Derma-HyEmble: A Green AI Framework for Lightweight and Interpretable Dermoscopic Skin Lesion Classification Using Multiscale CLAHE, Handcrafted Features, and Optimized Ensemble Learning
Procedia Computer Science · 2026 · 283 · 2187-2200
DOI: 10.1016/j.procs.2026.06.291 ↗Kurz gefasst
Fachartikel in der Zeitschrift Procedia Computer Science, 2026. Verfasst von Usha Mary Sharma und 2 weiteren Autor:innen. Die Arbeit nennt 54 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Early and precise diagnosis of skin cancer is essential and remains a challenge due to the visual similarities between benign and malignant skin lesions. In this study, we propose Derma-HyEmble, a Green AI framework designed for lightweight, interpretable and resource-efficient classification of dermoscopic skin lesion. Unlike, black-box deep learning models, our framework integrates Multiscale Contrast Limited Adaptive Histogram Equalization (CLAHE) for improved contrast, handcrafted texture features (first-order statistics, GLCM, GLRLM, the Gabor filters and Uniform LBP) and optimized ensemble classifiers. The proposed framework, Derma-HyEmble, was executed on publicly available benchmark ISIC 2018 dataset from the International Skin Imaging Collaboration (ISIC Archive) and the model outperformed the remaining ensemble models by achieving an accuracy of 82% and AUC of 0.90. These results underscores the need for better pre-processing, advanced texture analysis, and effective model tuning to develop a more accurate, sustainable, and reliable automated skin cancer diagnostic system that focuses on early detection thereby reducing the mortality rate. Furthermore, deploying the proposed model in edge or embedded environments can facilitate preventive interventions in skin cancer screening, particularly in rural and resource-limited settings.
Zitieren
APA
Sharma, U. M., Gupta, D., & Sambyo, K. (2026). Derma-HyEmble: A Green AI Framework for Lightweight and Interpretable Dermoscopic Skin Lesion Classification Using Multiscale CLAHE, Handcrafted Features, and Optimized Ensemble Learning. Procedia Computer Science, 283, 2187-2200. https://doi.org/10.1016/j.procs.2026.06.291
Vancouver
Sharma UM, Gupta D, Sambyo K. Derma-HyEmble: A Green AI Framework for Lightweight and Interpretable Dermoscopic Skin Lesion Classification Using Multiscale CLAHE, Handcrafted Features, and Optimized Ensemble Learning. Procedia Computer Science. 2026;283:2187-2200. doi:10.1016/j.procs.2026.06.291
BibTeX
@article{Sharma2026,
author = {Sharma, Usha Mary and Gupta, Deepak and Sambyo, Koj},
title = {Derma-HyEmble: A Green AI Framework for Lightweight and Interpretable Dermoscopic Skin Lesion Classification Using Multiscale CLAHE, Handcrafted Features, and Optimized Ensemble Learning},
journal = {Procedia Computer Science},
year = {2026},
volume = {283},
pages = {2187--2200},
doi = {10.1016/j.procs.2026.06.291}
}
EUVIMED