Research Information System
Thesis Type: Postgraduate
Institution Of The Thesis: Gazi Üniversitesi, Fen Bilimleri Enstitüsü, Turkey
Approval Date: 2019
Thesis Language: Turkish
Student: ZEYNEP CANTEMİR
Supervisor: HACER KARACAN
Open Archive Collection: AVESIS Open Access Collection
Abstract:Alzheimer's disease is a sustained progressive and irreversible neurodegenerative disease that causes loss of memory and thinking skills and the slow destruction of brain cells. One of the most commonly used neuroimaging methods in the diagnosis of the disease is magnetic resonance imaging (MRI), which is an important part of clinical evaluation. In Alzheimer's cases, the forms in the brain may have patterns that overlap with different diseases. Diagnosis of the disease is therefore a difficult task by distinguishing between Alzheimer's and healthy brain from MR image data. Recent approaches to challenging visual classification problems, such as in this study, are based on the use of deep learning methods to enhance classification success with stronger property extraction than previous methods. Therefore, the use of deep learning algorithms to assist in the diagnosis of the disease will help researchers make accurate and rapid computer-assisted diagnostics. The aim of this study was to determine the patterns in functional MR images by convolutional neural network, which is one of the deep learning architectures, to classify the images correctly and to facilitate diagnosis for the researchers. In this study as a data set, a subset of the ADNI neuroimaging database used like many studies in literature related to Alzheimer's. According to the classification pipeline which created, firstly preprocessing step is applied to the raw functional MRI data set. In the preprocessing step, reduce of the noise in the images, remove of non-brain tissue parts such as the skull from images, because of the need to motionless during the MR imaging correction of motion and such procedures was performed with the FSL analysis tool. After preprocessing, convolutional neural network training was performed to extract the features from low to high level, and then according to these features as a result of the test, high classification accuracy rate was obtained.