MUET logo

A Hybrid U-Net and EfficientNet-B0 Framework for Skin Lesion Segmentation and Classification with Grad-CAM Explainability

Authors

  • Saad Rehman Lodhi

    PhD Student, HITEC University, Taxila, Pakistan
    Author

  • Sobia Arshad

    Department of Computer Engineering, HITEC University, Taxila, Pakistan
    Author

DOI:

https://doi.org/10.22581/0868

Keywords:

Skin Lesion, Melasma, Dice-score, Segmentation, Classification, IoU

Abstract

The aggressive nature of skin cancer requires timely and accurate diagnosis to improve patient outcomes. This study proposes an integrated deep learning framework that performs skin lesion segmentation and multi-class classification within a unified pipeline to streamline diagnosis and enhance predictive performance. The framework combines U-Net for lesion segmentation with EfficientNet-B0 for diagnostic classification, fusing dermoscopic image features with structured patient metadata (age, sex, and anatomical site). The segmentation model is evaluated on the ISIC 2018 dataset, whereas the classification model is trained and validated on the HAM10000 dataset. At a 0.55 threshold, the segmentation model achieved an Intersection over Union (IoU) of 0.9327 and a Dice score of 0.9652, demonstrating precise boundary delineation. The classification model achieved 86.0\% accuracy, a macro AUC of 0.9773, and a macro F1-score of 0.7753, indicating strong multi-class generalization. Furthermore, Gradient-weighted Class Activation Mapping (Grad-CAM) visualizations improved model interpretability by highlighting the spatial regions driving diagnostic predictions, particularly in misclassified cases.

Downloads

References

[1] A. Abdel-Wahab, V. Pulipati, S. Joshi, R. Srikanteswara, S. Menon, and P. Dutta, “Deep Learning Framework for Breast Cancer Detection and Segmentation Using EfficientNet and U-Net with Hyperparameter Optimization,” SN Computer Science, vol. 6, no. 5, p. 528, 2025.

[2] I. Matas, C. Serrano, F. Silva, A. Serrano, T. Toledo-Pastrana, and B. Acha, “Clinically inspired enhanced explainability and interpretability of an AI tool for BCC diagnosis based on expert annotation,” arXiv preprint arXiv:2407.00104, 2024.

[3] N. Rane, “Transformers for medical image analysis: Applications, challenges, and future scope,” Challenges and Future Scope (November 2, 2023), 2023.

[4] D. Wen, S. M. Khan, A. J. Xu, H. Ibrahim, L. Smith, J. Caballero, L. Zepeda, C. de Blas Perez, A. K. Denniston, X. Liu, and R. N. Matin, “Characteristics of publicly available skin cancer image datasets: A systematic review,” The Lancet Digital Health, vol. 4, no. 1, pp. e64–e74, 2022.

[5] S. S. I. Badhon, S. A. Khushbu, N. C. Saha, A. H. Anik, M. A. Ali, and K. T. Hossain, “Explainable AI for Skin Disease Classification Using Grad-CAM and Transfer Learning to Identify Contours,” 2024.

[6] M. A. Abdou, “Literature review: Efficient deep neural network techniques for medical image analysis,” Neural Computing and Applications, vol. 34, no. 8, pp. 5791–5812, 2022.

[7] T. P. Fowdur and L. Ajageer, “A hybrid online learning-based predictive maintenance system for Industry 4.0,” Computing, vol. 107, no. 6, p. 142, 2025.

[8] G. Dheepak and D. Vaishali, “Brain tumor classification: a novel approach integrating GLCM, LBP, and composite features,” Frontiers in oncology, vol. 13, p. 1248452, 2024.

[9] M. Chowdhury, S. R. Shrima, and M. S. Islam, “Comparative analysis of deep learning architectures for multi-class mineral classification: a study using EfficientNet and ResNet models,” Earth Science Informatics, vol. 18, no. 3, p. 485, 2025.

[10] A. M. Zbrzezny and T. Krzywicki, “Artificial intelligence in dermatology: A review of methods, clinical applications, and perspectives,” Applied Sciences, vol. 15, no. 14, p. 7856, 2025.

[11] S. Chen, Y. Zhang, and Q. Yang, “Multi-task learning in natural language processing: An overview,” ACM Computing Surveys, vol. 56, no. 12, pp. 1–32, 2024.

[12] S. Kumar, A. A. Abdelhamid, and Z. Tarek, “Visualizing the unseen: Exploring GRAD-CAM for interpreting convolutional image classifiers,” J. Full Length Artic, vol. 4, pp. 34–42, 2023.

[13] M. P. Bates, “Visualizing deep learning decisions: grad-CAM-based explainable AI for medical image analysis,” Journal of Scalable Data Engineering and Intelligent Computing, pp. 26–33, 2024.

[14] B. Oliveira, H. R. Torres, P. Morais, F. Veloso, A. L. Baptista, J. C. Fonseca, and J. L. Vilaça, “A multi-task convolutional neural network for classification and segmentation of chronic venous disorders,” Scientific Reports, vol. 13, no. 1, p. 761, 2023.

[15] T. M. Alam, K. Shaukat, W. A. Khan, I. A. Hameed, L. A. Almuqren, M. A. Raza, M. Aslam, and S. Luo, “An efficient deep learning-based skin cancer classifier for an imbalanced dataset,” Diagnostics, vol. 12, no. 9, p. 2115, 2022.

[16] M. E. Atiq and S. A. Fattah, “Towards Explainable Skin Cancer Classification: A DualNetwork Attention Model with Lesion Segmentation and Clinical Metadata Fusion.” arXiv preprint arXiv:2510.17773, 2025.

[17] A. Akram, J. Rashid, M. A. Jaffar, M. Faheem, and R. U. Amin, “Segmentation and classification of skin lesions using a hybrid deep learning method in the Internet of Medical Things,” Skin Research and Technology, vol. 29, no. 11, p. e13524, 2023.

[18] A. Patil, A. Mehto, and S. Nalband, “Enhancing skin lesion diagnosis with data augmentation techniques: a review of the state-of-the-art,” Multimedia Tools and Applications, vol. 84, no. 22, pp. 25325–25364, 2025.

[19] L. Wang and H. Xiao, “A grid-level segmentation model based on an encoder-decoder structure with multi-source features for crop lodging detection,” Applied Soft Computing, vol. 151, p. 111113, 2024.

[20] X. Polisi, A. Uka, D. Avdiu, and D. Karras, “Efficient and accurate cell image segmentation through optimizer and loss function fine-tuning with quantization and pruning,” Edelweiss Appl. Sci. Technol, vol. 9, pp. 1675–1691, 2025.

[21] M. Alruwaili and M. Mohamed, “An integrated deep learning model with EfficientNet and ResNet for accurate multi-class skin disease classification,” Diagnostics, vol. 15, no. 5, p. 551, 2025.

[22] S. M. Hosseini and M. S. Baghshah, “Dilated balanced cross-entropy loss for medical image segmentation,” BMC Medical Imaging, 2026.

[23] M. W. L. Zheng, “Assessment of Consistency Detection in Mandibular First Premolars Root Canal Morphology Using U-net Model: Evaluation With Dice Coefficient and Intersection Over Union – A Pilot Study,” Master’s thesis, University of Malaya (Malaysia), 2024.

[24] D. Müller, I. Soto-Rey, and F. Kramer, “Towards a guideline for evaluation metrics in medical image segmentation,” BMC research notes, vol. 15, no. 1, p. 210, 2022.

[25] A. M. Carrington, D. G. Manuel, P. W. Fieguth, T. Ramsay, V. Osmani, B. Wernly, C. Bennett, S. Hawken, O. Magwood, Y. Sheikh, and M. McInnes, “Deep ROC analysis and AUC as balanced average accuracy, for improved classifier selection, audit, and explanation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 1, pp. 329–341, 2022.

[26] T. Hussain and H. Shouno, “Explainable deep learning approach for multi-class brain magnetic resonance imaging tumor classification and localization using gradient-weighted class activation mapping,” Information, vol. 14, no. 12, p. 642, 2023.

[27] Y. Weng, M. Han, H. He, M. Li, L. Yao, X. Chang, and B. Zhuang, “Mask propagation for efficient video semantic segmentation,” Advances in Neural Information Processing Systems, vol. 36, pp. 7170– 7183, 2023.

[28] P. K. Parhi, S. Mohanty, P. K. Patra, and A. Mohanty, “Deep Learning Techniques to Detect Brain Tumors Using the EfficientNet-B0 CNN Architecture,” in Artificial Intelligence in Oncology: Cancer Diagnosis and Treatment, Medical Imaging, and Personalized Medicine, pp. 115–126, Cham: Springer Nature Switzerland, 2025.

[29] S. R. Lodhi, S. Arshad, and J. Ali Khan, “A Semi-Supervised Approach utilizing Deep Learning Pseudo-Label Refinement and Confidence Filtering for Enhanced Image Segmentation: An application towards Skin Lesion Analysis,” MURJET, Pakistan, pp. 234–243, 2026.

Published

2026-10-01

Issue

Vol. 45 No. 4 (2026): October Issue

Section

Articles

License

Copyright (c) 2026 Mehran University Research Journal of Engineering and Technology

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

How to Cite

A Hybrid U-Net and EfficientNet-B0 Framework for Skin Lesion Segmentation and Classification with Grad-CAM Explainability. (2026). Mehran University Research Journal of Engineering and Technology, 45(4), 204-214. https://doi.org/10.22581/0868

Similar Articles

11-16 of 16

You may also start an advanced similarity search for this article.