A fine tune robust transfer learning based approach for brain tumor detection using VGG-16

Abstract

Brain tumor recognition by magnetic resonance imaging (MRI) is crucial because it improves survival rates and allows them to plan treatments accordingly. An accumulation of abnormal cells known as a brain tumor can spread to nearby tissues and endanger the patient. Magnetic resonance imagery is the primary imaging technique which determines the extent of brain tumors. Deep learning techniques rapidly grew in computer vision due to ample data for model training and improved designs on applications. MRI has shown promising results when using deep learning approaches to identify and classify brain tumors. This study uses MRI data and a convolutional neural network (CNN) to create a reliable transfer learning model that classifies tumors under four classes. Brain tumors' unwanted parts are excised, the quality is improved, and the cancer is coloured. By eliminating artefacts, decreasing noise, and boosting the image. The number of MRI images has increased using two augmentation techniques. A number of CNN architectures, including VGG19, VGG16, MobileNet, InceptionV3, and MobileNetV2 analyzed the augmented dataset. Where VGG-16 provides the accuracy of highest level. The best model underwent a hyperparameter ablation investigation, which led to the suggested hyper-tuned VGG16 obtaining 99.21% test and validation accuracy and 99.01% test accuracy.