MobileNet-V2: An Enhanced Skin Disease Classification by Attention and Multi-Scale Features

Area of Science:

• Dermatology
• Artificial Intelligence
• Medical Imaging

Background:

• Skin diseases are increasingly prevalent, demanding improved diagnostic methods.
• Accurate and early classification of skin conditions is crucial for effective treatment.

Purpose of the Study:

• To develop and evaluate a novel deep learning model for enhanced skin disease classification.
• To improve the accuracy and efficiency of dermatological diagnoses using AI.

Main Methods:

• A deep learning architecture combining MobileNet-V2, Squeeze-and-Excitation (SE) blocks, Atrous Spatial Pyramid Pooling (ASPP), and Channel Attention Mechanism was designed.
• The model was trained and validated on four diverse datasets: PH2, HAM10000, DermNet, and ISIC.
• Image preprocessing techniques including resizing and normalization were applied to optimize performance.

Main Results:

• The proposed model achieved an overall accuracy of 98.6% in skin disease classification.
• It outperformed traditional machine learning models and demonstrated competitive performance against state-of-the-art methods.
• Attention mechanisms significantly enhanced feature extraction and discriminative power.

Conclusions:

• The developed deep learning model shows significant potential as a valuable tool for dermatologists in early skin disease classification.
• The model's high accuracy and efficiency suggest practical applicability in clinical dermatology.
• Further research is recommended to explore limitations and expand practical implementation.