Skin lesion segmentation using deep learning algorithm with ant colony optimization

Area of Science:

• Artificial Intelligence in Medical Imaging
• Computational Pathology
• Machine Learning for Dermatology

Background:

• Accurate skin lesion segmentation is crucial for diagnosis and surveillance.
• Deep learning models offer advancements in medical image analysis.
• The Hybrid ResUNet model with Ant Colony Optimization (ACO) aims to improve skin lesion diagnosis efficiency.

Purpose of the Study:

• To evaluate the Hybrid ResUNet model's effectiveness in skin lesion classification.
• To assess the impact of ACO on optimizing the Hybrid ResUNet model.
• To bridge the gap between computational efficiency and clinical utility in AI-driven dermatology.

Main Methods:

• A deep learning approach using a Hybrid ResUNet model trained on diverse skin lesion data.
• Hyperparameter optimization of the Hybrid ResUNet model using Ant Colony Optimization (ACO).
• Performance evaluation using accuracy, Dice coefficient, and Jaccard index, compared against ResNet and U-Net.

Main Results:

• The Hybrid ResUNet model achieved high classification accuracy (95.8%), Dice coefficient (93.1%), and Jaccard index (87.5%).
• Demonstrated superior performance compared to existing state-of-the-art methods.
• Showcased exceptional ability in segmenting complex skin lesions, enhancing diagnostic precision.

Conclusions:

• Integrating ResUNet with ACO significantly enhances skin lesion classification accuracy.
• The Hybrid ResUNet model presents a viable strategy for clinical deployment of AI tools.
• Future work includes exploring multi-modal imaging, alternative optimization algorithms, and clinical applicability.