Convolutional neural network based brain MRI segmentation

Download
2018
Baydar, Bora
Visualization of the inner parts of human body is crucial in modern medicine and magnetic resonance imaging(MRI) is one of the widely used medical imaging methods. Manual analysis of MRIs, however, wastes the valuable time of experts. Development of an automatic segmentation method for brain MRIs can save time spent by the experts and can avoid human error factor. In this thesis, convolutional neural network (CNN) based methods are applied on brain MRI segmentation problem. The basic architectures used are FCN-8 and U-NET. Performance of different approaches has been analyzed by focusing on structural modifications, upsampling methods, activation functions, loss functions, pre-processing and post-processing methods. For the activation functions, ReLU, LReLU, PReLU and tanh are experimented. Histogram matching, normalization and histogram equalization have been applied for pre-processing. Conditional random fields and a 3 dimensional connected component analysis are separately integrated to the network as post-processors. Results are compared in terms of dice score, sensitivity and specificity. The experimental results show that the combination of two separate U-Nets has the best performance. Dilation modules also improve the results when inserted on a shal-low network. When combined with additional residual connections, they have alsoimproved the overall results. Inception modules do not provide a remarkable performance improvement. ReLU and PReLU have shown the best performance. No significant difference have been observed between results obtained from different upsampling methods, although bilinear interpolation, despite being non-trainable, has slightly better results.

Suggestions

Functional magnetic resonance imaging (fMRI) simulator
Arslankoz, Kamil; Gökçay, Didem; Department of Medical Informatics (2015)
Functional magnetic resonance imaging (fMRI) utilizes the change in the oxygenation of blood to predict active areas in the brain. fMRI consists of multiple low resolution whole brain images, for which, the contrast difference in corresponding voxels among all images are studied. In this study, an fMRI simulator has been developed which generates customized 4D fMRI data that can be used as a ground truth for comparing/benchmarking different fMRI analysis methods. This simulator can be also used for educatio...
Improving the sub-cortical gm segmentation using evolutionary hierarchical region merging
Çiftçioğlu, Mustafa Ulaş; Gökçay, Didem; Department of Medical Informatics (2011)
Segmentation of sub-cortical Gray Matter (GM) structures in magnetic resonance brain images is crucial in clinic and research for many purposes such as early diagnosis of neurological diseases, guidance of surgical operations and longitudinal volumetric studies. Unfortunately, the algorithms that segment the brain into 3 tissues usually suffer from poor performance in the sub-cortical region. In order to increase the detection of sub-cortical GM structures, an evolutionary hierarchical region merging approa...
An Information theoretic representation of brain connectivity for cognitive state classification using functional magnetic resonance imaging
Önal, Itır; Yarman Vural, Fatoş Tunay; Department of Computer Engineering (2013)
In this study, a new method for analyzing and representing the discriminative information, distributed in functional Magnetic Resonance Imaging (fMRI) data, is proposed. For this purpose, a local mesh with varying size is formed around each voxel, called the seed voxel. The relationships among each seed voxel and its neighbors are estimated using a linear regression equation by minimizing the expectation of the squared error. This squared error coming from linear regression is used to calculate various info...
CEREBRA: a 3-D visualization and processing tool for brain network extracted from fMRI data
Nasır, Barış; Yarman Vural, Fatoş Tunay; Department of Computer Engineering (2017)
In this thesis, we introduce a new tool, CEREBRA, for visualizing 3D network of human brain, extracted from the functional magnetic resonance imaging (fMRI) data. The tool aims to visualize the selected voxels as the nodes of the network and the edge weights are estimated by modeling the relationships among the voxel time series as a set of linear regression equations. This way, researchers can analyze the active brain regions/voxels and observe the interactions among them by analyzing the edge weights and ...
Magnetohydrodynamic Flow Imaging Using Spin-Echo Pulse Sequence
Eroğlu, Hasan Hüseyin; SADIGHI, MEHDI; Eyüboğlu, Behçet Murat (2019-04-24)
In this study, magnetohydrodynamic (MHD) flow of conductive liquids due to injection of electrical current during magnetic resonance imaging (MRI) is investigated. A spin-echo based MRI pulse sequence is proposed to image the MHD flow. Magnetic resonance (MR) phase effects of the MHD flow is related to the MRI pulse parameters and injected current. Average velocity distributions of the MHD flow are reconstructed using the MR phase images. The method is validated by numerical simulations. The reconstruction ...
Citation Formats
B. Baydar, “Convolutional neural network based brain MRI segmentation,” M.S. - Master of Science, Middle East Technical University, 2018.