EUVIMEDEUVIMED entdecken
Fall-Kontroll-Studie

Large Vision-Language Models for Pediatric Motor Tic Detection: Case-Control Diagnostic Study

Yujun Wu iD, Yue Niu iD, Zhao Xu iD, Jing Liu iD, Xianru Jiao iD, Xia Cui iD, Dan Sun iD, Zhixian Yang iD

JMIR AI · 06.11.2026 · 5 · e98782

DOI: 10.2196/98782 ↗

Kurz gefasst

Fall-Kontroll-Studie in der Zeitschrift JMIR AI, online erschienen am 06.11.2026. Verfasst von Yujun Wu und 7 weiteren Autor:innen. Beteiligte Einrichtung: Department of Pediatrics, Peking University People's Hospital, Beijing, China. Die Arbeit nennt 37 Literaturangaben.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Clinical diagnosis of tic disorders (TDs) is often delayed by fluctuating symptoms and the subjective nature of assessments. While automated methods have been proposed to bridge this gap, traditional machine learning models frequently lack robustness in unconstrained environments and fail to capture the complex, multisite tics characteristic of severe TDs.

Objective

This study aimed to develop and evaluate a fine-tuned large vision-language model framework for the automated, interpretable detection of pediatric motor tics, addressing the clinical need for objective and efficient symptom assessment in unconstrained environments.

Methods

In this multicenter study, 168 pediatric participants (n=83, 49.4% with TDs and n=85, 50.6% healthy controls) underwent 12-hour continuous video-electroencephalography monitoring. We fine-tuned the Qwen3-VL-8B model using 3- to 5-minute video segments via low-rank adaptation. The framework used chain-of-thought prompting to generate both binary classifications and textual justifications. A strict patient-wise split ensured zero identity overlap between training and test sets. Diagnostic precision was evaluated using 95% CIs calculated with the Clopper-Pearson exact method.

Results

In total, 168 pediatric participants (age: mean 8.3, SD 2.3 years) were enrolled, providing 559 expert-validated video segments for analysis. On the independent test set (n=55), the framework achieved 89.1% (49/55; 95% CI 77.8%-95.9%) accuracy and an area under the curve of 0.950 (95% CI 0.892-1.000), significantly outperforming the landmark-based deep neural network (accuracy 74.1%; P=.004); recall reached 100% (25/25 positive episodes; 95% CI 86.3%-100.0%). Model-generated justifications showed substantial agreement with expert annotations (κ=0.784, 95% CI 0.625-0.943; P

Schlagwörter

Stichwörter der Autor:innen

AICoTLVLMartificial intelligenceautomated tic detectionchain-of-thought reasoningclinical interpretabilitylarge vision-language modelparameter-efficient fine-tuningtic disordersvideo-based monitoring

Autor:innen und Institutionen

  • Department of Pediatrics, Peking University People's Hospital, Beijing, China.
  • Department of Pediatrics, the Third Affiliated Hospital of Beijing University of Chinese Medicine, Beijing University of Chinese Medicine, Beijing, Beijing, China.
  • Department of Neurology, Tongji Medical College, Wuhan Children's Hospital, Huazhong University of Science and Technology, Hubei, China.

ORCID-Profile: Yujun Wu, Yue Niu, Zhao Xu, Jing Liu, Xianru Jiao, Xia Cui, Dan Sun, Zhixian Yang

Zitieren

APA

Wu, Y., Niu, Y., Xu, Z., Liu, J., Jiao, X., Cui, X., Sun, D., & Yang, Z. (2026). Large Vision-Language Models for Pediatric Motor Tic Detection: Case-Control Diagnostic Study. JMIR AI, 5, e98782. https://doi.org/10.2196/98782

Vancouver

Wu Y, Niu Y, Xu Z, Liu J, Jiao X, Cui X, et al. Large Vision-Language Models for Pediatric Motor Tic Detection: Case-Control Diagnostic Study. JMIR AI. 2026;5:e98782. doi:10.2196/98782

BibTeX
@article{Wu2026,
  author = {Wu, Yujun and Niu, Yue and Xu, Zhao and Liu, Jing and Jiao, Xianru and Cui, Xia and Sun, Dan and Yang, Zhixian},
  title = {Large Vision-Language Models for Pediatric Motor Tic Detection: Case-Control Diagnostic Study},
  journal = {JMIR AI},
  year = {2026},
  volume = {5},
  pages = {e98782},
  doi = {10.2196/98782}
}
Kostenlos starten →