Automatic Segmentation of Intracranial Arteries in 4-Dimensional Phase Contrast Magnetic Resonance Angiography
Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
The development of a highly effective phase-contrast magnetic resonance imag- ing (PC-MRI) technique named phase-contrast vastly undersampled isotropic projection reconstruction (PC-VIPR) has improved the ability to image blood flow in the brain. The technique allows for the acquisition of temporally resolved volume images with high spatial resolution. Utilizing these improvements is of importance for diagnostic and research applications. The aim in this project was to investigate and implement PC-VIPR suitable segmentation techniques for blood flow quantification and labeling of blood vessels. The aim was to con- struct an automatic segmentation tool being able to accurately quantify blood flow and to label a few key arteries in the Circle of Willis.
A vascular tree construction was performed in which the vascular tree was separated into individual branches. Four different methods of boundary detec- tion of blood vessels were implemented and evaluated on their performance in quantifying blood flow based on conservation of mass principles and internal er- ror. A labeling algorithm was constructed in which labels of a few key arteries were assigned. A total of ten subjects were analyzed to provide evaluation of the segmentation tool.
The segmentation tool was constructed with complete automaticity. The best method for boundary detection showed an average mass conservation error of −0.9 ± 6.5 % where the internal carotid artery splits into the middle cerebral artery, the anterior cerebral artery and the posterior communicating artery. A labeling accuracy of 83 % was acquired. The results of the blood flow quan- tifications and the labeling of arteries were partly a result of the vascular tree construction, which were considered effective.
Keywords. 4D PC-VIPR, segmentation, intracranial arteries, flow quantifica- tion, artery labeling
Place, publisher, year, edition, pages
2013. , 54 p.
Medical Image Processing
IdentifiersURN: urn:nbn:se:umu:diva-73548OAI: oai:DiVA.org:umu-73548DiVA: diva2:632538
Master of Science Programme in Engineering Physics
Wåhlin, AndersEklund, Anders