Advancing Skin Cancer Diagnostics: A Comparative Analysis of Convolutional Neural Network Models for Skin Lesion Classification

Jesse Osuji, Technological University DublinFollow
Tania Malik, Technological University DublinFollow

Abstract

For many years, skin cancer diagnosis and cancer research have been major focal points in the healthcare sector. Accurate classification of different types of skin lesions, including cancerous ones, is essential for proper diagnostics. Many machine learning (ML) techniques and models are used to solve various problems across various domains. One of such techniques is called neural networks. Convolutional Neural Networks (CNNs) are primarily used in computer vision-based problems, allowing computer systems to process and understand images. CNNs offer a potential tool for diagnosing skin lesions from images, occasionally demonstrating superior accuracy compared to trained dermatologists. This project explores various neural network models, including CNN-SVM, DenseNet201, and a sequential model, to diagnose skin lesions. It proposes developing multiple machine learning models using advanced techniques like transfer learning and hybrid models. The performance of these models is compared to determine which architecture yields the best results. Furthermore, this paper compares the results of these models with findings from other research studies to assess the validity of using computer vision for skin cancer detection and to explore whether such models can support or potentially replace dermatologists in clinical practice. The three models focused on this project also differed in accuracy values. Our experimental results show that the CNN-SVM solution is the best-performing of the three models, followed by the DenseNet201 model.

Furthermore, the project has led to the creation of an interactive website where users can upload images of skin lesions and observe how the three different models classify the images. To support early detection and ensure access to professional medical advice, the website includes a built-in hospital locator system, guiding users to nearby healthcare facilities if they have concerns about the classifications provided by the models.

Recommended Citation

Osuji, Jesse and Malik, Tania (2025) "Advancing Skin Cancer Diagnostics: A Comparative Analysis of Convolutional Neural Network Models for Skin Lesion Classification," SURE Journal: Science Undergraduate Research Experience Journal: Vol. 7: Iss. 2, Article 3.
Available at: https://arrow.tudublin.ie/sure_j/vol7/iss2/3