<![CDATA[To build an accurate and reliable clinical decision support system, this study seeks to create a classification system using deep learning as a better approach in the analysis of oral cancer histopathological images. The dataset used consisted of 10,002 images, of which the two more balanced classes were normal oral and oral squamous cell carcinoma. Some pre-trained deep learning architectures are taken as baseline models and then optimized using the Whale Optimization Algorithm to obtain the best hyperparameter configuration. Performance evaluation was carried out on test data using accuracy, precision, recall, F1-score, confusion matrix, and operational efficiency metrics as well as evaluation of a trust-based decision support system with the same mechanism as the reject option system. The models are better optimized and all models show improved performance. From the results of the experiments, the model that was best optimized with an F1-score and an accuracy of 98.73%, and also showed the best performance, was the EfficientNet B3 model. This is accompanied by a stable training process and adequate generalization. Therefore, the model shows results with adequate performance in the coverage range of 0.60 - 0.90 and still provides a reasonable inference time for use in the clinic. These results show that this model has high potential to be integrated with clinical decision support systems. Therefore, this model can be used as a diagnostic tool in clinics that is more accurate and ensures consistency in each clinical practice and can also build a better diagnostic decision support system.]]>
Copyrights © 2026
