An Adaptive Generative 3D VNet Model for Enhanced Monkeypox Lesion Classification Using Deep Learning and Augmented...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Dermatology

Background:

• Monkeypox incidence is rising, necessitating accurate diagnostic tools for effective treatment.
• Limited labeled data presents a significant challenge for developing robust disease detection models.

Purpose of the Study:

• To design and evaluate an effective monkeypox detection and classification model using deep learning.
• To address the challenge of limited labeled data by generating synthetic augmented images.

Main Methods:

• A novel Adaptive Generative 3D VNet model integrating data augmentation, deep learning, and adaptive fusion.
• Utilizing an Adaptive Generative Network for data augmentation (cropping, rotation, flipping) to increase dataset diversity.
• Employing a 3D VNet for volumetric image processing to capture spatial lesion relationships and an adaptive fusion layer for combining predictions.

Main Results:

• The Adaptive Generative 3D VNet model demonstrated superior performance compared to traditional 2D models.
• Achieved high classification accuracy (98.8%) and precision (98.5%) on the Monkeypox Skin Lesion Dataset.
• Significantly improved classification robustness, particularly with limited labeled data.

Conclusions:

• The proposed Adaptive Generative 3D VNet model offers a robust and accurate solution for monkeypox lesion classification.
• Effective mitigation of limited data challenges through synthetic data generation and adaptive fusion.
• The model shows promise for enhancing diagnostic capabilities in clinical settings for monkeypox.