Oral cancer is a major health problem in the world, with significant morbidity and mortality rates that are mostly because they are not diagnosed early enough or because the lesions are often not detected at an early stage. Intraoral imaging has become a convenient and non-invasive screening tool, but diagnostic performance is highly reliant on clinician experience, and there is a wide range of heterogeneous imaging conditions across devices and clinical environments. This paper will attempt to overcome these shortcomings by introducing OralLesionNet, a dense multi-scale convolutional neural network (CNN) architecture that can effectively detect oral cancer using intraoral images with a higher degree of domain generalisation. The proposed model incorporates a dense multi-scale backbone to achieve fine-grained textural features, as well as coarse morphological features, to facilitate a better ability to differentiate benign and malignant lesions. The domain generalisation module is included in the training to promote the learning of invariant features to mitigate the performance decline due to inter-centre and device-related variations. The model was tested on a free, publicly available intraoral image dataset of benign and malignant cases that were clinically verified. To obtain a credible assessment, extensive preprocessing, augmentation, and train-validation-test splitting were used. The experimental outcomes indicate that OralLesionNet is more accurate, more precise, more recalls, more F1-score, and more area under the ROC curve (AUC) than baseline CNN, ResNet and DenseNet architectures. Ablation experiments validate the claims that multi-scale design and domain generalisation strategy play an important role in enhancing performance. The sensitivity obtained with the proposed framework is very high; this is very important in screening applications where false negatives should be reduced to a minimum. These results suggest that oral lesion classification can be made more robust and generalizable with dense multi-scale representation and domain-aware learning. OralLesionNet has a good potential of becoming a computer-aided diagnostic support tool that could enhance the ability to detect at earlier stages, minimise the inter-observer variability, and expand the screening to resource-limited settings.

Oral cancer, Deep learning, Dense multi-scale CNN, Intraoral image classification, Domain generalization, Computer-aided diagnosis

H. Rumgay et al., “Global incidence of lip, oral cavity, and pharyngeal cancers by subsite in 2022,” CA: A Cancer Journal for Clinicians, vol. 76, no. 1, Dec. 2025, doi: https://doi.org/10.3322/caac.70048.

C. Rivera, “Essentials of oral cancer,” International Journal of Clinical and Experimental Pathology, vol. 8, no. 9, p. 11884, Sep. 2015, Available: https://pmc.ncbi.nlm.nih.gov/articles/PMC4637760/

O. Di Fede et al., “Automated Detection of Oral Malignant Lesions Using Deep Learning: Scoping Review and Meta‐Analysis,” Oral Diseases, Nov. 2024, doi: https://doi.org/10.1111/odi.15188.

N. Sharma et al., “Automated medical image segmentation techniques,” Journal of Medical Physics, vol. 35, no. 1, p. 3, 2010, doi: https://doi.org/10.4103/0971-6203.58777.

Andreea Ionela Dumachi and Cătălin Buiu, “Applications of Machine Learning in Cancer Imaging: A Review of Diagnostic Methods for Six Major Cancer Types,” Electronics, vol. 13, no. 23, pp. 4697–4697, Nov. 2024, doi: https://doi.org/10.3390/electronics13234697.

A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks,” Communications of the ACM, vol. 60, no. 6, pp. 84–90, May 2012.

Babina Banjara, “Deep Residual Learning for Image Recognition (ResNet Explained),” Medium, Feb. 07, 2023. https://medium.com/@bbabina/deep-residual-learning-for-image-recognition-resnet-explained-d2b3c06f7c0a (accessed Feb. 28, 2026).

T. Zhou, X. Ye, H. Lu, X. Zheng, S. Qiu, and Y. Liu, “Dense Convolutional Network and Its Application in Medical Image Analysis,” BioMed Research International, vol. 2022, pp. 1–22, Apr. 2022, doi: https://doi.org/10.1155/2022/2384830.

D. Raval and J. N. Undavia, “A Comprehensive assessment of Convolutional Neural Networks for skin and oral cancer detection using medical images,” Healthcare analytics, vol. 3, pp. 100199–100199, Nov. 2023, doi: https://doi.org/10.1016/j.health.2023.100199.

G. Tanriver, M. Soluk Tekkesin, and O. Ergen, “Automated Detection and Classification of Oral Lesions Using Deep Learning to Detect Oral Potentially Malignant Disorders,” Cancers, vol. 13, no. 11, p. 2766, Jun. 2021, doi: https://doi.org/10.3390/cancers13112766.

C. Görürgöz et al., “Performance of a convolutional neural network algorithm for tooth detection and numbering on periapical radiographs,” Dentomaxillofacial Radiology, vol. 51, no. 3, Mar. 2022, doi: https://doi.org/10.1259/dmfr.20210246.

S. Zhang, C. Liu, Y. Zhang, S. Liu, and X. Wang, “Multi-Scale Feature Learning Convolutional Neural Network for Image Denoising,” Sensors, vol. 23, no. 18, p. 7713, Sep. 2023, doi: https://doi.org/10.3390/s23187713.

Y. Du, X. Chen, and Y. Fu, “Multiscale transformers and multi-attention mechanism networks for pathological nuclei segmentation,” Scientific Reports, vol. 15, no. 1, Apr. 2025, doi: https://doi.org/10.1038/s41598-025-90397-2.

S. Matta et al., “A systematic review of generalization research in medical image classification,” Computers in Biology and Medicine, vol. 183, p. 109256, Oct. 2024, doi: https://doi.org/10.1016/j.compbiomed.2024.109256.

M. Xu, S. Yoon, A. Fuentes, and D. S. Park, “A Comprehensive Survey of Image Augmentation Techniques for Deep Learning,” Pattern Recognition, vol. 137, p. 109347, May 2023, doi: https://doi.org/10.1016/j.patcog.2023.109347.

T. Fawcett, “An introduction to ROC analysis,” Pattern Recognition Letters, vol. 27, no. 8, pp. 861–874, Jun. 2006, doi: https://doi.org/10.1016/j.patrec.2005.10.010.

D. Wei, “Demystifying the Adam Optimizer in Machine Learning,” Medium, Jan. 30, 2024. https://medium.com/@weidagang/demystifying-the-adam-optimizer-in-machine-learning-4401d162cb9e

J. Hu, L. Shen, and G. Sun, “Squeeze-and-Excitation Networks,” 2018. Available: https://openaccess.thecvf.com/content_cvpr_2018/papers/Hu_Squeeze-and-Excitation_Networks_CVPR_2018_paper.pdf

J. Kijowska, J. Grzegorczyk, K. Gliwa, Aleksandra Jędras, and M. Sitarz, “Epidemiology, Diagnostics, and Therapy of Oral Cancer—Update Review,” Cancers, vol. 16, no. 18, pp. 3156–3156, Sep. 2024, doi: https://doi.org/10.3390/cancers16183156.

National Cancer Institute, “Cancer of the Oral Cavity and Pharynx - Cancer Stat Facts,” SEER, 2025. https://seer.cancer.gov/statfacts/html/oralcav.html

J. Xu et al., “Deep reconstruction-recoding network for unsupervised domain adaptation and multi-center generalization in colonoscopy polyp detection,” Computer methods and programs in biomedicine, vol. 214, p. 106576, Feb. 2022, doi: https://doi.org/10.1016/j.cmpb.2021.106576.

C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens, and Z. Wojna, “Rethinking the Inception Architecture for Computer Vision,” Thecvf.com, pp. 2818–2826, 2016, Available: https://openaccess.thecvf.com/content_cvpr_2016/html/Szegedy_Rethinking_the_Inception_CVPR_2016_paper.html

N. Al-Rawi et al., “The Effectiveness of Artificial Intelligence in Detection of Oral Cancer,” International Dental Journal, vol. 72, no. 4, pp. 436–447, Aug. 2022, doi: https://doi.org/10.1016/j.identj.2022.03.001.

A. Patel et al., “Attention-guided convolutional network for bias-mitigated and interpretable oral lesion classification,” Scientific reports, vol. 14, no. 1, p. 31700, 2024, doi: https://doi.org/10.1038/s41598-024-81724-0.

P. Wighton, T. K. Lee, G. Mori, H. Lui, and M. S. Atkins, “Conditional Random Fields and Supervised Learning in Automated Skin Lesion Diagnosis,” International Journal of Biomedical Imaging, vol. 2011, no. 6833, p. 846312, Jan. 2011, doi: https://doi.org/10.1155/2011/846312.

Will, “Wrapping Up CNN Models: Shifting Focus to Attention-Based Architectures,” Medium, Oct. 25, 2023. https://blog.gopenai.com/wrapping-up-cnn-models-shifting-focus-to-attention-based-architectures-5029b87034d9 (accessed Feb. 28, 2026).

C.-L. Fan, “Multiscale Feature Extraction by Using Convolutional Neural Network: Extraction of Objects from Multiresolution Images of Urban Areas,” ISPRS International Journal of Geo-Information, vol. 13, no. 1, p. 5, Jan. 2024, doi: https://doi.org/10.3390/ijgi13010005.

Y. Wang, S. Hu, G. Wang, C. Chen, and Z. Pan, “Multi-scale dilated convolution of convolutional neural network for crowd counting,” Multimedia Tools and Applications, vol. 79, no. 1–2, pp. 1057–1073, Oct. 2019, doi: https://doi.org/10.1007/s11042-019-08208-6.

N. Al-Rawi et al., “The Effectiveness of Artificial Intelligence in Detection of Oral Cancer,” International Dental Journal, vol. 72, no. 4, pp. 436–447, Aug. 2022, doi: https://doi.org/10.1016/j.identj.2022.03.001