Class Fusion of Support Vector Machines with Deep Learning Features for Oral Cancer Histopathology Classification

This study explores the fusion of support vector machine (SVM) classifiers trained on features extracted from InceptionResNet-v2 and vision transformer (ViT) models to classify histopathological images for detecting dysplasia in oral cancer. The InceptionResNet-v2 with SVM excels at identifying the presence of dysplasia, while the ViT with SVM performs better in detecting its absence. Fusing these models through class selection yielded superior balanced accuracy, precision, sensitivity, and area under the curve compared to individual models and other existing methods. This fusion approach addresses class imbalance effectively, showcasing the advantages of combining complementary classifiers to improve diagnostic accuracy and clinical outcomes.