Gastric Cancer Detection Using Vision Transformers: Feature Extraction and Classical Classifier Performance Evaluation

Authors : Uğur Demiroğlu, Bilal Şenol

Pages : 92-100

Doi:10.53070/bbd.1652603

View : 33 | Download : 22

Publication Date : 2025-06-01

Article Type : Research Paper

Abstract :Gastric cancer remains one of the most prevalent and deadly forms of cancer worldwide, necessitating advanced computational methods for early and accurate detection. This study explores the effectiveness of Vision Transformers (ViTs) in feature extraction for gastric cancer image classification. A publicly available dataset was sourced from Kaggle, consisting of three categories: Normal, Stage-1, and Stage-2 gastric cancer images. Using a pre-trained Google Vision Transformer model, 1000 deep features were extracted from the fully connected head layer without additional training. These extracted features were then used as input for various classical classifiers, including Support Vector Machines (SVM), k-Nearest Neighbors (KNN), Decision Trees, and Random Forest, to evaluate their classification performance. The effectiveness of these classifiers was assessed based on classification accuracies. Comparative analysis of classifier results demonstrated the impact of feature extraction via Vision Transformers on improving gastric cancer detection. The findings highlight the potential of Vision Transformers in medical image analysis and emphasize the role of feature-based classification in aiding early diagnosis. This study provides insights into the applicability of deep learning models in feature extraction and their integration with traditional machine learning classifiers for medical diagnostics.
Keywords : Mide Kanseri, Görüntü Sınıflandırma, Görme Dönüştürücüler, Özellik Çıkarımı

ORIGINAL ARTICLE URL