Development and validation of a video-based deep learning model for distinguishing epileptic seizures from non-epileptic events in a pediatric cohort

IF 2.3 3区医学 Q2 BEHAVIORAL SCIENCES

Epilepsy & Behavior Pub Date : 2025-10-18 DOI:10.1016/j.yebeh.2025.110785

Ping Ding , Tinghong Liu , Jinshan Xu, Liu Yuan, Liwei Zhang, Zhirong Wei, Yuchen Tai, Shuli Liang

{"title":"Development and validation of a video-based deep learning model for distinguishing epileptic seizures from non-epileptic events in a pediatric cohort","authors":"Ping Ding , Tinghong Liu , Jinshan Xu, Liu Yuan, Liwei Zhang, Zhirong Wei, Yuchen Tai, Shuli Liang","doi":"10.1016/j.yebeh.2025.110785","DOIUrl":null,"url":null,"abstract":"<div><h3>Objective</h3><div>This study aimed to develop and validate a video-based deep learning system for distinguishing epileptic seizures (ES) from non-epileptic events (NEE) in a pediatric cohort. Using a prospective validation cohort, we further assessed the diagnostic performance and clinical applicability of the artificial intelligence (AI) model, investigated potential factors contributing to its diagnostic errors, and benchmarked its clinical utility against clinicians grouped by different levels of expertise.</div></div><div><h3>Methods</h3><div>An enhanced multiscale vision transformer was trained on 438 retrospectively collected videos, with benchmark comparisons against MViTv2 and SlowFast architectures. Prospective validation was performed using 130 consecutive videos to assess the diagnostic performance of the AI system against tiered clinician groups (interns, attending physicians, and chief physicians). A generalized linear mixed model (GLMM) was employed to identify factors associated with AI misdiagnosis, with further comparative analysis of diagnostic performance between AI and human clinicians.</div></div><div><h3>Results</h3><div>Our model demonstrated significantly higher accuracy (p = 0.001) and sensitivity (p = 0.004) compared to the MViTv2 model. Although all performance metrics were numerically higher than those of the SlowFast model, these differences did not reach statistical significance. GLMM analysis indicated that event type (motor vs. non-motor) was a significant factor influencing model misclassification (p = 0.020). The model achieved substantially higher diagnostic accuracy for motor events compared to non-motor events (p < 0.001).</div></div><div><h3>Conclusion</h3><div>The video-based AI classifier shows promise as an assistive tool for clinicians in differentiating ES from NEE based on video evidence in a pediatric cohort. Our AI model demonstrated notably effective diagnostic performance for motor events, while its accuracy was more limited for non-motor events.</div></div>","PeriodicalId":11847,"journal":{"name":"Epilepsy & Behavior","volume":"173 ","pages":"Article 110785"},"PeriodicalIF":2.3000,"publicationDate":"2025-10-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Epilepsy & Behavior","FirstCategoryId":"3","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1525505025005256","RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"BEHAVIORAL SCIENCES","Score":null,"Total":0}

引用次数: 0

Abstract

Objective

This study aimed to develop and validate a video-based deep learning system for distinguishing epileptic seizures (ES) from non-epileptic events (NEE) in a pediatric cohort. Using a prospective validation cohort, we further assessed the diagnostic performance and clinical applicability of the artificial intelligence (AI) model, investigated potential factors contributing to its diagnostic errors, and benchmarked its clinical utility against clinicians grouped by different levels of expertise.

Methods

An enhanced multiscale vision transformer was trained on 438 retrospectively collected videos, with benchmark comparisons against MViTv2 and SlowFast architectures. Prospective validation was performed using 130 consecutive videos to assess the diagnostic performance of the AI system against tiered clinician groups (interns, attending physicians, and chief physicians). A generalized linear mixed model (GLMM) was employed to identify factors associated with AI misdiagnosis, with further comparative analysis of diagnostic performance between AI and human clinicians.

Results

Our model demonstrated significantly higher accuracy (p = 0.001) and sensitivity (p = 0.004) compared to the MViTv2 model. Although all performance metrics were numerically higher than those of the SlowFast model, these differences did not reach statistical significance. GLMM analysis indicated that event type (motor vs. non-motor) was a significant factor influencing model misclassification (p = 0.020). The model achieved substantially higher diagnostic accuracy for motor events compared to non-motor events (p < 0.001).

Conclusion

The video-based AI classifier shows promise as an assistive tool for clinicians in differentiating ES from NEE based on video evidence in a pediatric cohort. Our AI model demonstrated notably effective diagnostic performance for motor events, while its accuracy was more limited for non-motor events.

查看原文本刊更多论文

开发和验证基于视频的深度学习模型，用于区分儿童队列中的癫痫发作和非癫痫事件

本研究旨在开发和验证一种基于视频的深度学习系统，用于在儿童队列中区分癫痫发作（ES）和非癫痫事件（NEE）。通过前瞻性验证队列，我们进一步评估了人工智能（AI）模型的诊断性能和临床适用性，调查了导致其诊断错误的潜在因素，并对不同专业水平分组的临床医生进行了基准测试。方法对438个回顾性收集的视频进行增强型多尺度视觉转换器的训练，并与MViTv2和SlowFast架构进行基准比较。使用130个连续视频进行前瞻性验证，以评估人工智能系统对分层临床医生组（实习生、主治医生和主任医生）的诊断性能。采用广义线性混合模型（GLMM）识别与人工智能误诊相关的因素，并进一步比较分析人工智能与人类临床医生的诊断表现。结果与MViTv2模型相比，我们的模型具有更高的准确性（p = 0.001）和灵敏度（p = 0.004）。尽管所有性能指标在数值上都高于SlowFast模型，但这些差异没有达到统计学意义。GLMM分析显示，事件类型（运动与非运动）是影响模型误分类的重要因素（p = 0.020）。与非运动事件相比，该模型对运动事件的诊断准确率显著提高（p < 0.001）。结论基于视频的人工智能分类器有望成为临床医生根据儿童队列视频证据区分ES和NEE的辅助工具。我们的人工智能模型对运动事件的诊断表现出了显著的有效性，而对非运动事件的准确性则更为有限。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Epilepsy & Behavior 医学-行为科学

CiteScore

5.40

自引率

15.40%

发文量

385

审稿时长

43 days

期刊介绍： Epilepsy & Behavior is the fastest-growing international journal uniquely devoted to the rapid dissemination of the most current information available on the behavioral aspects of seizures and epilepsy. Epilepsy & Behavior presents original peer-reviewed articles based on laboratory and clinical research. Topics are drawn from a variety of fields, including clinical neurology, neurosurgery, neuropsychiatry, neuropsychology, neurophysiology, neuropharmacology, and neuroimaging. From September 2012 Epilepsy & Behavior stopped accepting Case Reports for publication in the journal. From this date authors who submit to Epilepsy & Behavior will be offered a transfer or asked to resubmit their Case Reports to its new sister journal, Epilepsy & Behavior Case Reports.