Ahmed A. Abdulredha Al-Taie

• Uncertainty is widespread when interpreting medical imaging data sets due to different sources of artifacts such as signal measurement errors or noise and partial volume effects in the acquisition process. The effectiveness of visualization methods for supporting decision making and diagnostic exploration is limited by the lack of suitable uncertainty estimation and visualization tools. A striking example of uncertainty visualization importance is given in neurosurgery, where errors in the range of a millimeter can have dramatic effects. Image segmentation plays an essential role in a broad range of computer vision and image processing applications such as pattern recognition, geographical imaging, geology, remote sensing, and medical imaging. In image segmentation, images are partitioned into disjoint regions of homogeneous properties. This study is concerned with deriving estimates of uncertainty associated with the segmentation result of single- and multi-modal medical imaging data, developing methods to visualize these estimates intuitively, and using these estimation in both evaluating and improving image segmentation results and the analysis of these results. As domain of applications, we use synthetic images that simulate the main brain structures in magnetic resonance imaging (MRI) data, simulated MRI data from BrainWeb, and Real MRI data as well. The main objectives of our study are to distinguish the certain and uncertain areas in the segmentation result, and to categorize the uncertain areas according to their level of uncertainty. By uncertain area, we basically understand the area that is affected by one or more sources of artifacts. Note: (see the complete abstract in the thesis)