Web-based Melanoma Detection

TL;DR

A unified melanoma classification approach that supports 54 combinations of 11 datasets and 24 state-of-the-art deep learning architectures is introduced that enables a fair comparison of 1,296 experiments and results in a lightweight model deployable to the web-based MeshNet architecture named Mela-D.

Abstract

Melanoma is the most aggressive form of skin cancer, and early detection can significantly increase survival rates and prevent cancer spread. However, developing reliable automated detection techniques is difficult due to the lack of standardized datasets and evaluation methods. This study introduces a unified melanoma classification approach that supports 54 combinations of 11 datasets and 24 state-of-the-art deep learning architectures. It enables a fair comparison of 1,296 experiments and results in a lightweight model deployable to the web-based MeshNet architecture named Mela-D. This approach can run up to 33x faster by reducing parameters 24x to yield an analogous 88.8\% accuracy comparable with ResNet50 on previously unseen images. This allows efficient and accurate melanoma detection in real-world settings that can run on consumer-level hardware.

Figures (9)

• Figure 1: Examples of melanoma images from different datasets illustrate the diversity in appearance and challenges for automated detection. Variations in size, shape, color, and texture across datasets highlight the need for a robust and generalizable melanoma detection approach.
• Figure 2: The typical workflow of existing melanoma classifiers is non-deployable for web-based applications as each model is generated from incompatible frameworks or undergoes different preprocessing. This limits practical utility for end-users such as dermatologists.
• Figure 3: The figure is an overview of our proposed melanoma classification framework, which is deployable on the web through a user-friendly interface. It enables end-users to easily classify melanoma images.
• Figure 4: The 11 melanoma databases containing images and labels are combined into 54 different sets. These sets are trained and evaluated with 24 network architectures, yielding 1,296 experiments.
• Figure 5: Mela-D architecture uses Dilated convolutions dilation with increasing dilation factors in each layer to achieve independent computation during inference with lower computational cost.