الفهرس 
CHAPTER 1Only 14 pages are availabe for public view
 |  | from | 119 |  |  |
| | | from | 119 | | |
Abstract
A brain tumour is a highly serious condition that can have diverse effects on an individual’s overall well-being. Identified as a neoplasm characterized by abnormal cell proliferation, these tumours are typically located within or near the cerebral region. Brain masses can manifest as either benign or malignant neoplasms. In medical practice, various diagnostic methods are employed to determine the nature of a patient’s brain tumour, differentiating between benign and malignant tumours.Presently, the analysis of radiology images heavily relies on deep learning techniques, utilizing common imaging methods such as CT, MRI, PET, and ultrasound. Among these, CT, and MRI scans, each with its own advantages and disadvantages, emerge as the most widely utilized imaging modalities.To automate the detection of brain masses, a dedicated system has been developed specifically for CT and MRI scans, capitalizing on the unique advantages offered by each imaging method. The system initially verifies the type of input image. If identified as a CT-scan image, the recommended Convolutional Neural Network (CNN) architecture is employed for diagnosis. The CNN architecture has showcased exceptional performance, achieving impressive accuracy, F1-score, precision, and recall values of 98.01%, 98%, 99.7%, and 98.84%, respectively. Alternatively, for MRI scans, the system can employ Reset101, a pre-trained convolutional neural network, to ensure accurate diagnosis.Chapter 1 The initial section provides an overview of the structure, functions, and medical concerns related to the brain, along with a discussion on the role of artificial intelligence in medical imaging. A thorough examination of relevant literature has been conducted to underscore the importance of our research and showcase the progress made in utilizing deep learning methods for the diagnosis of brain tumours.Chapter 2 examined the fundamentals of medical diagnostic imaging’s physical principles. Four modalities: X-rays for diagnosis and treatment, CT, ultrasound, and magnetic resonance imaging.Chapter 3 introduces machine learning and deep learning approaches for image processing. The many phases of medical image pre-processing are where it all begins. Afterwards, the definitions of the tools for segmenting data and extracting features to comprehend the identification process.Chapter 4 explored the two types of X-rays possess distinct advantages, providing a valuable method for radiologists. The system subjects the input image to testing, and if identified as a CT-scan image, it employs the recommended Convolutional Neural Network (CNN) architecture for diagnosis. The CNN architecture has demonstrated exceptional functionality, achieving impressive accuracy, F1-score, precision, and recall values of 98.01%, 98%, 99.7%, and 98.84%, respectively. Alternatively, if the image is identified as an MRI scan, the system can utilize Reset101, a pre-trained convolutional neural network, for diagnosis. The test results for Reset101 show 99.8%, 99.9%, 99.2%, and 99.55% accuracy, precision, recall, and F1-score, respectively.