Internet of Things-Assisted Smart Skin Cancer Detection Using Metaheuristics with Deep Learning Model

Area of Science:

• Medical Informatics
• Artificial Intelligence
• Dermatology

Background:

• Skin cancer diagnosis is challenging due to lesion variability and imaging artifacts.
• Deep learning (DL) and Internet of Things (IoT) offer potential for improved early detection.
• Existing methods face difficulties in accurately analyzing diverse skin lesion characteristics.

Purpose of the Study:

• To present an optimal deep learning-based skin cancer detection and classification (ODL-SCDC) methodology within an IoT environment.
• To enhance early skin cancer analysis and monitoring through advanced DL techniques.
• To leverage metaheuristic algorithms for optimizing DL model performance in skin cancer classification.

Main Methods:

• The ODL-SCDC methodology utilizes the Arithmetic Optimization Algorithm (AOA) with EfficientNet for feature extraction.
• A Stacked Denoising Autoencoder (SDAE) model is employed for skin cancer detection.
• The Dragonfly Algorithm (DFA) optimizes the hyperparameters of the SDAE model for improved classification.
• Validation was performed on the ISIC skin lesion database.

Main Results:

• The ODL-SCDC methodology achieved high performance metrics on the ISIC dataset.
• Achieved maximum sensitivity of 97.74%, specificity of 99.71%, and accuracy of 99.55%.
• Demonstrated superior performance compared to existing models in skin cancer classification.

Conclusions:

• The proposed ODL-SCDC model significantly improves skin cancer detection and classification accuracy in an IoT setting.
• This AI-driven approach can aid medical professionals, including dermatologists, in the diagnostic process.
• The integration of DL and IoT presents a promising avenue for advancing dermatological care and early cancer detection.