A ROBUST EXPLAINABLE ARTIFICIAL INTELLIGENCE FRAMEWORK FOR MELANOMA SKIN CANCER DETECTION USING DEEP LEARNING
Keywords:
melanoma classification, explainable artificial intelligence, deep convolutional neural network, transfer learning, Grad-CAM, algorithmic fairness, Fitzpatrick skin type, dermoscopyAbstract
Melanoma is the most lethal form of skin cancer, and its prognosis depends critically on early and accurate diagnosis. Deep learning has achieved dermatologist-level performance on dermoscopic image classification, yet clinical translation remains limited by two persistent barriers: the opacity of "black-box" convolutional models and inconsistent performance across patients with different skin tones. This paper presents an analytical framework for explainable melanoma detection that integrates classical machine learning, custom deep convolutional architectures, and transfer learning within a single comparative pipeline, coupled with post-hoc explainability and fairness auditing stratified by Fitzpatrick skin type. We synthesize methodological evidence from the dermatological AI literature to justify a five-stage pipeline — preprocessing, segmentation, handcrafted and learned feature extraction, multi-model classification (Support Vector Machine with a radial basis function kernel, a custom deep convolutional neural network, and fine-tuned AlexNet, VGG16, and VGG19), and explainability-fairness auditing using Gradient-weighted Class Activation Mapping and SHapley Additive exPlanations. We discuss dataset considerations (ISIC archive, HAM10000, Fitzpatrick17k), evaluation protocols, and the trade-offs between diagnostic accuracy and demographic equity. The analysis indicates that transfer-learning architectures with attention-based explainability modules offer the strongest balance of accuracy, interpretability, and generalizability, but that accuracy gains are not guaranteed to be equitable unless fairness constraints are explicitly incorporated into model design and evaluation. We conclude with recommendations for building clinically trustworthy, explainable, and equitable melanoma-screening systems.
References
Ahmad, N., Shah, J. H., Khan, M. A., Baili, J., Ansari, G. J., Tariq, U., Kim, Y. J., & Cha, J.-H. (2023). A novel framework of multiclass skin lesion recognition from dermoscopic images using deep learning and explainable artificial intelligence. Frontiers in Oncology, 13, 1151257.
Du, S., Hers, B., Bayasi, N., Hamarneh, G., & Garbi, R. (2022). FairDisCo: Fairer AI in dermatology via disentanglement contrastive learning. In Proceedings of the ISIC Skin Image Analysis Workshop, European Conference on Computer Vision (ECCV) Workshops.
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.
Gamage, L., Isuranga, U., De Silva, S., & Meedeniya, D. (2023). Melanoma skin cancer classification with explainability. In Proceedings of the 2023 3rd International Conference on Advanced Research in Computing (ICARC) (pp. 30–35). IEEE.
Groh, M., Harris, C., Soenksen, L., Lau, F., Han, R., Kim, A., Koochek, A., & Badri, O. (2021). Evaluating deep neural networks trained on clinical images in dermatology with the Fitzpatrick 17k dataset. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) (pp. 1820–1828).
Downloads
How to Cite
Issue
Section
License
Copyright (c) 2025 International Journal of Engineering Science & Humanities

This work is licensed under a Creative Commons Attribution 4.0 International License.


