EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Improving Diagnostic Sensitivity in Imbalanced Oral Cancer Image Classification: A Comparative Study of CNN and Transformer Architectures

Pablo Ormeño-Arriagada iD, Valentina Zúñiga iD, Carlos Toro iD, Gastón Márquez iD, David Araya iD, Diego Mellado iD, Carla Taramasco iD

Cancers · 09.09.2026 · 18 (18) · 2926

DOI: 10.3390/cancers18182926 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Cancers, online erschienen am 09.09.2026. Verfasst von Pablo Ormeño-Arriagada und 6 weiteren Autor:innen. Beteiligte Einrichtung: Instituto de Tecnología para la Innovación en Salud y Bienestar, Facultad de Ingenieria, Universidad Andres Bello, Viña del Mar 2530958, Chile. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Agencia Nacional de Investigación y Desarrollo of Chile, Centro de Prevención y Control de Cáncer.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Class imbalance remains a major limitation in artificial intelligence based oral cancer diagnosis, particularly in small clinical image datasets where malignant lesions are underrepresented.

Methods

Using a dataset of 3000 mobile-acquired oral cavity images categorized into four diagnostic classes (healthy, benign, oral potentially malignant disorders, and oral cancer), we systematically evaluated the impact of imbalance mitigation strategies on diagnostic performance. Three backbone architectures, EfficientNet, Vision Transformer, and Swin Transformer, were trained under four conditions: raw imbalanced training, random under-sampling, random over-sampling, and medical-safe data augmentation. Performance was evaluated using stratified five-fold cross-validation, and all reported metrics correspond to the mean performance across the validation folds.

Results

In multiclass evaluation, augmentation yielded the strongest overall performance, with EfficientNet achieving the highest macro-F1 score (0.669 ± 0.024) and area under curve (0.881 ± 0.015). Under clinically oriented binary malignant-risk evaluation (high-risk vs. low-risk lesions), sensitivity reached 0.817 ± 0.030 with EfficientNet, while data augmentation maintained competitive malignant-risk detection across the evaluated architectures. In contrast, random undersampling consistently yielded lower descriptive performance across architectures.

Conclusions

These findings indicate that augmentation-based imbalance mitigation is associated with improved multiclass and oral cancer specific performance while maintaining competitive overall discrimination. The results provide practical guidance for developing clinically robust artificial intelligence systems for early oral cancer detection in imbalanced real-world datasets.

Schlagwörter

Stichwörter der Autor:innen

class imbalanceconvolutional neural networksdata augmentationmalignant-risk detectionmedical image classificationoral cancervision transformer

Autor:innen und Institutionen

  • Instituto de Tecnología para la Innovación en Salud y Bienestar, Facultad de Ingenieria, Universidad Andres Bello, Viña del Mar 2530958, Chile.
  • CECAN, Centro para la Prevención y Control del Cancer, Santiago 8330077, Chile.
  • Faculty of Engineering, Universidad Andres Bello, Autopista Concepción-Talcahuano CCP-THNO 7100, Talcahuano 4260000, Chile.
  • Departamento de Ciencias de la Computación y Tecnología de la Información, Universidad del Biobio, Chillán 4081112, Chile.

ORCID-Profile: Pablo Ormeño-Arriagada, Valentina Zúñiga, Carlos Toro, Gastón Márquez, David Araya, Diego Mellado, Carla Taramasco

Förderung

  • Agencia Nacional de Investigación y Desarrollo of Chile
  • Centro de Prevención y Control de Cáncer

Zitieren

APA

Ormeño-Arriagada, P., Zúñiga, V., Toro, C., Márquez, G., Araya, D., Mellado, D., & Taramasco, C. (2026). Improving Diagnostic Sensitivity in Imbalanced Oral Cancer Image Classification: A Comparative Study of CNN and Transformer Architectures. Cancers, 18(18), 2926. https://doi.org/10.3390/cancers18182926

Vancouver

Ormeño-Arriagada P, Zúñiga V, Toro C, Márquez G, Araya D, Mellado D, et al. Improving Diagnostic Sensitivity in Imbalanced Oral Cancer Image Classification: A Comparative Study of CNN and Transformer Architectures. Cancers. 2026;18(18):2926. doi:10.3390/cancers18182926

BibTeX
@article{OrmeoArriagada2026,
  author = {Ormeño-Arriagada, Pablo and Zúñiga, Valentina and Toro, Carlos and Márquez, Gastón and Araya, David and Mellado, Diego and Taramasco, Carla},
  title = {Improving Diagnostic Sensitivity in Imbalanced Oral Cancer Image Classification: A Comparative Study of CNN and Transformer Architectures},
  journal = {Cancers},
  year = {2026},
  volume = {18},
  number = {18},
  pages = {2926},
  doi = {10.3390/cancers18182926}
}
Kostenlos starten →