Includes CD

Includes references (95-109 p.)

'ABSTRACT A brain haemorrhage is a class of stroke which is caused through an injury that occurs in the artery of the brain which leads to bleeding around the tissues of the brain cells. As a result, the brain cells get damaged. The major causes of brain haemorrhage are mainly due to smoking, alcohol usage, brain tumours, drug abuse, high blood pressure and trauma. Recently, Deep learning based networks have shown a great generalization capability when applied to solve challenging medical problems such as medical image classification, medical image analysis, medical organs detection, disease detection. This thesis addresses the problem of identifying brain haemorrhage which is considered as a tedious task for radiologists, especially in the early stages of the haemorrhage. The problem is solved using a deep learning approach where convolutional neural networks (CNN) are employed, CNN were the most effective networks among deep neural networks, because of its architecture and has the paradigms of that is biologically inspired structured more than other traditional networks. Well-known AlexNet neural network, and also a modified novel version of AlexNet with support vector machine (AlexNet-SVM) classifier are trained to classify the brain Computed tomography (CT) images into haemorrhage or non-haemorrhage images. The major advantage of employing deep learning model is to address the primary question in medical image analysis and classification: can a sufficient fine-tuning of a pre-trained model (transfer learning) eliminate the need of building a CNN from scratch? Moreover, to investigate the advantages of using SVM as a classifier instead of a three-layer neural network. The same classification task was applied to three deep networks; one is created from scratch, another pre-trained model that was fine-tuned to the brain CT haemorrhage classification task, and the modified novel AlexNet model which uses SVM classifier. The three networks were trained using the same number of brain CT images available. The experiments show that the transfer of knowledge from natural images to medical images classification is possible. In addition, our results proved that the proposed modified pre-trained model "AlexNet-SVM" can outperform a convolutional neural network created from scratch and the original AlexNet in identifying the brain haemorrhage. Furthermore, Weighted Cross Entropy and Sequential features selection were introduced to AlexNet-SVM in an attempt to improve the performance of the algorithm. Keywords: Haemorrhage, convolutional neural network, fine-tuning, AlexNet, AlexNet-SVM, transfer learning, transfer of knowledge, deep learning.   '