Citation

Ⅴ., Aravinda C. and Emerson Raja, Joseph and Alasmari, Sultan (2025) Optimized DenseNet121 and Quantum PennyLane Fusion for Explainable Skin Disease Recognition and Classification. IEEE Access. p. 1. ISSN 2169-3536

Text 9.pdf - Submitted Version Restricted to Repository staff only Download (8MB)

Abstract

The accurate diagnosis of skin diseases, particularly in distinguishing between visually similar conditions such as Measles, Monkeypox, Chickenpox, and Normal skin states, remains a critical challenge in clinical dermatology. To address this, we propose a novel hybrid framework that seamlessly integrates classical convolutional neural networks (CNNs) with quantum computing principles for enhanced multi-class classification. Our approach utilizes DenseNet121 as the backbone feature extractor, which is then coupled with a quantum-enhanced classifier implemented using the PennyLane framework. This fusion not only improves classification accuracy but also ensures interpretability, a key requirement in medical diagnostics. To further enhance transparency and trust, we incorporate multiple Explainable AI (XAI) techniques, including SHAP, LIME, Grad-CAM, and XRAI, to provide both pixel-level visualizations and region-based annotations through bounding boxes that highlight infection zones. An embedding-based image recognition module is also introduced to retrieve visually similar cases from a database, aiding clinicians in comparative analysis and decision-making. We employ Optuna-based hyperparameter optimization to fine-tune the CNN components, achieving a validation accuracy exceeding 91.5% on a dataset of 770 real-world skin lesion images. The robustness of our pipeline is validated through rigorous random sampling and multi-class evaluation. This work demonstrates the potential of combining quantum machine learning with interpretable CNNs to advance medical image analysis in dermatology, offering a system that balances performance, transparency, and clinical utility.

Item Type:	Article
Uncontrolled Keywords:	Skin Disease Classification, Quantum Machine Learning, DenseNet121, Explainable AI, SHAP, LIME, Grad-CAM, XRAI, Clinical Decision Support, Hybrid Framework.
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5101-6720 Telecommunication. Including telegraphy, telephone, radio, radar, television
Divisions:	Faculty of Engineering and Technology (FET)
Depositing User:	Ms Suzilawati Abu Samah
Date Deposited:	30 Sep 2025 07:25
Last Modified:	05 Oct 2025 11:22
URII:	http://shdl.mmu.edu.my/id/eprint/14594

Downloads

Edit (login required)