الفهرس 
Medical Imaging Segmentation TechniquesOnly 14 pages are availabe for public view
 |  | from | 125 |  |  |
| | | from | 125 | | |
Abstract
Vascular structures inspection is important to assess patient’s cardiovascular risk. Current developmental techniques in imaging modalities including CTA (Computed Tomography Angiography) and PC-MRI (Phase-Contrast Magnetic Resonance Images) has brought attention to the possibility for non-invasive assessment of CAD (Coronary Artery Disease) patients.
This thesis proposes a fully automatic technique to segment the beginning of the cardiac tree which is the ascending aorta beginning from the aortic arch down to the ostia points and a fully automatic technique to detect ostia points’ coordinates. Also, a model for the ascending aorta is proposed and built from its anatomical features. The whole system is implemented for the purpose of quantitative analysis of thoracic aortic diseases to assess patient’s cardiovascular risk.
Automatic Ascending Aorta Segmentation is done using a proposed algorithm that automatically detects the ascending aorta using model fitting augmented with Hough transform. After detection, the whole ascending aorta is segmented from the aortic arch down to the ostia points using a proposed automatically seeded region growing algorithm. After segmentation is done an output volume of segmented ascending aorta cross sections is provided.
The automatic detection of the ostia points has been done using two approaches: Template Matching and Harris Corner detection. The output volume of segmented ascending aorta is input to the automatic ostia detection techniques. For template matching two templates are provided; one for the left and one for the right ostium. On the other hand Harris corner detection exploits the anatomical feature of the ostia that they appear like a corner/curvature on an ascending aorta contour.
The proposed detection and segmentation algorithms are tested and validated on the Computed Tomography Angiography database provided by the Rotterdam Coronary Artery Algorithm Evaluation Framework and Phase-Contrast Magnetic Resonance Images acquired using 1.5 Tesla - MRI scanner, for eight patients containing overall time frames of ninety time frames.
The automatic detection and segmentation of the ascending aorta succeeded in all test cases acquired from the two imaging modalities; proving the robustness of the ascending aorta model and the proposed algorithm in the automatic segmentation process even on data from different modalities and different scanner types. The accuracy of the segmentation has a mean Dice Similarity Coefficient (DSC) of 94.72% for CTA datasets and 97.13% for PC-MRI datasets.
The automatic detection of ostia points succeeded in all test cases from the CTA database with 100% detection rate attained by the template matching algorithm and 95.83% attained by the Harris Corner detection algorithm. Concerning accuracy; ostia coordinates’ points are detected with deviation of 0 to 10 pixels from the ground truth for both algorithms. Measurement of the diameter of the segmented ascending aorta cross sections is done to detect abnormalities. Our system detects thoracic aortic diseases including ascending aortic aneurysm and ascending aortic dissection. The abnormal measurements are alerted to the physician to take the appropriate action, together with the ostia position to help in surgical planning; in case coronaries ostia re-implantation is needed.
The proposed techniques are fully automatic, works in real-time, robust as parameters set values are the same for all tested datasets and accuracy attained is competitive with respect to other previous work.