One of the more severe disorders that affect both children and adults is a brain tumour as a higher percentage of all primary Central Nervous System (CNS) malignancies are brain tumours. About 11,700 patients are given a brain tumour diagnosis each year. For those who have a malignant brain or CNS tumour, the 5-year survival rate is roughly 34% for males and 36% for women (Cancer.Net, 2022). Benign, malignant, pituitary, and other types of brain tumours are all categorised. To extend patient lives, appropriate care, careful planning, and precise diagnostics must be used. Magnetic Resonance Imaging is the most effective method for finding brain cancers (MRI). The scans provide an enormous amount of picture data. The radiologist examines these pictures, but because of the complexity of brain tumours and their characteristics, a manual examination can be prone to inaccuracy. Automated classification methods based on Artificial Neural Networks have consistently outperformed manual categorization in terms of accuracy. So, it would be beneficial for Neurosurgeons all over the world to propose a system that does detection and classification utilising Deep Learning Algorithms like Convolution-Neural Network (CNN) and Transfer-Learning (TL) like ResNet50. Thus, this Course Work employs Multilayer Neural Perceptron which is also known as feed forward neural networks, Convolutional-Neural Network and RESNET 50 (Transfer learning) as deep learning tool to detect and classify brain tumours, given a dataset collected from Kaggle. The accuracy and loss results of the three models applied on the image datasets are as follows: ResNet 50 (Accuracy 0.99, Loss 0.0718) , MLP (Accuracy 0.47, Loss 0.6945) and CNN (Accuracy 0.97, Loss 0.1351).
Keywords: Resnet50, Convolutional Neural Networks, Multilayer Neural Perceptron, Artificial Neural Networks, Central Nervous System