagent-induced tissue transitions across examination stages. (b) Colonoscopy: modeling temporal polyp emergence. Row 1: benign case; Rows 2&3: malignant cases [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
read the original abstract

Medical video diagnosis involves inferring clinical decisions from dynamic tissue responses throughout examination processes. Existing methods rely on an end-to-end learning paradigm that i) focuses on appearance rather than pathology, ii) lacks clinical priors, and iii) reasons solely from observations without counterfactual comparison. This work introduces MedVCR, a counterfactual reasoning framework that mimics clinical diagnostic thinking. MedVCR comprises three components: a Counterfactual Generator that synthesizes tissue evolution under specified pathological states via a diffusion-based manner; a Counterfactual Representation Learning module that encodes diagnostic knowledge through clinical rules (i.e., temporal consistency, pathological separability, and counterfactual alignment); and a Dual Diagnostic Prediction strategy that integrates video-level assessment with frame-level counterfactual analysis. MedVCR is evaluated under both fully supervised (e.g., colposcopy) and weakly supervised (e.g., colonoscopy) video diagnosis settings, yielding 2.6%-10.2% performance gains compared with leading baselines. Comprehensive ablation studies further validate the effectiveness of each component. The code will be released.

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. The paper introduces MedVCR, a counterfactual reasoning framework for medical video diagnosis. It consists of a diffusion-based Counterfactual Generator that synthesizes tissue evolution under specified pathological states, a Counterfactual Representation Learning module that encodes diagnostic knowledge via three clinical rules (temporal consistency, pathological separability, counterfactual alignment), and a Dual Diagnostic Prediction strategy integrating video-level and frame-level analysis. The framework is evaluated on fully supervised (colposcopy) and weakly supervised (colonoscopy) video diagnosis tasks, reporting 2.6%-10.2% gains over leading baselines, with ablation studies validating each component.

Significance. If the counterfactuals prove clinically faithful, the work could meaningfully advance video-based diagnosis by shifting from appearance-driven end-to-end learning to pathology-aware counterfactual comparison with explicit clinical priors. The diffusion generator and rule-based representation learning are technically interesting contributions, and the dual supervised/weakly-supervised evaluation plus code-release commitment are positive.

major comments (3)
  1. [Methods (Counterfactual Generator) and Experiments] The central claim that MedVCR performs 'clinically-grounded' reasoning rests on the diffusion Counterfactual Generator producing pathologically accurate tissue evolution. No section reports independent clinical validation (blinded expert ratings, Dice overlap with real follow-up frames, or artifact quantification), so downstream task gains alone cannot confirm the generator avoids non-diagnostic cues that the representation learner might exploit.
  2. [§4 (Ablation Studies)] §4 (Ablation Studies): The ablations demonstrate that removing each clinical rule or the generator reduces accuracy, but provide no test of whether the three rules actually enforce pathological separability or merely improve internal consistency; this leaves open the possibility that gains arise from spurious correlations rather than the claimed clinical grounding.
  3. [Table 1 / main results] Table 1 / main results: The 2.6%-10.2% gains are presented without statistical significance tests, confidence intervals, or multi-run variance, which is load-bearing for interpreting whether the improvements are reliable under both fully and weakly supervised regimes.
minor comments (2)
  1. [Abstract] The abstract states performance gains but supplies no dataset sizes, exact metrics, or baseline names; these details should appear in the abstract or be cross-referenced to a table.
  2. [§3.2] Notation for the three clinical rules is introduced descriptively; adding explicit loss equations or pseudocode would improve reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback on MedVCR. The comments highlight opportunities to strengthen evidence for clinical grounding and statistical rigor. We address each point below with proposed revisions where feasible.

read point-by-point responses

  1. Referee: [Methods (Counterfactual Generator) and Experiments] The central claim that MedVCR performs 'clinically-grounded' reasoning rests on the diffusion Counterfactual Generator producing pathologically accurate tissue evolution. No section reports independent clinical validation (blinded expert ratings, Dice overlap with real follow-up frames, or artifact quantification), so downstream task gains alone cannot confirm the generator avoids non-diagnostic cues that the representation learner might exploit.

    Authors: We acknowledge that direct clinical validation (e.g., blinded ratings or Dice scores against follow-up frames) is absent and would provide stronger support for pathological fidelity. The current evidence relies on downstream gains and ablations. In revision we will add an explicit Limitations subsection discussing this gap and outlining plans for future expert studies; we will also expand qualitative counterfactual visualizations for reader assessment. revision: partial

  2. Referee: [§4 (Ablation Studies)] §4 (Ablation Studies): The ablations demonstrate that removing each clinical rule or the generator reduces accuracy, but provide no test of whether the three rules actually enforce pathological separability or merely improve internal consistency; this leaves open the possibility that gains arise from spurious correlations rather than the claimed clinical grounding.

    Authors: The rules are explicitly derived from clinical diagnostic principles to target separability. Ablations already show performance impact, but to directly test separability we will augment §4 with quantitative metrics (e.g., inter-class embedding distances before/after each rule) demonstrating enhanced pathological distinctions. This will be included in the revised manuscript. revision: yes

  3. Referee: [Table 1 / main results] Table 1 / main results: The 2.6%-10.2% gains are presented without statistical significance tests, confidence intervals, or multi-run variance, which is load-bearing for interpreting whether the improvements are reliable under both fully and weakly supervised regimes.

    Authors: We agree statistical reporting is essential. In the revision we will rerun all experiments across multiple random seeds, report means ± standard deviation, and add paired statistical tests (e.g., t-tests with p-values) to Table 1 for both supervised and weakly-supervised settings. revision: yes

Circularity Check

0 steps flagged

No circularity: framework components defined independently of outputs

full rationale

The provided abstract and description introduce MedVCR as a composite framework with a diffusion generator, rule-based representation learning (temporal consistency, pathological separability, counterfactual alignment), and dual prediction. No equations, self-definitions, fitted parameters renamed as predictions, or load-bearing self-citations are visible that would make any claimed result equivalent to its inputs by construction. The performance gains are presented as empirical outcomes from external evaluations rather than tautological derivations. This is the common case of a self-contained proposal without detectable circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available, so no concrete free parameters, axioms, or invented entities can be extracted. The diffusion generator and clinical-rule modules almost certainly contain many unspecified hyperparameters and modeling assumptions about what constitutes valid counterfactual tissue evolution.

pith-pipeline@v0.9.1-grok · 5728 in / 1175 out tokens · 52357 ms · 2026-06-29T18:33:12.302192+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

90 extracted references

  1. [1]

    Vivit: A video vi- sion transformer

    Anurag Arnab, Mostafa Dehghani, Georg Heigold, Chen Sun, Mario Lu ˇci´c, and Cordelia Schmid. Vivit: A video vi- sion transformer. InProceedings of International Conference on Computer Vision (ICCV), 2021. 2

    2021

  2. [2]

    Dig-in: Diffusion guidance for investigating networks- uncovering classifier differences neuron visualisations and visual counterfactual explanations

    Maximilian Augustin, Yannic Neuhaus, and Matthias Hein. Dig-in: Diffusion guidance for investigating networks- uncovering classifier differences neuron visualisations and visual counterfactual explanations. InProceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2024. 2

    2024

  3. [3]

    Is space-time attention all you need for video understanding? In Proceedings of International Conference on Machine Learn- ing (ICML), 2021

    Gedas Bertasius, Heng Wang, and Lorenzo Torresani. Is space-time attention all you need for video understanding? In Proceedings of International Conference on Machine Learn- ing (ICML), 2021. 6

    2021

  4. [4]

    Hyperkvasir, a comprehensive multi-class im- age and video dataset for gastrointestinal endoscopy.Scien- tific data, 7(1):283, 2020

    Hanna Borgli, Vajira Thambawita, Pia H Smedsrud, Steven Hicks, Debesh Jha, Sigrun L Eskeland, Kristin Ranheim Randel, Konstantin Pogorelov, Mathias Lux, Duc Tien Dang Nguyen, et al. Hyperkvasir, a comprehensive multi-class im- age and video dataset for gastrointestinal endoscopy.Scien- tific data, 7(1):283, 2020. 7

    2020

  5. [5]

    Weakly supervised causal representation learning

    Johann Brehmer, Pim De Haan, Phillip Lippe, and Taco S Cohen. Weakly supervised causal representation learning. In Proceedings of Advances in Neural Information Processing Systems (NeurIPS), 2022. 2

    2022

  6. [6]

    Global elimination of cer- vical cancer as a public health problem.The Lancet Oncol- ogy, 20(3):319–321, 2019

    Marc Brisson and M ´elanie Drolet. Global elimination of cer- vical cancer as a public health problem.The Lancet Oncol- ogy, 20(3):319–321, 2019. 6

    2019

  7. [7]

    Robotic wireless capsule endoscopy: recent advances and upcoming technologies.Nature Communications, 15(1): 4597, 2024

    Qing Cao, Runyi Deng, Yue Pan, Ruijie Liu, Yicheng Chen, Guofang Gong, Jun Zou, Huayong Yang, and Dong Han. Robotic wireless capsule endoscopy: recent advances and upcoming technologies.Nature Communications, 15(1): 4597, 2024. 6

    2024

  8. [8]

    Quo vadis, action recognition? a new model and the kinetics dataset

    Joao Carreira and Andrew Zisserman. Quo vadis, action recognition? a new model and the kinetics dataset. InPro- ceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2017. 2, 4, 5, 7, 8

    2017

  9. [9]

    Causality matters in medical imaging.Nature Communications, 11(1): 3673, 2020

    Daniel C Castro, Ian Walker, and Ben Glocker. Causality matters in medical imaging.Nature Communications, 11(1): 3673, 2020. 2

    2020

  10. [10]

    Stddnet: Harness- ing mamba for video polyp segmentation via spatial-aligned temporal modeling and discriminative dynamic representa- tion learning

    Guilian Chen, Huisi Wu, and Jing Qin. Stddnet: Harness- ing mamba for video polyp segmentation via spatial-aligned temporal modeling and discriminative dynamic representa- tion learning. InProceedings of International Conference on Computer Vision (ICCV), 2025. 2, 6

    2025

  11. [11]

    Ultrasound image-to-video synthesis via latent dynamic diffusion mod- els

    Tingxiu Chen, Yilei Shi, Zixuan Zheng, Bingcong Yan, Jingliang Hu, Xiao Xiang Zhu, and Lichao Mou. Ultrasound image-to-video synthesis via latent dynamic diffusion mod- els. InInternational Conference on Medical Image Comput- ing and Computer Assisted Intervention (MICCAI), 2024. 1

    2024

  12. [12]

    Surgical video captioning with mutual-modal concept alignment

    Zhen Chen, Qingyu Guo, Leo KT Yeung, Danny TM Chan, Zhen Lei, Hongbin Liu, and Jinqiao Wang. Surgical video captioning with mutual-modal concept alignment. InIn- ternational Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), 2023. 2

    2023

  13. [13]

    Stargan: Unified genera- tive adversarial networks for multi-domain image-to-image translation

    Yunjey Choi, Minje Choi, Munyoung Kim, Jung-Woo Ha, Sunghun Kim, and Jaegul Choo. Stargan: Unified genera- tive adversarial networks for multi-domain image-to-image translation. InProceedings of Conference on Computer Vi- sion and Pattern Recognition (CVPR), 2018. 2

    2018

  14. [14]

    Polypsegtrack: uni- fied foundation model for colonoscopy video analysis

    Anwesa Choudhuri, Zhongpai Gao, Meng Zheng, Benjamin Planche, Terrence Chen, and Ziyan Wu. Polypsegtrack: uni- fied foundation model for colonoscopy video analysis. In International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), 2025. 1, 2, 6

    2025

  15. [15]

    Opera: Attention-regularized transformers for surgical phase recog- nition

    Tobias Czempiel, Magdalini Paschali, Daniel Ostler, Seong Tae Kim, Benjamin Busam, and Nassir Navab. Opera: Attention-regularized transformers for surgical phase recog- nition. InInternational Conference on Medical Image Com- puting and Computer Assisted Intervention (MICCAI), 2021. 1, 2

    2021

  16. [16]

    Memsam: Taming segment anything model for echocardiog- raphy video segmentation

    Xiaolong Deng, Huisi Wu, Runhao Zeng, and Jing Qin. Memsam: Taming segment anything model for echocardiog- raphy video segmentation. InProceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2024. 2

    2024

  17. [17]

    Videocutmix: Temporal segmentation of surgical videos in scarce data scenarios

    Rohan Raju Dhanakshirur, Mrinal Tyagi, Britty Baby, Ashish Suri, Prem Kalra, and Chetan Arora. Videocutmix: Temporal segmentation of surgical videos in scarce data scenarios. In International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), 2024. 6

    2024

  18. [18]

    The automated learning of deep features for breast mass clas- sification from mammograms

    Neeraj Dhungel, Gustavo Carneiro, and Andrew P Bradley. The automated learning of deep features for breast mass clas- sification from mammograms. InInternational Conference on Medical Image Computing and Computer Assisted Inter- vention (MICCAI), 2016. A1

    2016

  19. [19]

    Exploring segment- level semantics for online phase recognition from surgical videos.IEEE Transactions on Medical Imaging (TMI), 41 (11):3309–3319, 2022

    Xinpeng Ding and Xiaomeng Li. Exploring segment- level semantics for online phase recognition from surgical videos.IEEE Transactions on Medical Imaging (TMI), 41 (11):3309–3319, 2022. 2

    2022

  20. [20]

    Pranet: Parallel reverse attention network for polyp segmentation

    Deng-Ping Fan, Ge-Peng Ji, Tao Zhou, Geng Chen, Huazhu Fu, Jianbing Shen, and Ling Shao. Pranet: Parallel reverse attention network for polyp segmentation. InInternational Conference on Medical Image Computing and Computer As- sisted Intervention (MICCAI), 2020. 6

    2020

  21. [21]

    Slowfast networks for video recognition

    Christoph Feichtenhofer, Haoqi Fan, Jitendra Malik, and Kaiming He. Slowfast networks for video recognition. In Proceedings of International Conference on Computer Vi- sion (ICCV), 2019. 2

    2019

  22. [22]

    Mist: Multiple instance self-training framework for video anomaly detection

    Jia-Chang Feng, Fa-Ting Hong, and Wei-Shi Zheng. Mist: Multiple instance self-training framework for video anomaly detection. InProceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2021. 7

    2021

  23. [23]

    Trans-svnet: Accurate phase recogni- tion from surgical videos via hybrid embedding aggregation transformer

    Xiaojie Gao, Yueming Jin, Yonghao Long, Qi Dou, and Pheng-Ann Heng. Trans-svnet: Accurate phase recogni- tion from surgical videos via hybrid embedding aggregation transformer. InInternational Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), pages 593–603, 2021. 1

    2021

  24. [24]

    Ganalyze: Toward visual definitions of cognitive im- age properties

    Lore Goetschalckx, Alex Andonian, Aude Oliva, and Phillip Isola. Ganalyze: Toward visual definitions of cognitive im- age properties. InProceedings of International Conference on Computer Vision (ICCV), 2019. 2

    2019

  25. [25]

    Mmsum- mary: Multimodal summary generation for fetal ultrasound video

    Xiaoqing Guo, Qianhui Men, and J Alison Noble. Mmsum- mary: Multimodal summary generation for fetal ultrasound video. InInternational Conference on Medical Image Com- puting and Computer Assisted Intervention (MICCAI), 2024. 1

    2024

  26. [26]

    Benefits and harms of cervical screening, triage and treatment strate- gies in women living with hiv.Nature Medicine, 29(12): 3059–3066, 2023

    Michaela T Hall, Kate T Simms, John M Murray, Adam Keane, Diep TN Nguyen, Michael Caruana, Gigi Lui, He- len Kelly, Linda O Eckert, Nancy Santesso, et al. Benefits and harms of cervical screening, triage and treatment strate- gies in women living with hiv.Nature Medicine, 29(12): 3059–3066, 2023. 1, A1

    2023

  27. [27]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. InProceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2016. 8, A1

    2016

  28. [28]

    Motionbench: Benchmarking and improving fine- grained video motion understanding for vision language models

    Wenyi Hong, Yean Cheng, Zhuoyi Yang, Weihan Wang, Lefan Wang, Xiaotao Gu, Shiyu Huang, Yuxiao Dong, and Jie Tang. Motionbench: Benchmarking and improving fine- grained video motion understanding for vision language models. InProceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2025. 1

    2025

  29. [29]

    Nurvid: A large expert-level video database for nursing pro- cedure activity understanding

    Ming Hu, Lin Wang, Siyuan Yan, Don Ma, Qingli Ren, Peng Xia, Wei Feng, Peibo Duan, Lie Ju, and Zongyuan Ge. Nurvid: A large expert-level video database for nursing pro- cedure activity understanding. InProceedings of Advances in Neural Information Processing Systems (NeurIPS), 2023. 1

    2023

  30. [30]

    Self- supervised learning for medical image classification: a sys- tematic review and implementation guidelines.NPJ Digital Medicine, 6(1):74, 2023

    Shih-Cheng Huang, Anuj Pareek, Malte Jensen, Matthew P Lungren, Serena Yeung, and Akshay S Chaudhari. Self- supervised learning for medical image classification: a sys- tematic review and implementation guidelines.NPJ Digital Medicine, 6(1):74, 2023. 2

    2023

  31. [31]

    Causal normalizing flows: from theory to practice

    Adri ´an Javaloy, Pablo S ´anchez-Mart´ın, and Isabel Valera. Causal normalizing flows: from theory to practice. InPro- ceedings of Advances in Neural Information Processing Sys- tems (NeurIPS), 2023. 2

    2023

  32. [32]

    Ad- versarial counterfactual visual explanations

    Guillaume Jeanneret, Lo ¨ıc Simon, and Fr´ed´eric Jurie. Ad- versarial counterfactual visual explanations. InProceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2023. 2

    2023

  33. [33]

    Sv-rcnet: workflow recogni- tion from surgical videos using recurrent convolutional net- work.IEEE Transactions on Medical Imaging (TMI), 37(5): 1114–1126, 2017

    Yueming Jin, Qi Dou, Hao Chen, Lequan Yu, Jing Qin, Chi- Wing Fu, and Pheng-Ann Heng. Sv-rcnet: workflow recogni- tion from surgical videos using recurrent convolutional net- work.IEEE Transactions on Medical Imaging (TMI), 37(5): 1114–1126, 2017. 2

    2017

  34. [34]

    Multi-task recurrent con- volutional network with correlation loss for surgical video analysis.Medical Image Analysis (MIA), 59:101572, 2020

    Yueming Jin, Huaxia Li, Qi Dou, Hao Chen, Jing Qin, Chi- Wing Fu, and Pheng-Ann Heng. Multi-task recurrent con- volutional network with correlation loss for surgical video analysis.Medical Image Analysis (MIA), 59:101572, 2020. 2

    2020

  35. [35]

    Smita Joshi, Richard Muwonge, Ramesh Bhosale, Pritam Chaudhari, Vinay Kulkarni, Mahesh Mandolkar, Kedar De- odhar, Seema Kand, Nikhil Phadke, Shobini Rajan, et al. A randomised controlled non-inferiority trial to compare the ef- ficacy of ‘hpv screen, triage and treat’with ‘hpv screen and treat’approach for cervical cancer prevention among women living wi...

    2025

  36. [36]

    Grounding counterfactual explanation of image classifiers to textual concept space

    Siwon Kim, Jinoh Oh, Sungjin Lee, Seunghak Yu, Jaeyoung Do, and Tara Taghavi. Grounding counterfactual explanation of image classifiers to textual concept space. InProceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2023. 2

    2023

  37. [37]

    Semi-supervised echocardiography video segmenta- tion via adaptive spatio-temporal tensor semantic awareness and memory flow.IEEE Transactions on Medical Imaging (TMI), 2025

    Xiaodi Li, Chen Cui, Siyuan Shi, Hongwen Fei, and Yue Hu. Semi-supervised echocardiography video segmenta- tion via adaptive spatio-temporal tensor semantic awareness and memory flow.IEEE Transactions on Medical Imaging (TMI), 2025. 2

    2025

  38. [38]

    Causal composition diffusion model for closed-loop traffic generation

    Haohong Lin, Xin Huang, Tung Phan, David Hayden, Huan Zhang, Ding Zhao, Siddhartha Srinivasa, Eric Wolff, and Hongge Chen. Causal composition diffusion model for closed-loop traffic generation. InProceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2025. 2

    2025

  39. [39]

    Skit: a fast key information video transformer for online surgical phase recognition

    Yang Liu, Jiayu Huo, Jingjing Peng, Rachel Sparks, Prokar Dasgupta, Alejandro Granados, and Sebastien Ourselin. Skit: a fast key information video transformer for online surgical phase recognition. InProceedings of International Conference on Computer Vision (ICCV), 2023. 1

    2023

  40. [40]

    Video swin transformer

    Ze Liu, Jia Ning, Yue Cao, Yixuan Wei, Zheng Zhang, Stephen Lin, and Han Hu. Video swin transformer. In Proceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2022. 6

    2022

  41. [41]

    Fadmb: Fully attention-based dual memory bank network for weakly supervised video anomaly detection

    Zhiming Luo, Shuheng Huang, Kun Yang, Jianzhe Gao, and Shaozi Li. Fadmb: Fully attention-based dual memory bank network for weakly supervised video anomaly detection. In Pattern Recognition (PR), 2025. 7

    2025

  42. [42]

    Ldpolypvideo benchmark: a large-scale colonoscopy video dataset of diverse polyps

    Yiting Ma, Xuejin Chen, Kai Cheng, Yang Li, and Bin Sun. Ldpolypvideo benchmark: a large-scale colonoscopy video dataset of diverse polyps. InInternational Conference on Medical Image Computing and Computer Assisted Interven- tion (MICCAI), 2021. 1, 7

    2021

  43. [43]

    Diffpo: A causal diffusion model for learning distributions of potential outcomes

    Yuchen Ma, Valentyn Melnychuk, Jonas Schweisthal, and Stefan Feuerriegel. Diffpo: A causal diffusion model for learning distributions of potential outcomes. InProceed- ings of Advances in Neural Information Processing Systems (NeurIPS), 2024. 2

    2024

  44. [44]

    Self-supervised masked convolu- tional transformer block for anomaly detection.Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 46 (1):525–542, 2023

    Neelu Madan, Nicolae-C ˘at˘alin Ristea, Radu Tudor Ionescu, Kamal Nasrollahi, Fahad Shahbaz Khan, Thomas B Moes- lund, and Mubarak Shah. Self-supervised masked convolu- tional transformer block for anomaly detection.Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 46 (1):525–542, 2023. 2

    2023

  45. [45]

    Benchmarking counterfactual image generation

    Thomas Melistas, Nikos Spyrou, Nefeli Gkouti, Pedro Sanchez, Athanasios Vlontzos, Yannis Panagakis, Giorgos Papanastasiou, and Sotirios A Tsaftaris. Benchmarking counterfactual image generation. InProceedings of Ad- vances in Neural Information Processing Systems (NeurIPS),

  46. [46]

    Mcat: Visual query-based localization of stan- dard anatomical clips in fetal ultrasound videos using multi- tier class-aware token transformer

    Divyanshu Mishra, Pramit Saha, He Zhao, Netzahualcoyotl Hernandez-Cruz, Olga Patey, Aris Papageorghiou, and J Al- ison Noble. Mcat: Visual query-based localization of stan- dard anatomical clips in fetal ultrasound videos using multi- tier class-aware token transformer. InProceedings of AAAI Conference on Artificial Intelligence (AAAI), 2025. 1

    2025

  47. [47]

    Tier-loc: Visual query-based video clip lo- calization in fetal ultrasound videos with a multi-tier trans- former.Medical Image Analysis (MIA), page 103611, 2025

    Divyanshu Mishra, Pramit Saha, He Zhao, Netzahualcoy- otl Hernandez-Cruz, Olga Patey, Aris T Papageorghiou, and J Alison Noble. Tier-loc: Visual query-based video clip lo- calization in fetal ultrasound videos with a multi-tier trans- former.Medical Image Analysis (MIA), page 103611, 2025. 2

    2025

  48. [48]

    Inbreast: toward a full-field digital mammographic database.Aca- demic radiology, 19(2):236–248, 2012

    In ˆes C Moreira, Igor Amaral, Inˆes Domingues, Ant´onio Car- doso, Maria Joao Cardoso, and Jaime S Cardoso. Inbreast: toward a full-field digital mammographic database.Aca- demic radiology, 19(2):236–248, 2012. A1

    2012

  49. [49]

    Maxime Oquab, Timoth ´ee Darcet, Th´eo Moutakanni, Huy V . V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel HAZIZA, Francisco Massa, Alaaeldin El-Nouby, Mido Assran, Nicolas Ballas, Wojciech Galuba, Russell Howes, Po-Yao Huang, Shang-Wen Li, Ishan Misra, Michael Rabbat, Vasu Sharma, Gabriel Synnaeve, Hu Xu, Herve Je- gou, Julien Mairal, Patr...

    2024

  50. [50]

    Mambasci: Efficient mamba-unet for quad-bayer patterned video snapshot compressive imag- ing

    Zhenghao Pan, Haijin Zeng, Jiezhang Cao, Yongyong Chen, Kai Zhang, and Yong Xu. Mambasci: Efficient mamba-unet for quad-bayer patterned video snapshot compressive imag- ing. InProceedings of Advances in Neural Information Pro- cessing Systems (NeurIPS), 2024. 1

    2024

  51. [51]

    Deep structural causal models for tractable counterfactual in- ference

    Nick Pawlowski, Daniel Coelho de Castro, and Ben Glocker. Deep structural causal models for tractable counterfactual in- ference. InProceedings of Advances in Neural Information Processing Systems (NeurIPS), 2020. 2

    2020

  52. [52]

    Learn- ing transferable visual models from natural language super- vision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learn- ing transferable visual models from natural language super- vision. InProceedings of International Conference on Ma- chine Learning (ICML), 2021. 8

    2021

  53. [53]

    Im- proving the accuracy of medical diagnosis with causal ma- chine learning.Nature communications, 11(1):3923, 2020

    Jonathan G Richens, Ciar ´an M Lee, and Saurabh Johri. Im- proving the accuracy of medical diagnosis with causal ma- chine learning.Nature communications, 11(1):3923, 2020. 2

    2020

  54. [54]

    On the pitfalls of batch normalization for end-to-end video learning: a study on surgical workflow analysis.Medical Image Anal- ysis (MIA), 94:103126, 2024

    Dominik Rivoir, Isabel Funke, and Stefanie Speidel. On the pitfalls of batch normalization for end-to-end video learning: a study on surgical workflow analysis.Medical Image Anal- ysis (MIA), 94:103126, 2024. 2

    2024

  55. [55]

    Cer- vical cancer screening using dna methylation triage in a real- world population.Nature medicine, 30(8):2251–2257, 2024

    Lena Schreiberhuber, James E Barrett, Jiangrong Wang, Elisa Redl, Chiara Herzog, Charlotte D Vavourakis, Karin Sundstr¨om, Joakim Dillner, and Martin Widschwendter. Cer- vical cancer screening using dna methylation triage in a real- world population.Nature medicine, 30(8):2251–2257, 2024. 1, A1

    2024

  56. [56]

    Image quality assessment: Investigating causal perceptual effects with abductive coun- terfactual inference

    Wenhao Shen, Mingliang Zhou, Yu Chen, Xuekai Wei, Yong Feng, Huayan Pu, and Weijia Jia. Image quality assessment: Investigating causal perceptual effects with abductive coun- terfactual inference. InProceedings of Conference on Com- puter Vision and Pattern Recognition (CVPR), 2025. 2

    2025

  57. [57]

    Deep unsupervised learning using nonequilibrium thermodynamics

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. InInternational confer- ence on machine learning, pages 2256–2265. pmlr, 2015. 3

    2015

  58. [58]

    Doubly abductive coun- terfactual inference for text-based image editing

    Xue Song, Jiequan Cui, Hanwang Zhang, Jingjing Chen, Richang Hong, and Yu-Gang Jiang. Doubly abductive coun- terfactual inference for text-based image editing. InProceed- ings of Conference on Computer Vision and Pattern Recog- nition (CVPR), 2024. 2

    2024

  59. [59]

    Using computer vision to improve endoscopic disease quantification in therapeutic clinical trials of ulcerative colitis.Gastroenterology, 166(1): 155–167, 2024

    Ryan W Stidham, Lingrui Cai, Shuyang Cheng, Flora Rajaei, Tadd Hiatt, Emily Wittrup, Michael D Rice, Shrinivas Bishu, Jan Wehkamp, Weiwei Schultz, et al. Using computer vision to improve endoscopic disease quantification in therapeutic clinical trials of ulcerative colitis.Gastroenterology, 166(1): 155–167, 2024. 1

    2024

  60. [60]

    Last: Latent space-constrained transformers for automatic surgical phase recognition and tool presence detection.IEEE Transactions on Medical Imaging (TMI), 42(11):3256–3268, 2023

    Rong Tao, Xiaoyang Zou, and Guoyan Zheng. Last: Latent space-constrained transformers for automatic surgical phase recognition and tool presence detection.IEEE Transactions on Medical Imaging (TMI), 42(11):3256–3268, 2023. 2

    2023

  61. [61]

    Weakly-supervised video anomaly detection with robust temporal feature mag- nitude learning

    Yu Tian, Guansong Pang, Yuanhong Chen, Rajvinder Singh, Johan W Verjans, and Gustavo Carneiro. Weakly-supervised video anomaly detection with robust temporal feature mag- nitude learning. InProceedings of International Conference on Computer Vision (ICCV), 2021. 7

    2021

  62. [62]

    Contrastive transformer-based multiple instance learning for weakly supervised polyp frame detection

    Yu Tian, Guansong Pang, Fengbei Liu, Yuyuan Liu, Chong Wang, Yuanhong Chen, Johan Verjans, and Gustavo Carneiro. Contrastive transformer-based multiple instance learning for weakly supervised polyp frame detection. In International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), 2022. 2, 7

    2022

  63. [63]

    A closer look at spatiotem- poral convolutions for action recognition

    Du Tran, Heng Wang, Lorenzo Torresani, Jamie Ray, Yann LeCun, and Manohar Paluri. A closer look at spatiotem- poral convolutions for action recognition. InProceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2018. 8

    2018

  64. [64]

    En- donet: a deep architecture for recognition tasks on laparo- scopic videos.IEEE Transactions on Medical Imaging (TMI), 36(1):86–97, 2016

    Andru P Twinanda, Sherif Shehata, Didier Mutter, Jacques Marescaux, Michel De Mathelin, and Nicolas Padoy. En- donet: a deep architecture for recognition tasks on laparo- scopic videos.IEEE Transactions on Medical Imaging (TMI), 36(1):86–97, 2016. 2

    2016

  65. [65]

    Rsdnet: Learning to predict remaining surgery duration from laparoscopic videos without manual annotations.IEEE Transactions on Medical Imaging (TMI), 38(4):1069–1078, 2018

    Andru Putra Twinanda, Gaurav Yengera, Didier Mutter, Jacques Marescaux, and Nicolas Padoy. Rsdnet: Learning to predict remaining surgery duration from laparoscopic videos without manual annotations.IEEE Transactions on Medical Imaging (TMI), 38(4):1069–1078, 2018. 2

    2018

  66. [66]

    Com- prehensive echocardiogram evaluation with view primed vi- sion language ai.Nature, 2025

    Milos Vukadinovic, Xiu Tang, Neal Yuan, Paul Cheng, De- biao Li, Susan Cheng, Bryan He, and David Ouyang. Com- prehensive echocardiogram evaluation with view primed vi- sion language ai.Nature, 2025. 6, 7

    2025

  67. [67]

    Bilateral asymmetry guided counterfactual generating network for mammogram classification.Transactions on Image Processing (TIP), 30: 7980–7994, 2021

    Churan Wang, Jing Li, Fandong Zhang, Xinwei Sun, Hao Dong, Yizhou Yu, and Yizhou Wang. Bilateral asymmetry guided counterfactual generating network for mammogram classification.Transactions on Image Processing (TIP), 30: 7980–7994, 2021. 2, A1

    2021

  68. [68]

    Videomae v2: Scaling video masked autoencoders with dual masking

    Limin Wang, Bingkun Huang, Zhiyu Zhao, Zhan Tong, Yi- nan He, Yi Wang, Yali Wang, and Yu Qiao. Videomae v2: Scaling video masked autoencoders with dual masking. In Proceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2023. 6

    2023

  69. [69]

    Omnidrive: A holistic vision-language dataset for au- tonomous driving with counterfactual reasoning

    Shihao Wang, Zhiding Yu, Xiaohui Jiang, Shiyi Lan, Min Shi, Nadine Chang, Jan Kautz, Ying Li, and Jose M Al- varez. Omnidrive: A holistic vision-language dataset for au- tonomous driving with counterfactual reasoning. InProceed- ings of Conference on Computer Vision and Pattern Recog- nition (CVPR), 2025. 2

    2025

  70. [70]

    Foun- dation model for endoscopy video analysis via large-scale self-supervised pre-train

    Zhao Wang, Chang Liu, Shaoting Zhang, and Qi Dou. Foun- dation model for endoscopy video analysis via large-scale self-supervised pre-train. InInternational Conference on Medical Image Computing and Computer Assisted Interven- tion (MICCAI), 2023. 2, 7

    2023

  71. [71]

    Deep neural networks im- prove radiologists’ performance in breast cancer screening

    Nan Wu, Jason Phang, Jungkyu Park, Yiqiu Shen, Zhe Huang, Masha Zorin, Stanisław Jastrzebski, Thibault F´evry, Joe Katsnelson, Eric Kim, et al. Deep neural networks im- prove radiologists’ performance in breast cancer screening. IEEE Transactions on Medical Imaging (TMI), 39(4):1184– 1194, 2019. A1

    2019

  72. [72]

    Masked images are counterfactual samples for robust fine- tuning

    Yao Xiao, Ziyi Tang, Pengxu Wei, Cong Liu, and Liang Lin. Masked images are counterfactual samples for robust fine- tuning. InProceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2023. 2

    2023

  73. [73]

    Rethinking spatiotemporal feature learn- ing: Speed-accuracy trade-offs in video classification

    Saining Xie, Chen Sun, Jonathan Huang, Zhuowen Tu, and Kevin Murphy. Rethinking spatiotemporal feature learn- ing: Speed-accuracy trade-offs in video classification. In Proceedings of European Conference on Computer Vision (ECCV), 2018. 8

    2018

  74. [74]

    A survey on incorporating do- main knowledge into deep learning for medical image anal- ysis.Medical Image Analysis (MIA), 69:101985, 2021

    Xiaozheng Xie, Jianwei Niu, Xuefeng Liu, Zhengsu Chen, Shaojie Tang, and Shui Yu. A survey on incorporating do- main knowledge into deep learning for medical image anal- ysis.Medical Image Analysis (MIA), 69:101985, 2021. 2

    2021

  75. [75]

    A whole-slide foundation model for digital pathology from real-world data.Nature, 630(8015):181–188, 2024

    Hanwen Xu, Naoto Usuyama, Jaspreet Bagga, Sheng Zhang, Rajesh Rao, Tristan Naumann, Cliff Wong, Zelalem Gero, Javier Gonz ´alez, Yu Gu, et al. A whole-slide foundation model for digital pathology from real-world data.Nature, 630(8015):181–188, 2024. 2

    2024

  76. [76]

    Deep learning enabled liquid-based cytology model for cervical precancer and cancer detection

    Peng Xue, Le Dang, Ling-Hua Kong, Hong-Ping Tang, Hai- Miao Xu, Hai-Yan Weng, Zhe Wang, Rong-Gan Wei, Lian Xu, Hong-Xia Li, et al. Deep learning enabled liquid-based cytology model for cervical precancer and cancer detection. Nature Communications, 16(1):3506, 2025. 6, A1

    2025

  77. [77]

    Causalvae: Structured causal disentanglement in variational autoencoder

    Mengyue Yang, Furui Liu, Zhitang Chen, Xinwei Shen, Jianye Hao, and Jun Wang. Causalvae: Structured causal disentanglement in variational autoencoder. InProceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2021. 2

    2021

  78. [78]

    Causalvae: Disentangled rep- resentation learning via neural structural causal models

    Mengyue Yang, Furui Liu, Zhitang Chen, Xinwei Shen, Jianye Hao, and Jun Wang. Causalvae: Disentangled rep- resentation learning via neural structural causal models. In Proceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2021. 2

    2021

  79. [79]

    Surgformer: Surgical transformer with hierarchical tempo- ral attention for surgical phase recognition

    Shu Yang, Luyang Luo, Qiong Wang, and Hao Chen. Surgformer: Surgical transformer with hierarchical tempo- ral attention for surgical phase recognition. InInternational Conference on Medical Image Computing and Computer As- sisted Intervention (MICCAI), 2024. 2, 6

    2024

  80. [80]

    Vivim: A video vision mamba for medical video segmentation.IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), 2025

    Yijun Yang, Zhaohu Xing, Lequan Yu, Chunwang Huang, Huazhu Fu, and Lei Zhu. Vivim: A video vision mamba for medical video segmentation.IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), 2025. 1, 2

    2025

Showing first 80 references.