Detection and analysis of brain tumours from Magnetic Resonance Imaging (MRI) scans are essential for the early detection and treatment of brain cancers. Manual analysis of the MRI images is a time-consuming process for radiologists, as there are possibilities of human errors, especially while handling large volumes of medical images. An automatic brain tumor image classification method utilizing Deep Learning algorithms is proposed to counter this problem. For this project, three different models of deep learning are applied. First, a personal solution of the Convolutional Neural Network (CNN) architecture is developed as a basic model to identify the spatial features of the MRI images. Second, VGG16, a transfer learning model developed earlier for other large-scale image databases, is utilized to tap into the experience gained from handling large-scale image databases. Third, ResNet50, a more complex model of the residual network architecture, is applied to overcome the problem of a disappearing gradient during the backward propagation process to achieve better results for the classification process of the brain tumor dataset. The proposed system can also provide real-time prediction, and any new MRI image can be uploaded by the user, showing the predicted tumor type with the confidence score. The experimental findings prove that the transfer learning models achieve higher performance compared to the basic CNN, among which ResNet50 obtained the top result. This work verifies deep learning and transfer learning in medical image analysis and delivers a trustworthy way of automatic brain tumor classification.
One of the In our technology, we successfully decode gene sequences and transplant seeds from variou...
One of the most common cancers in the world, thyroid cancer is becoming more common. Improving patie...
Nowadays, many people have pets, but because of busy daily schedules, they are not always able to fe...
Gridlock in metropolitan areas is the severe problem regarded within the gift lifestyles. The motors...
