Skin Cancer Detection Using Hybrid Deep Learning and Clinical Recommendation System

Abstract

Skin cancer is one of the most prevalent forms of cancer
worldwide, with early detection being crucial for effective
treatment. Traditional diagnostic methods rely heavily on
clinical expertise, leading to potential delays and misdiagnoses.
This study introduces a hybrid deep learning approach that
integrates Convolutional Neural Networks (CNNs) with GrayLevel Co-occurrence Matrix (GLCM) texture analysis for
improved multi-class skin cancer detection. The model was
trained on the HAM10000 dataset, consisting of 10,000
dermoscopic images spanning seven skin cancer types.
EfficientNet-B0 was used for deep feature extraction, while
GLCM provided texture-based insights. Our hybrid model
achieved an accuracy of 87.4%, outperforming benchmark
models ResNet50 (82.1%) and VGG16 (79.8%). Furthermore, we
developed a clinical recommendation system that provides
personalized precautionary guidelines, dietary suggestions, and
specialist referrals based on diagnostic results. The integration
of AI-powered diagnosis with real-time patient education
enhances accessibility and decision-making in dermatology. This
study demonstrates the potential of AI-driven diagnostic tools in
reducing biopsy rates and improving early detection, especially
in remote and under-resourced areas.