Improving Diagnostic Sensitivity in Imbalanced Oral Cancer Image Classification: A Comparative Study of CNN…

AI / informaticsCancers· Sep 2026

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

Pablo Ormeño-Arriagada, Valentina Zúñiga, Carlos Toro … Carla Taramasco

Medical Imaging AIAI / informaticsDiagnosis

EfficientNet with data augmentation achieved high sensitivity for oral cancer detection.

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

Pablo Ormeño-Arriagada … Carla Taramasco

Cancers · Sep 2026

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.

n = 3000 mobile-acquired oral cavity

0.817

Results

0.817

EfficientNet detected a high proportion of actual oral cancer cases

n = 3000 mobile-acquired oral cavity

More 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).

In contrast, random undersampling consistently yielded lower descriptive performance across architectures.

“

Conclusion · 1 of 2

These findings indicate that augmentation-based imbalance mitigation is associated with improved multiclass and oral cancer specific performance while maintaining competitive overall discrimination.

Conclusion · 2 of 2

The results provide practical guidance for developing clinically robust artificial intelligence systems for early oral cancer detection in imbalanced real-world datasets.

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

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

Related papers