Application of Deep Learning to CT Scan Image-Based Brain Tumor Classification System

Abstract

Brain tumor detection and classification is an important challenge in the medical field that requires a fast and accurate solution. In this study, we propose the application of Deep Learning to a CT scan image-based brain tumor classification system. We use DenseNet as the base model and train it using CT scan image dataset to distinguish between positive class (brain tumor) and negative class (no brain tumor). In addition, we conducted a series of experiments with varying number of epochs to understand the development of the model's performance during the training process. The evaluation results show that our model achieved the highest accuracy of 0.92 at epoch 100, with precision, recall, and F1 score stabilizing at high values. Although there are fluctuations in performance at some stages of training, the model still shows stable performance overall. These findings suggest that the application of Deep Learning can be an effective tool in supporting the diagnosis of complex brain diseases.