Skip to content
DraconiticResearch
Search Sources API
186,631 studies31 sourcesUpdated 37 min ago
Peer-reviewedCC BYJournal of healthcare informatics research

Weakly-Supervised Detection of Facial Motor Tics in Video Data Using Score-Driven Top-k Multiple Instance Learning

  • Nele Sophie Brügge1
  • Frédéric Li1
  • Ronja Schappert2
  • Gesine Marie Sallandt2,3
  • Christian Beste4,5
  • Sebastian Fudickar6,7
  • Marcin Grzegorzek1,8
  • Alexander Münchau2
  • Heinz Handels1,8
8 affiliations
  1. 1German Research Center for Artificial Intelligence, Lübeck, Germany
  2. 2Institute of Systems Motor Science, University of Luebeck, Lübeck, Germany
  3. 3Department of Neurology, University Hospital Medical Center Schleswig-Holstein, Lübeck, Germany
  4. 4Cognitive Neurophysiology, Department of Child and Adolescent Psychiatry, Faculty of Medicine, TU Dresden, Dresden, Germany
  5. 5Cognitive Psychology, Faculty of Psychology, Shandong Normal University, Jinan, China
  6. 6Fraunhofer Research Institution for Individualized and Cell-Based Medical Engineering (IMTE), Lübeck, Germany
  7. 7Section for Clinical Research IT (SKFIT), Institute of Medical Biometry and Statistics, University of Luebeck, Lübeck, Germany
  8. 8Institute of Medical Informatics, University of Luebeck, Lübeck, Germany
Read full textPDFDOICite

Abstract

In medicine, the laborious annotation process poses a significant limitation to the creation of large labelled datasets, a challenge that is particularly pronounced for medical video data, where numerous frames require expert annotation. Weakly supervised approaches, such as Multiple Instance Learning (MIL), offer a promising solution by leveraging only video-level labels rather than frame-by-frame annotations. In this work, we tailor a top-k MIL-based approach to medical video data to detect and localise facial motor tics in videos of patients diagnosed with Gilles de la Tourette Syndrome (GTS). Our adaptations include a score-driven top-k instance selection strategy that accounts for variability in tic occurrence between subjects. This enables the effective exploitation of large unlabelled datasets by incorporating weak labels where available. Evaluated against state-of-the-art MIL models, the proposed approach achieves the highest instance-level tic detection performance, with an area under the curve of 86.73 %. These results demonstrate that facial motor tics in videos of individuals with GTS can be effectively localised using only weak video-level annotations, highlighting the potential of the proposed approach for future clinical applications.

Related

  1. Weakly Supervised Multiple-Instance Learning for Seizure Onset Zone Identification
    J. Mao, A.-C. LESAGE, L. Camarillo-Rodriguez +4
    medRxivPreprintCC BY-NC-NDFull textNeurologymedRxiv
  2. Weakly-Supervised Gated Attention Multiple Instance Learning for Multi-Center Cervical Cytology Slide Triage
    S. Mayala, S. Sang'udi, D. Mzurikwao +1
    medRxivPreprintCC BYFull textEpidemiology & public healthmedRxiv
  3. Weakly supervised attention-based learning supports breast cancer detection on deep ultraviolet-excitation fluorescence images
    Ryuta Nakao, Tomoya Matsui, Masatoshi Abe +8
    Scientific ReportsPeer-reviewedCC BYFull textGeneral medicineDOAJ
  4. FragMIL: A Fragment-Token Multiple-Instance Learning Framework for Cancer Detection from Cell-Free DNA Fragmentomics
    Tien-Dung Bui, Yung-Feng Lin, Mai-Hanh Nguyen +3
    CancersPeer-reviewedCC BY 4.0Full textOncologyCrossrefEurope PMC
  5. Multimodal Multiple Instance Learning for Fibroepithelial Tumor Diagnosis in Breast Ultrasound
    F. F. Abir, M. Ali, M. S. Hossain +5
    medRxivPreprintAll rights reservedFull textMedical imaging & radiologymedRxiv

Cite

BibTeXRIS
BibTeX
@article{brugge2026weaklysupervised,
  title = {{Weakly-Supervised Detection of Facial Motor Tics in Video Data Using Score-Driven Top-k Multiple Instance Learning}},
  author = {Nele Sophie Brügge and Frédéric Li and Ronja Schappert and Gesine Marie Sallandt and Christian Beste and Sebastian Fudickar and Marcin Grzegorzek and Alexander Münchau and Heinz Handels},
  journal = {Journal of healthcare informatics research},
  year = {2026},
  month = jul,
  doi = {10.1007/s41666-026-00233-z},
  url = {https://europepmc.org/article/PMC/PMC13719929},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Weakly-Supervised Detection of Facial Motor Tics in Video Data Using Score-Driven Top-k Multiple Instance Learning
AU  - Nele Sophie Brügge
AU  - Frédéric Li
AU  - Ronja Schappert
AU  - Gesine Marie Sallandt
AU  - Christian Beste
AU  - Sebastian Fudickar
AU  - Marcin Grzegorzek
AU  - Alexander Münchau
AU  - Heinz Handels
PY  - 2026
DA  - 2026/07/30
JO  - Journal of healthcare informatics research
DO  - 10.1007/s41666-026-00233-z
UR  - https://europepmc.org/article/PMC/PMC13719929
AB  - In medicine, the laborious annotation process poses a significant limitation to the creation of large labelled datasets, a challenge that is particularly pronounced for medical video data, where numerous frames require expert annotation. Weakly supervised approaches, such as Multiple Instance Learning (MIL), offer a promising solution by leveraging only video-level labels rather than frame-by-frame annotations. In this work, we tailor a top-k MIL-based approach to medical video data to detect and localise facial motor tics in videos of patients diagnosed with Gilles de la Tourette Syndrome (GTS). Our adaptations include a score-driven top-k instance selection strategy that accounts for variability in tic occurrence between subjects. This enables the effective exploitation of large unlabelled datasets by incorporating weak labels where available. Evaluated against state-of-the-art MIL models, the proposed approach achieves the highest instance-level tic detection performance, with an area under the curve of 86.73 %. These results demonstrate that facial motor tics in videos of individuals with GTS can be effectively localised using only weak video-level annotations, highlighting the potential of the proposed approach for future clinical applications.
SN  - 2509-4971
SN  - 2509-498X
ER  -