An Efficient Artificial Rabbits Optimization Based on Mutation Strategy For Skin Cancer Prediction

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Dermatology

Background:

• Accurate melanoma diagnosis is crucial for early detection, but current methods lack sufficient accuracy.
• Deep learning (DL) models, particularly pre-trained ones, offer potential for improving skin cancer detection efficiency.
• Existing DL approaches often require training from scratch, limiting their efficiency.

Purpose of the Study:

• To develop a robust deep learning model for accurate skin cancer detection.
• To enhance feature extraction and selection for improved diagnostic accuracy.
• To validate the model's performance on diverse skin lesion datasets.

Main Methods:

• Utilized MobileNetV3 architecture as a deep learning backbone for feature extraction.
• Introduced the Improved Artificial Rabbits Optimizer (IARO) with Gaussian mutation and crossover for feature selection.
• Validated the model on the PH2, ISIC-2016, and HAM10000 skin lesion datasets.

Main Results:

• Achieved high accuracy rates: 96.79% on PH2, 87.17% on ISIC-2016, and 88.71% on HAM10000 datasets.
• Demonstrated that the IARO algorithm effectively ignores unimportant features, enhancing prediction.
• The combined DL and IARO approach significantly improved skin cancer prediction accuracy.

Conclusions:

• The developed DL model with IARO offers a robust and accurate solution for skin cancer detection.
• This approach shows significant potential for early melanoma diagnosis and improved patient outcomes.
• The study highlights the effectiveness of combining advanced DL architectures with novel optimization algorithms.