CHASHNIt for enhancing skin disease classification using GAN augmented hybrid model with LIME and SHAP based XAI...

Area of Science:

• Dermatology
• Artificial Intelligence
• Medical Imaging

Background:

• Accurate skin disease classification is crucial for timely diagnosis.
• Deep learning models face challenges like data imbalance and lack of interpretability in medical applications.

Purpose of the Study:

• To introduce CHASHNIt, a novel hybrid deep learning model integrating EfficientNetB7, DenseNet201, and InceptionResNetV2.
• To enhance skin disease classification by addressing data imbalance and improving model interpretability.

Main Methods:

• Utilized Generative Adversarial Networks (GANs) for data augmentation to ensure balanced class representation.
• Implemented sophisticated preprocessing techniques including normalization and feature selection.
• Integrated Explainable AI (XAI) methods like SHAP and LIME for model transparency.

Main Results:

• Achieved high performance metrics: 97.8% accuracy, 98.1% precision, 97.5% recall, 97.6% F1 Score, and 92.3% IoU.
• Outperformed benchmark models including Swin Transformer, ResNet101, and ConvNeXt.
• Ablation studies confirmed the synergistic benefits of the hybrid architecture.

Conclusions:

• CHASHNIt offers an advanced, automated framework for skin disease classification, balancing scalability, accuracy, and explainability.
• The model demonstrates superior performance and transparency compared to existing methods.
• Future work will focus on optimizing computational efficiency for low-resource devices.