EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Feasibility of improving vocal fold pathology image classification with synthetic images generated by DDPM-based GenAI: a pilot study

Iman Khazrak iD, Shahryar Zainaee iD, Mostafa M Rezaee iD, Mehran Ghasemi iD, Robert C Green iD

European Archives of Oto-Rhino-Laryngology · 2025

DOI: 10.1007/s00405-025-09443-4 ↗

Kurz gefasst

Fachartikel in der Zeitschrift European Archives of Oto-Rhino-Laryngology, 2025. Verfasst von Iman Khazrak und 4 weiteren Autor:innen. Beteiligte Einrichtung: Department of Computer Science, Bowling Green State University, Bowling Green, OH, 43403, USA. ikhazra@bgsu.edu. Hauptthemen laut MEDLINE: Vocal Cords, Laryngoscopy, Voice Disorders, Laryngeal Diseases. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 45 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Voice disorders (VD) are often linked to vocal fold structural pathologies (VFSP). Laryngeal imaging plays a vital role in assessing VFSPs and VD in clinical and research settings, but challenges like scarce and imbalanced datasets can limit the generalizability of findings. Denoising Diffusion Probabilistic Models (DDPMs), a subtype of Generative AI, has gained attention for its ability to generate high-quality and realistic synthetic images to address these challenges.

Purpose

This study explores the feasibility of improving VFSP image classification by generating synthetic images using DDPMs.

Methods

404 laryngoscopic images depicting VF without and with VFSP were included. DDPMs were used to generate synthetic images to augment the original dataset. Two convolutional neural network architectures, VGG16 and ResNet50, were applied for model training. The models were initially trained only on the original dataset. Then, they were trained on the augmented datasets. Evaluation metrics were analyzed to assess the performance of the models for both binary classification (with/without VFSPs) and multi-class classification (seven specific VFSPs).

Results

Realistic and high-quality synthetic images were generated for dataset augmentation. The model first failed to converge when trained only on the original dataset, but they successfully converged and achieved low loss and high accuracy when trained on the augmented datasets. The best performance was gained for both binary and multi-class classification when the models were trained on an augmented dataset.

Conclusion

Generating realistic images of VFSP using DDPMs is feasible and can enhance the classification of VFSPs by an AI model and may support VD screening and diagnosis.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Vocal CordsLaryngoscopyVoice DisordersLaryngeal DiseasesHumansPilot ProjectsFeasibility StudiesNeural Networks, ComputerModels, StatisticalFemaleMale

Stichwörter der Autor:innen

Denoising diffusion probabilistic modelsDysphoniaGenerative artificial intelligenceVocal fold imagingVocal fold pathologyVoice disorders

Autor:innen und Institutionen

  • Department of Computer Science, Bowling Green State University, Bowling Green, OH, 43403, USA. ikhazra@bgsu.edu.
  • Department of Communication Sciences and Disorders, Bowling Green State University, Bowling Green, OH, 43403, USA.
  • Department of Computer Science, Bowling Green State University, Bowling Green, OH, 43403, USA.

ORCID-Profile: Iman Khazrak, Shahryar Zainaee, Mostafa M Rezaee, Mehran Ghasemi, Robert C Green

Zitieren

APA

Khazrak, I., Zainaee, S., M Rezaee, M., Ghasemi, M., & C Green, R. (2025). Feasibility of improving vocal fold pathology image classification with synthetic images generated by DDPM-based GenAI: a pilot study. European Archives of Oto-Rhino-Laryngology. https://doi.org/10.1007/s00405-025-09443-4

Vancouver

Khazrak I, Zainaee S, M Rezaee M, Ghasemi M, C Green R. Feasibility of improving vocal fold pathology image classification with synthetic images generated by DDPM-based GenAI: a pilot study. European Archives of Oto-Rhino-Laryngology. 2025. doi:10.1007/s00405-025-09443-4

BibTeX
@article{Khazrak2025,
  author = {Khazrak, Iman and Zainaee, Shahryar and M Rezaee, Mostafa and Ghasemi, Mehran and C Green, Robert},
  title = {Feasibility of improving vocal fold pathology image classification with synthetic images generated by DDPM-based GenAI: a pilot study},
  journal = {European Archives of Oto-Rhino-Laryngology},
  year = {2025},
  doi = {10.1007/s00405-025-09443-4}
}
Kostenlos starten →