Contrastive Learning for Dermatological Imaging: Progress, Limitations, and Future Prospects in Smart Device Engineering and AI-Driven Skin Lesion Analysis

Akinrotimi Akinyemi Omololu, Omotosho Israel Oluwabusayo, Owolabi Olugbenga Olayinka, Omude Paul Onome

Abstract


The integration of artificial intelligence (AI) and mobile imaging has transformed dermatology, enabling early, accessible skin disease screening through smartphones and portable dermatoscopes. Yet, the dependence of deep learning models on large annotated datasets restricts their scalability and fairness across diverse populations. Contrastive learning, a branch of self-supervised representation learning, has emerged as a promising alternative by learning discriminative image features without extensive labeling. This paper reviews developments in contrastive learning for dermatological imaging, emphasizing algorithmic innovations, device-level engineering, and fairness considerations. Drawing from recent literature between 2019 and 2025, the study identifies how contrastive pretraining improves lesion classification and segmentation while reducing annotation costs. It also highlights persistent challenges related to hardware variability, underrepresentation of darker skin tones, and the computational limitations of on-device models. The paper concludes with future research directions for integrating contrastive pipelines with smart imaging hardware, explainable AI (XAI), and equitable data governance frameworks to achieve trustworthy and accessible dermatological diagnostics. 


Full Text:

PDF

References


Akram, S., Khan, M., & Javed, A. (2023). Hybrid CNN approaches for lesion recognition in skin cancer diagnosis. Journal of Medical Imaging and Health Informatics, 13(2), 187–198.

Alzahrani, S. (2024). SkinLiTE: Supervised contrastive learning for imbalanced skin lesion datasets. Biomedical Signal Processing and Control, 93, 106054.

Azizi, S., Kornblith, S., Norouzi, M., Abid, A., Lee, H., & Hinton, G. (2021). Big self-supervised models advance medical image classification. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 3478–3488.

Brinker, T. J., Hekler, A., & von Kalle, C. (2022). Deep neural networks for skin cancer classification: Current state and future directions. European Journal of Cancer, 166, 181–190.

Celebi, M. E., Codella, N. C. F., & Halpern, A. (2015). Dermoscopy image analysis: Overview and future directions. IEEE Journal of Biomedical and Health Informatics, 19(2), 543–546.

Chen, T., Kornblith, S., Norouzi, M., & Hinton, G. (2020). A simple framework for contrastive learning of visual representations. Proceedings of the 37th International Conference on Machine Learning (ICML), 1597–1607.

Christopoulos, M., Tzoumas, I., & Kyriakopoulos, K. (2025). SLIMP: Multi-modal contrastive learning for skin lesion diagnosis. Medical Image Analysis, 93, 103172.

Doersch, C., Gupta, A., & Efros, A. A. (2015). Unsupervised visual representation learning by context prediction. IEEE International Conference on Computer Vision (ICCV), 1422–1430.

Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017). Dermatologist-level classification of skin cancer with deep neural networks. Nature, 542(7639), 115–118.

Gharawi, M. A., El-Gamal, F. E., & Basha, S. H. (2023). Self-supervised segmentation of skin lesions using generative models. Mathematics, 11(18), 3805.

Goyal, P., Zhai, X., Misra, I., & He, K. (2021). Self-supervised pretraining improves dermatology image classification on limited datasets. IEEE Transactions on Medical Imaging, 40(10), 2746–2757.

Hasan, R., Rezk, E., & Wong, A. (2024). Beyond accuracy: Fairness-aware skin lesion analysis using cost-sensitive learning. Journal of Biomedical Informatics, 152, 104714.

He, K., Fan, H., Wu, Y., Xie, S., & Girshick, R. (2020). Momentum contrast for unsupervised visual representation learning. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 9729–9738.

Kundu, A., Sharma, S., & Banerjee, P. (2022). On-device self-supervised learning for efficient skin lesion recognition. Sensors, 22(14), 5411.

Nagpal, K., Shah, S., & Esteva, A. (2023). Towards fair dermatological AI: Skin-tone-balanced contrastive learning. Journal of Biomedical Informatics, 146, 104655.

Rezk, E., Wong, A., & Hussain, S. (2023). Addressing representation bias in dermatology AI through dataset diversification. Artificial Intelligence in Medicine, 139, 102473.

Science Advances DDI Dataset Study. (2023). Diversity in dermatology images: Addressing racial bias in AI datasets. Science Advances, 9(24), eabq6147.

Sun, Z., Xu, J., & Li, T. (2023). Transformer architectures in dermatological imaging: A review. Computers in Biology and Medicine, 163, 107136.

Tschandl, P., Rinner, C., & Argenziano, G. (2020). Data-driven approaches to detecting melanoma using deep learning and smartphone imagery. The Lancet Digital Health, 2(10), e512–e520.

Zhang, Y., Liu, R., & Zhao, D. (2024). Multimodal contrastive learning for dermatological diagnosis integrating images and clinical text. Artificial Intelligence in Medicine, 152, 102560.


Refbacks

  • There are currently no refbacks.