pith. sign in

arxiv: 2605.21852 · v1 · pith:NL3ZLD77new · submitted 2026-05-21 · 💻 cs.CV

Seizure-Semiology-Suite (S3): A Clinically Multimodal Dataset, Benchmark, and Models for Seizure Semiology Understanding

Lina Zhang , Tonmoy Monsoor , Peizheng Li , Jiarui Cui , Xinyi Peng , Chong Han , Prateik Sinha , Siyuan Dai
show 18 more authors
Jessica Nichole Pasqua Colin M McCrimmon Weiting Liu Hailey Marie Miranda Bing Hu Xiangting Wu Tengyou Xu Chunhan Li Jiaye Tian Jiarui Tang Detao Ma Lingye Kong Junnan Lyu Jungang Li Yan Zan Junhua Huang Rajarshi Mazumder Vwani Roychowdhury
This is my paper

Pith reviewed 2026-05-22 08:08 UTC · model grok-4.3

classification 💻 cs.CV
keywords seizure semiologymultimodal large language modelsmedical video understandingepilepsy classificationclinical datasetneuro-symbolic frameworkfine-tuningreport quality evaluation
0
0 comments X

The pith

A dataset of 438 annotated seizure videos and a seven-task benchmark enable multimodal models to reach 0.96 F1 on epileptic versus non-epileptic classification after seizure-specific fine-tuning.

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

The paper introduces Seizure-Semiology-Suite, a collection of 438 clinical seizure videos annotated with more than 35,000 labels across 20 ILAE-defined semiological features. It pairs this resource with a hierarchical seven-task benchmark that tests models on visual perception, temporal sequencing, report generation, and diagnosis, plus the Seizure-RQI metric to score how clinically faithful generated reports are. Baselines on eleven open-weight multimodal models expose consistent failures in laterality judgment, symptom timing, and accurate reporting. Seizure-specific fine-tuning lifts results across tasks, and a two-stage neuro-symbolic pipeline reaches an F1 of 0.96 on distinguishing epileptic from non-epileptic events. A sympathetic reader would care because reliable video understanding could support faster, more consistent interpretation of seizures in epilepsy monitoring units where visual analysis remains central to diagnosis.

Core claim

The central claim is that a clinically grounded dataset of 438 videos with dense annotations on 20 semiological features, combined with a seven-task benchmark and the Seizure-RQI metric, reveals systematic weaknesses in current multimodal large language models; seizure-specific fine-tuning substantially improves performance across the tasks, while a two-stage neuro-symbolic framework achieves an F1 score of 0.96 on epileptic versus non-epileptic seizure classification.

What carries the argument

The seven-task hierarchical benchmark that moves from low-level visual perception through temporal sequencing and narrative report generation to final seizure diagnosis, evaluated with the Seizure-RQI for clinical faithfulness.

If this is right

  • General multimodal models exhibit repeatable failures in laterality reasoning, temporal localization, and symptom sequencing on seizure videos.
  • Seizure-specific fine-tuning produces measurable gains on every task in the seven-task hierarchy.
  • The two-stage neuro-symbolic framework delivers an F1 of 0.96 on the binary epileptic versus non-epileptic classification task.
  • The Seizure-RQI metric supplies a structured way to judge whether generated reports match clinical expectations for detail and accuracy.
  • The benchmark can serve as a testbed for developing domain-adapted multimodal systems intended for safety-critical medical video analysis.

Where Pith is reading between the lines

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

  • Similar annotated video collections could be created for other paroxysmal neurological events such as syncope or movement disorders to test cross-domain transfer.
  • If the reported gains hold on prospective clinical recordings, the approach might reduce inter-rater variability in epilepsy monitoring unit reviews.
  • The identified weaknesses in temporal ordering suggest that future video models will need explicit mechanisms for modeling symptom evolution over seconds to minutes.
  • The dataset size and annotation density make it feasible to explore whether smaller, specialized models can match the performance of large general models after targeted training.

Load-bearing premise

The expert annotations on the 438 videos are accurate, consistent, and representative of real-world clinical seizure diversity so the benchmark and Seizure-RQI capture meaningful requirements.

What would settle it

Performance of the fine-tuned models and the neuro-symbolic framework drops sharply when the same tasks are run on a fresh set of seizure videos annotated by an independent group of clinicians using different labeling criteria.

Figures

Figures reproduced from arXiv: 2605.21852 by Bing Hu, Chong Han, Chunhan Li, Colin M McCrimmon, Detao Ma, Hailey Marie Miranda, Jessica Nichole Pasqua, Jiarui Cui, Jiarui Tang, Jiaye Tian, Jungang Li, Junhua Huang, Junnan Lyu, Lina Zhang, Lingye Kong, Peizheng Li, Prateik Sinha, Rajarshi Mazumder, Siyuan Dai, Tengyou Xu, Tonmoy Monsoor, Vwani Roychowdhury, Weiting Liu, Xiangting Wu, Xinyi Peng, Yan Zan.

Figure 1
Figure 1. Figure 1: Seizure Semiology Dataset Information. (The patient depicted in this figure is AI-generated.) trum of semiological features. They first trained the annota￾tors using 76 videos through live annotations, detailed ex￾planations, and real-time question and answer. To quantify reliability, the annotators independently labeled the other 75-video subset. As summarized in [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Seizure Semiology Benchmark. (The patients in this figure are AI-generated.) denkemper & Kotagal, 2005), including risk of injury and sudden unexpected death in epilepsy (SUDEP) (Harden et al., 2017). In contrast, features like pelvic thrusting and asynchronous movements are more indicative of NES (Perez & LaFrance, 2016). This task evaluates whether an MLLM can detect the presence or absence of 20 core se… view at source ↗
Figure 3
Figure 3. Figure 3: Task 1 F1-scores across 20 semiological features [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Clinical Alignment of Evaluation Metrics. Distance (5.15) and highest LCS ratio (0.43), its Temporal F1 is low (0.18 vs. InternVL3.5-8B’s 0.57), indicating poor pairwise order preservation ( [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Dataset Construction Pipeline [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of epileptologist and trained-annotator seizure semiology distributions across epileptic (ES) and nonepileptic (NES) events. (a) Sorted bar plot of ES semiology feature frequencies annotated by epileptologists and trained annotators. Features are ordered by relative difference between groups, highlighting minimal divergence for high-frequency ES behaviors such as tonic, clonic, and ictal vocaliz… view at source ↗
Figure 8
Figure 8. Figure 8: Confusion matrix of Qwen2.5-VL-72B predictions on Task 1. 25 [PITH_FULL_IMAGE:figures/full_fig_p025_8.png] view at source ↗
read the original abstract

While Multimodal Large Language Models (MLLMs) have demonstrated remarkable proficiency in general video understanding, their capacity to interpret involuntary, and spatio-temporally evolving pathologic motor behaviors such as seizure semiology remains largely untested. To address this gap, we introduce Seizure-Semiology-Suite, a clinically grounded dataset and benchmark for fine-grained, structured seizure semiology understanding. The dataset includes 438 seizure videos annotated with over 35,000 dense labels covering 20 ILAE-defined semiological features. Building on this dataset, we propose a seven-task hierarchical benchmark that systematically evaluates MLLMs from low-level visual perception to temporal sequencing, narrative report generation, and seizure diagnosis. To enable clinically meaningful evaluation of generated reports, we further introduce the Report Quality Index for Seizure Semiology (Seizure-RQI). Extensive baselines across 11 open-weight MLLMs reveal systematic weaknesses in laterality reasoning, temporal localization, symptom sequencing, and clinically faithful reporting. We show that seizure-specific fine-tuning substantially improves performance across tasks, and that a two-stage neuro-symbolic framework achieves an F1 score of 0.96 on epileptic versus non-epileptic seizure classification. Seizure-Semiology-Suite establishes a rigorous benchmark for evaluating multimodal models in safety-critical medical video understanding and guides the development of clinically reliable, domain-adaptive multimodal intelligence.

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

2 major / 2 minor

Summary. The paper introduces Seizure-Semiology-Suite (S3), a dataset of 438 seizure videos with over 35,000 dense labels for 20 ILAE-defined semiological features. It defines a seven-task hierarchical benchmark evaluating MLLMs from low-level visual perception through temporal sequencing and narrative generation to seizure diagnosis, introduces the Seizure-RQI metric for clinically meaningful report assessment, and reports baselines on 11 open-weight MLLMs that reveal weaknesses in laterality, localization, sequencing, and faithful reporting. Seizure-specific fine-tuning yields consistent gains, and a two-stage neuro-symbolic framework reaches an F1 of 0.96 on epileptic versus non-epileptic classification.

Significance. If the annotations prove reliable and representative, the work supplies a much-needed structured benchmark and evaluation protocol for safety-critical medical video understanding. The systematic comparison of 11 models, the demonstration of domain-specific fine-tuning benefits, and the introduction of Seizure-RQI constitute concrete contributions that could guide future model development in this domain.

major comments (2)
  1. [Dataset section] Dataset section: the manuscript provides no inter-rater reliability statistics (e.g., Cohen’s or Fleiss’ kappa), expert count, annotation protocol, or consensus procedure for the 35,000 labels across the 438 videos and 20 ILAE features. This is load-bearing for the central empirical claims because both the seven-task benchmark and the reported 0.96 F1 on epileptic versus non-epileptic classification presuppose that the ground-truth labels are accurate, consistent, and clinically representative.
  2. [Evaluation section] Evaluation section: exact definitions of the per-task metrics and the precise formulation of Seizure-RQI are not supplied, nor are the train/validation/test splits or any cross-validation procedure. Without these details the quantitative improvements attributed to seizure-specific fine-tuning cannot be fully reproduced or interpreted.
minor comments (2)
  1. [Abstract] The abstract and introduction could more explicitly reference the ILAE 2017 classification standards when listing the 20 semiological features.
  2. [Figures] Figure captions for example video frames and annotation visualizations would benefit from additional detail on what each label represents.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We sincerely thank the referee for their detailed and constructive comments. We have carefully considered each point and provide our responses below. We will make the necessary revisions to address the concerns regarding dataset annotation details and evaluation specifics.

read point-by-point responses

  1. Referee: [Dataset section] Dataset section: the manuscript provides no inter-rater reliability statistics (e.g., Cohen’s or Fleiss’ kappa), expert count, annotation protocol, or consensus procedure for the 35,000 labels across the 438 videos and 20 ILAE features. This is load-bearing for the central empirical claims because both the seven-task benchmark and the reported 0.96 F1 on epileptic versus non-epileptic classification presuppose that the ground-truth labels are accurate, consistent, and clinically representative.

    Authors: We agree that these details are critical for validating our ground-truth annotations. In the revised manuscript, we will add comprehensive information on the annotation protocol, including the number of experts (clinical neurologists), the consensus procedure, and inter-rater reliability measures such as Fleiss' kappa for the multi-label annotations across the 20 features. This will strengthen the credibility of the benchmark and the reported performance metrics. revision: yes

  2. Referee: [Evaluation section] Evaluation section: exact definitions of the per-task metrics and the precise formulation of Seizure-RQI are not supplied, nor are the train/validation/test splits or any cross-validation procedure. Without these details the quantitative improvements attributed to seizure-specific fine-tuning cannot be fully reproduced or interpreted.

    Authors: We acknowledge the need for precise specifications to ensure reproducibility. The revised manuscript will include exact definitions and formulas for all per-task metrics, the complete mathematical formulation of the Seizure-RQI metric, detailed descriptions of the train/validation/test splits (with patient-level stratification to avoid data leakage), and any cross-validation methods used in the experiments. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical dataset and benchmark paper with direct evaluations only

full rationale

This is a dataset introduction and empirical benchmark paper. The abstract and described content contain no mathematical derivations, equations, fitted parameters, predictions, or self-referential chains. Performance claims (seizure-specific fine-tuning gains and 0.96 F1 on epileptic vs. non-epileptic classification) are presented as direct model evaluation results on the 438-video dataset with 35,000 labels. No load-bearing steps reduce to inputs by construction, self-citation, or renaming; the seven-task benchmark and Seizure-RQI are defined explicitly from the new annotations rather than derived from prior results. The paper is self-contained against external benchmarks as an empirical contribution.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an empirical dataset creation and benchmarking paper with no mathematical derivations, free parameters, axioms, or postulated entities.

pith-pipeline@v0.9.0 · 5899 in / 1162 out tokens · 56340 ms · 2026-05-22T08:08:58.077292+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

118 extracted references · 118 canonical work pages · 8 internal anchors

  1. [1]

    Langley , title =

    P. Langley , title =. Proceedings of the 17th International Conference on Machine Learning (ICML 2000) , address =. 2000 , pages =

    work page 2000

  2. [2]

    T. M. Mitchell. The Need for Biases in Learning Generalizations. 1980

    work page 1980

  3. [3]

    M. J. Kearns , title =

    work page

  4. [4]

    Machine Learning: An Artificial Intelligence Approach, Vol. I. 1983

    work page 1983

  5. [5]

    R. O. Duda and P. E. Hart and D. G. Stork. Pattern Classification. 2000

    work page 2000

  6. [6]

    Suppressed for Anonymity , author=

    work page

  7. [7]

    Newell and P

    A. Newell and P. S. Rosenbloom. Mechanisms of Skill Acquisition and the Law of Practice. Cognitive Skills and Their Acquisition. 1981

    work page 1981

  8. [8]

    A. L. Samuel. Some Studies in Machine Learning Using the Game of Checkers. IBM Journal of Research and Development. 1959

    work page 1959

  9. [9]

    Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19) , year =

    CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison , author =. Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19) , year =

    work page

  10. [10]

    Epilepsy & Behavior , volume = 154, pages =

    Ahmedt-Aristizabal, David and Armin, Mohammad Ali and Hayder, Zeeshan and Garcia-Cairasco, Norberto and Petersson, Lars and Fookes, Clinton and Denman, Simon and McGonigal, Aileen , title =. Epilepsy & Behavior , volume = 154, pages =

    work page

  11. [11]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Activitynet-qa: A dataset for understanding complex web videos via question answering , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    work page

  12. [12]

    MSR-VTT: A Large Video Description Dataset for Bridging Video and Language , year=

    Xu, Jun and Mei, Tao and Yao, Ting and Rui, Yong , booktitle=. MSR-VTT: A Large Video Description Dataset for Bridging Video and Language , year=

    work page

  13. [13]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Hong, Wenyi and Cheng, Yean and Yang, Zhuoyi and Wang, Weihan and Wang, Lefan and Gu, Xiaotao and Huang, Shiyu and Dong, Yuxiao and Tang, Jie , title =. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    work page

  14. [14]

    Journal of Clinical Neurology , year =

    Krikor Tufenkjian and Hans O Lüders , title =. Journal of Clinical Neurology , year =

    work page

  15. [15]

    Scientific Data , volume =

    Gupta, Deepak and Attal, Kush and Demner-Fushman, Dina , title =. Scientific Data , volume =. 2023 , doi =

    work page 2023

  16. [16]

    Epilepsy & Behavior , year =

    Beniczky, Simona Alexandra and Fogarasi, András and Neufeld, Miri and Andersen, Noémi Becser and Wolf, Peter and van Emde Boas, Walter and Beniczky, Sándor , title =. Epilepsy & Behavior , year =

    work page

  17. [17]

    IEEE Journal of Biomedical and Health Informatics , volume =

    Yang, Yonghua and Sarkis, Rani A and Atrache, Rima EI and Loddenkemper, Tobias and Meisel, Christian , title =. IEEE Journal of Biomedical and Health Informatics , volume =. 2021 , doi =

    work page 2021

  18. [18]

    and Kotagal, Prakash , title =

    Elwan, Sherif and Alexopoulos, Andreas and Silveira, Diosely C. and Kotagal, Prakash , title =. Seizure , volume =. 2018 , doi =

    work page 2018

  19. [19]

    arXiv preprint arXiv:2505.02064 , year=

    RTV-Bench: Benchmarking MLLM Continuous Perception, Understanding and Reasoning through Real-Time Video , author=. arXiv preprint arXiv:2505.02064 , year=

    work page arXiv

  20. [20]

    arXiv preprint arXiv:2512.22905 , year=

    Javisgpt: A unified multi-modal llm for sounding-video comprehension and generation , author=. arXiv preprint arXiv:2512.22905 , year=

    work page arXiv

  21. [21]

    SV-RCNet: Workflow Recognition From Surgical Videos Using Recurrent Convolutional Network , year=

    Jin, Yueming and Dou, Qi and Chen, Hao and Yu, Lequan and Qin, Jing and Fu, Chi-Wing and Heng, Pheng-Ann , journal=. SV-RCNet: Workflow Recognition From Surgical Videos Using Recurrent Convolutional Network , year=

    work page

  22. [22]

    Perez, D. L. and LaFrance, W. C., Jr. , title =. CNS Spectrums , volume =. 2016 , doi =

    work page 2016

  23. [24]

    Neurology , volume=

    Practice guideline summary: Sudden unexpected death in epilepsy incidence rates and risk factors: Report of the Guideline Development, Dissemination, and Implementation Subcommittee of the American Academy of Neurology and the American Epilepsy Society , author=. Neurology , volume=. 2017 , publisher=

    work page 2017

  24. [26]

    Epilepsia , volume=

    Frontal lobe seizures: from clinical semiology to localization , author=. Epilepsia , volume=. 2012 , doi=

    work page 2012

  25. [27]

    Neurology , volume=

    Lateralizing significance of head and eye deviation in secondary generalized tonic-clonic seizures , author=. Neurology , volume=. 1993 , doi=

    work page 1993

  26. [28]

    Epilepsia , volume=

    Lateralizing value of asymmetric tonic limb posturing observed in secondarily generalized tonic-clonic seizures , author=. Epilepsia , volume=. 2000 , doi=

    work page 2000

  27. [30]

    Epilepsia Open , volume=

    Prospective study of epilepsy with generalized tonic--clonic seizures alone: Clinical features, response to treatment, and likelihood of medication withdrawal , author=. Epilepsia Open , volume=. 2024 , publisher=

    work page 2024

  28. [31]

    Journal of Clinical Neurophysiology , year =

    Value and limitations of seizure semiology in localizing seizure onset , author =. Journal of Clinical Neurophysiology , year =. doi:10.1097/01.wnp.0000228498.71365.7b , pmid =

    work page doi:10.1097/01.wnp.0000228498.71365.7b

  29. [32]

    Journal of clinical neurology (Seoul, Korea) , volume=

    Seizure semiology: its value and limitations in localizing the epileptogenic zone , author=. Journal of clinical neurology (Seoul, Korea) , volume=

    work page

  30. [34]

    CONTINUUM: Lifelong Learning in Neurology , volume=

    Diagnosis and treatment of nonepileptic seizures , author=. CONTINUUM: Lifelong Learning in Neurology , volume=. 2016 , publisher=

    work page 2016

  31. [35]

    Annals of internal medicine , volume=

    Clinical documentation in the 21st century: executive summary of a policy position paper from the American College of Physicians , author=. Annals of internal medicine , volume=. 2015 , publisher=

    work page 2015

  32. [36]

    JAMA internal medicine , volume=

    Medical documentation burden among US office-based physicians in 2019: a national study , author=. JAMA internal medicine , volume=. 2022 , publisher=

    work page 2019

  33. [43]

    Advances in Neural Information Processing Systems , volume=

    Audio flamingo 3: Advancing audio intelligence with fully open large audio language models , author=. Advances in Neural Information Processing Systems , volume=

    work page

  34. [45]

    Advances in neural information processing systems , volume =

    Training language models to follow instructions with human feedback , author =. Advances in neural information processing systems , volume =

    work page

  35. [47]

    Nature Communications , volume =

    The Medical Segmentation Decathlon , author =. Nature Communications , volume =. 2022 , doi =

    work page 2022

  36. [50]

    What’s left can’t be right – the remaining positional incompetence of contrastive vision-language models, 2023

    What’s left can’t be right – The remaining positional incompetence of contrastive vision-language models , author =. 2023 , eprint =. doi:10.48550/arXiv.2311.11477 , url =

    work page doi:10.48550/arxiv.2311.11477 2023

  37. [51]

    Epileptic Disorders , year =

    Seizure semiology: ILAE glossary of terms and their significance , author =. Epileptic Disorders , year =. doi:10.1684/epd.2022.1430 , pmid =

    work page doi:10.1684/epd.2022.1430 2022

  38. [52]

    Epilepsy & Behavior , volume=

    The diagnostic utility of the ictal cry , author=. Epilepsy & Behavior , volume=. 2010 , publisher=

    work page 2010

  39. [53]

    Neuropsychiatric disease and treatment , pages=

    Long-term video EEG monitoring for diagnosis of psychogenic nonepileptic seizures , author=. Neuropsychiatric disease and treatment , pages=. 2014 , publisher=

    work page 2014

  40. [54]

    Journal of Clinical Neuroscience , volume=

    Epilepsy with eyelid myoclonia: A systematic review and meta-analysis , author=. Journal of Clinical Neuroscience , volume=. 2025 , publisher=

    work page 2025

  41. [55]

    MAMC Journal of Medical Sciences , volume=

    Psychogenic nonepileptic seizures (PNES): A review , author=. MAMC Journal of Medical Sciences , volume=. 2019 , publisher=

    work page 2019

  42. [56]

    Convulsive

    “Convulsive” nonepileptic seizures have a characteristic pattern of rhythmic artifact distinguishing them from convulsive epileptic seizures , author=. Epilepsia , volume=. 2004 , publisher=

    work page 2004

  43. [57]

    Neurology: Clinical Practice , volume=

    Using semiology to classify epileptic seizures vs psychogenic nonepileptic seizures: a meta-analysis , author=. Neurology: Clinical Practice , volume=. 2022 , publisher=

    work page 2022

  44. [58]

    Journal of neurology , volume=

    Predictive semiology of psychogenic non-epileptic seizures in an epilepsy monitoring unit , author=. Journal of neurology , volume=. 2022 , publisher=

    work page 2022

  45. [59]

    The Lancet Neurology , volume=

    Impaired consciousness in epilepsy , author=. The Lancet Neurology , volume=. 2012 , publisher=

    work page 2012

  46. [60]

    Seizure , volume=

    Clinical features of automatisms and correlation with the seizure onset zones: A cluster analysis of 74 surgically-treated cases , author=. Seizure , volume=. 2022 , publisher=

    work page 2022

  47. [61]

    Journal of Clinical Neurophysiology , volume=

    Mesial temporal lobe epilepsy , author=. Journal of Clinical Neurophysiology , volume=. 2012 , publisher=

    work page 2012

  48. [62]

    Sleep , volume=

    NREM arousal parasomnias and their distinction from nocturnal frontal lobe epilepsy: a video EEG analysis , author=. Sleep , volume=. 2009 , publisher=

    work page 2009

  49. [63]

    Journal of personalized medicine , volume=

    The reciprocal relationship between sleep and epilepsy , author=. Journal of personalized medicine , volume=. 2024 , publisher=

    work page 2024

  50. [64]

    Neurology: Clinical Practice , volume=

    Semiology and neurophysiology of clonic seizures: a report of 39 patients , author=. Neurology: Clinical Practice , volume=. 2024 , publisher=

    work page 2024

  51. [65]

    Epilepsia , volume=

    The generalized tonic-clonic seizure in partial versus generalized epilepsy: semiologic differences , author=. Epilepsia , volume=. 1999 , publisher=

    work page 1999

  52. [66]

    Handbook of Clinical Neurology , volume=

    Motor seizure semiology , author=. Handbook of Clinical Neurology , volume=. 2023 , publisher=

    work page 2023

  53. [67]

    Epilepsia , year =

    Glossary of descriptive terminology for ictal semiology: Report of the ILAE task force on classification and terminology , author =. Epilepsia , year =

    work page

  54. [69]

    Npj health systems , volume=

    Current and future state of evaluation of large language models for medical summarization tasks , author=. Npj health systems , volume=. 2025 , publisher=

    work page 2025

  55. [70]

    Findings of the association for computational linguistics: EMNLP 2024 , pages=

    Green: Generative radiology report evaluation and error notation , author=. Findings of the association for computational linguistics: EMNLP 2024 , pages=

    work page 2024

  56. [71]

    International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=

    Privacy-preserving early detection of epileptic seizures in videos , author=. International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=. 2023 , organization=

    work page 2023

  57. [72]

    Epilepsia , pages=

    Video-based detection of tonic–clonic seizures using a three-dimensional convolutional neural network , author=. Epilepsia , pages=

    work page

  58. [73]

    Epilepsy & Behavior , pages=

    Automated detection of tonic–clonic seizures using 3-D accelerometry and surface electromyography in pediatric patients , author=. Epilepsy & Behavior , pages=

    work page

  59. [74]

    Medical Engineering & Physics , pages=

    Automatic segmentation of episodes containing epileptic clonic seizures in video sequences , author=. Medical Engineering & Physics , pages=

    work page

  60. [75]

    IEEE Transactions on Biomedical Engineering , year =

    Automatic Segmentation of Episodes Containing Epileptic Clonic Seizures in Video Sequences , author =. IEEE Transactions on Biomedical Engineering , year =

    work page

  61. [76]

    Epilepsia , year =

    Convulsive seizure detection using a wrist-worn electrodermal activity and accelerometry biosensor , author =. Epilepsia , year =

    work page

  62. [77]

    Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =

    What's ``up'' with vision-language models? Investigating their struggle with spatial reasoning , author =. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =. 2023 , address =

    work page 2023

  63. [79]

    2026 , howpublished =

    Introducing GPT-5.4 , author =. 2026 , howpublished =

    work page 2026

  64. [80]

    S., Khashman, M., and Ahmed, M

    Ahmed, H. S., Khashman, M., and Ahmed, M. T. Epilepsy with eyelid myoclonia: A systematic review and meta-analysis. Journal of Clinical Neuroscience, 140: 0 111535, 2025

    work page 2025

  65. [82]

    Blair, R. D. G. Temporal lobe epilepsy semiology. Epilepsy Research and Treatment, 2012: 0 751510, 2012. doi:10.1155/2012/751510. URL https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3420439/

    work page doi:10.1155/2012/751510 2012

  66. [83]

    T., L \"u ders, H., Mizrahi, E., Tassinari, C., van Emde Boas, W., and Engel, J

    Blume, W. T., L \"u ders, H., Mizrahi, E., Tassinari, C., van Emde Boas, W., and Engel, J. Glossary of descriptive terminology for ictal semiology: Report of the ilae task force on classification and terminology. Epilepsia, 42 0 (9): 0 1212--1218, 2001. doi:10.1046/j.1528-1157.2001.22001.x

    work page doi:10.1046/j.1528-1157.2001.22001.x 2001

  67. [84]

    Impaired consciousness in epilepsy

    Blumenfeld, H. Impaired consciousness in epilepsy. The Lancet Neurology, 11 0 (9): 0 814--826, 2012

    work page 2012

  68. [85]

    C., Goyal, S., Sabtala, M

    Bonini, F., McGonigal, A., Trébuchon, A., Gavaret, M., Bartolomei, F., and Guye, M. Frontal lobe seizures: from clinical semiology to localization. Epilepsia, 53 0 (5): 0 815--826, 2012. doi:10.1111/j.1528-1167.2012.03440.x

    work page doi:10.1111/j.1528-1167.2012.03440.x 2012

  69. [86]

    Boyne, J. et al. Video-based detection of tonic–clonic seizures using a three-dimensional convolutional neural network. Epilepsia, pp.\ 2495--2506, 2025

    work page 2025

  70. [87]

    Psychogenic nonepileptic seizures (pnes): A review

    Chaudhry, N., Dhamija, K., and Puri, V. Psychogenic nonepileptic seizures (pnes): A review. MAMC Journal of Medical Sciences, 5 0 (1): 0 1--7, 2019

    work page 2019

  71. [88]

    Chen, D. K. and LaFrance Jr, W. C. Diagnosis and treatment of nonepileptic seizures. CONTINUUM: Lifelong Learning in Neurology, 22 0 (1): 0 116--131, 2016

    work page 2016

  72. [89]

    A., Silva, R., Whatley, B., and Walker, M

    Chowdhury, F. A., Silva, R., Whatley, B., and Walker, M. C. Localisation in focal epilepsy: a practical guide. Practical Neurology, 21 0 (6): 0 481--491, 2021. doi:10.1136/practneurol-2019-002341

    work page doi:10.1136/practneurol-2019-002341 2021

  73. [90]

    Current and future state of evaluation of large language models for medical summarization tasks

    Croxford, E., Gao, Y., Pellegrino, N., Wong, K., Wills, G., First, E., Liao, F., Goswami, C., Patterson, B., and Afshar, M. Current and future state of evaluation of large language models for medical summarization tasks. Npj health systems, 2 0 (1): 0 6, 2025

    work page 2025

  74. [91]

    P., Harvey, A

    Derry, C. P., Harvey, A. S., Walker, M. C., Duncan, J. S., and Berkovic, S. F. Nrem arousal parasomnias and their distinction from nocturnal frontal lobe epilepsy: a video eeg analysis. Sleep, 32 0 (12): 0 1637--1644, 2009

    work page 2009

  75. [92]

    J., Peric, I., Boston, R., and Seneviratne, U

    Duncan, A. J., Peric, I., Boston, R., and Seneviratne, U. Predictive semiology of psychogenic non-epileptic seizures in an epilepsy monitoring unit. Journal of neurology, 269 0 (4): 0 2172--2178, 2022

    work page 2022

  76. [93]

    S., Lin, K., and Benbadis, S

    Elzawahry, H., Do, C. S., Lin, K., and Benbadis, S. The diagnostic utility of the ictal cry. Epilepsy & Behavior, 18 0 (3): 0 306--307, 2010

    work page 2010

  77. [94]

    Fotedar, N., Acar, A., Hakami, S., Praditukrit, K., Morris, A., dela Vega, M., Fernandez-BacaVaca, G., and L \"u ders, H. O. Semiology and neurophysiology of clonic seizures: a report of 39 patients. Neurology: Clinical Practice, 14 0 (2): 0 e200252, 2024

    work page 2024

  78. [95]

    Gedzelman, E. R. and LaRoche, S. M. Long-term video eeg monitoring for diagnosis of psychogenic nonepileptic seizures. Neuropsychiatric disease and treatment, pp.\ 1979--1986, 2014

    work page 1979

  79. [96]

    Audio flamingo 3: Advancing audio intelligence with fully open large audio language models

    Ghosh, S., Goel, A., Kim, J., Kumar, S., Kong, Z., Lee, S.-g., Yang, C.-H., Duraiswami, R., Manocha, D., Valle, R., et al. Audio flamingo 3: Advancing audio intelligence with fully open large audio language models. Advances in Neural Information Processing Systems, 38: 0 41819--41886, 2026

    work page 2026

  80. [97]

    A dataset for medical instructional video classification and question answering

    Gupta, D., Attal, K., and Demner-Fushman, D. A dataset for medical instructional video classification and question answering. Scientific Data, 10 0 (1): 0 158, 2023. doi:10.1038/s41597-023-02036-y

    work page doi:10.1038/s41597-023-02036-y 2023

Showing first 80 references.