Hybrid Deep Learning Strategies Leveraging Cutting-Edge VGG Architectures for Advanced Oral Cancer Diagnosis
Kırklareli Üniversitesi Mühendislik ve Fen Bilimleri Dergisi · 31.12.2025 · 11 (2) · 320-335
DOI: 10.34186/klujes.1834277 ↗Kurz gefasst
Fachartikel in der Zeitschrift Kırklareli Üniversitesi Mühendislik ve Fen Bilimleri Dergisi, online erschienen am 31.12.2025. Verfasst von Cem Baydogan. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 21 Literaturangaben.
Abstract (Originalsprache)
Oral Cancer (OC) has become a critical public health problem, with its increasing prevalence worldwide and high mortality rate when diagnosed late. Tobacco and alcohol use, Human Papilloma Virus (HPV) infections, and various environmental factors play a significant role in the development of the disease. Early detection of the disease significantly improves treatment success and quality of life. However, traditional clinical examinations and manual assessment methods are both time-consuming and can lead to high misclassification rates due to expert dependency. In this study, a deep learning-based hybrid approach for the automatic classification of OC is proposed. The proposed model utilizes different variants of the Visual Geometry Group (VGG) architecture, namely VGG11, VGG13, VGG16, and VGG19, to extract deep features from OC images. The resulting deep features were processed with various classifiers, including Extreme Gradient Boosting (XGBoost), K-Nearest Neighbors (kNN), Support Vector Machines (SVM), Random Forest (RF), and Light Gradient Boosting Machine (LGBM), and a comprehensive experimental analysis was conducted. Experimental findings demonstrate that the VGG19+SVM hybrid model, in particular, demonstrated superior performance, achieving the highest AUC score (0.9144) for inter-class discrimination. Furthermore, the VGG19+LGBM model achieved the highest accuracy rate (0.9158), demonstrating strong classification performance. The results demonstrate that VGG-based deep feature extraction provides high accuracy and strong discrimination in OC classification. These findings demonstrate that the proposed hybrid approach is a reliable diagnostic tool that can be effectively used in clinical decision support systems.
Zitieren
APA
Baydogan, C. (2025). Hybrid Deep Learning Strategies Leveraging Cutting-Edge VGG Architectures for Advanced Oral Cancer Diagnosis. Kırklareli Üniversitesi Mühendislik ve Fen Bilimleri Dergisi, 11(2), 320-335. https://doi.org/10.34186/klujes.1834277
Vancouver
Baydogan C. Hybrid Deep Learning Strategies Leveraging Cutting-Edge VGG Architectures for Advanced Oral Cancer Diagnosis. Kırklareli Üniversitesi Mühendislik ve Fen Bilimleri Dergisi. 2025;11(2):320-335. doi:10.34186/klujes.1834277
BibTeX
@article{Baydogan2025,
author = {Baydogan, Cem},
title = {Hybrid Deep Learning Strategies Leveraging Cutting-Edge VGG Architectures for Advanced Oral Cancer Diagnosis},
journal = {Kırklareli Üniversitesi Mühendislik ve Fen Bilimleri Dergisi},
year = {2025},
volume = {11},
number = {2},
pages = {320--335},
doi = {10.34186/klujes.1834277}
}
EUVIMED