Skin Cancer Detection Based on Deep Learning.

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Dermatology

Background:

• Conventional skin cancer detection relies on visual inspection and biopsy.
• Deep convolutional neural networks (CNNs) show promise in automated classification.
• Clinical adoption of automated systems for skin cancer detection remains limited.

Purpose of the Study:

• To propose a deep learning (DL) based method for enhanced skin cancer detection in lesion images.
• To improve diagnostic accuracy for physicians by integrating patient data.
• To develop a clinically viable system for early skin cancer identification.

Main Methods:

• A novel DL model utilizing Inception-ResNet-v2 CNN was developed.
• The model incorporated both lesion images and patient metadata (anatomical site, age, gender).
• The model was trained and evaluated on a large dataset of dermoscopic images.

Main Results:

• The proposed DL model achieved high accuracy in discriminating skin conditions and classifying benign vs. malignant lesions.
• Incorporating patient metadata improved classification accuracy by at least 5%.
• The model demonstrated 89.3% accuracy for 4 major skin conditions and 94.5% for benign vs. malignant classification.

Conclusions:

• The study highlights the efficacy of the proposed DL approach for skin cancer detection.
• Integrating patient metadata significantly enhances diagnostic performance.
• The findings support the development of more accurate and clinically trusted automated skin cancer detection systems.