Master thesis presentation by Fredrik Elofsson: “Detection of Coronary Arteries in Contrast-Enhanced 3D CT-Images”
On Wednesday 15 June, Fredrik Elofsson, MPCAS, will present his master thesis with the title “Detection of Coronary Arteries in Contrast-Enhanced 3D CT-Images”.
When? 11:00 am. Wednesday, 15 June, 2016
Where? Landahlsrummet (room 7430), Hörsalsvägen 11, 7th floor
Supervisor: Behrooz Nasihatkon
Examiner: Fredrik Kahl
Within cardiology, assessing the function of coronary arteries is of vital importance as these arteries supply the heart with oxygenated blood. A method routinely used in this assessment is x-ray computer tomography (CT), where a 3D image of the heart is analyzed by an expert. However, this is a laborious task, and thus it is tempting to automate parts of this process. In this thesis, a coronary artery segmentation method is presented. Consisting of two steps, the origins of the right coronary artery (RCA) and the left anterior descending artery (LAD) are estimated via a feature-based image registration method. Secondly, an incremental shortest-path algorithm is presented to track the vessels.
The segmentation method is evaluated on the public data set of the Rotterdam Coronary Artery Algorithm Evaluation Framework (RCAAEF). Using the training set of the RCAAEF, the average error of estimating the origin of the RCA and LAD is found to be 5 mm. Furthermore, the vessel segmentation is evaluated by tracking the RCA in the training set. The tracking performance is found to be on average 53% in the framework’s standardized overlap measure (0-100%, where “100%” means that the entire segmented vessel is within the radius of the true vessel). The results are worse than that of other methods submitted to the RCAAEF, and potential improvements of the proposed method, such as choice of data cost function and post-processing options, are discussed.