Performance Evaluation of Deep Learning Models for Detection of Brain Tumors from CT Scans

Abstract

With brain tumors being a leading cause of death in the world, this paper explores the use of deep learning algorithms in medical imaging tasks specifically the detection of brain tumors from CT scans. The study employs a dataset comprising 230 images of normal and brain tumor patients which are used to train and evaluate the performance of seven deep learning models including six pre-trained models: ResNet-50, MobileNet-V2, InceptionNet, VGG-16, Xception, and DenseNet along with a custom baseline CNN model. Transfer learning is used for pre-trained model detection of brain tumors from CT scans while the custom CNN model is compared against pre-trained models with specific hyperparameters. Results indicate that MobileNet-V2 outperforms other models with a test accuracy of 98%, making it a promising candidate for efficient brain tumor detection. The baseline CNN model and InceptionNet closely followed, achieving a 97% test accuracy. The study shows the innovative potential of using deep learning in medical imaging and provides valuable insights for optimization of deep learning models for brain tumor detection addressing a significant need in medical diagnostics.