Enhancing Brain Tumor Diagnosis: A CNN-Based Multi-Class Classification Approach
Rohith Krishnan, P G Gokul, Gopikrishnan Sujith, T Anjali, S Abhishek
Abstract
A modified convolutional neural network (CNN) has been used to accomplish the multi-classification of brain tumors utilizing deep learning. The model architecture consists of group normalization, maximum pooling, and many convolutional layers with varying-sized filters. The model’s ability to effectively extract intricate characteristics from brain tumor pictures is facilitated by its design. The model achieved 99.8% training accuracy, 98.6% validation accuracy, and groundbreaking 99.3% final test accuracy, the deep learning model showed outstanding performance when trained and also validated on a dataset of patients with brain tumors. The model was trained using the Adam optimization function for faster learning to improve its performance. Image augmentation techniques were also used widely for improving performance. This model is an effective way to identify certain brain tumors because of its improvements in design and training methods, which made the model better than previous models.