Méthodes de segmentation d'images médicales basées sur la fusion d'information clinique : application à l'ouverture de la valve aortique et à la réalisation des contours de la prostate
|Advisor:||Laurendeau, Denis; Beaulieu, Luc|
|Abstract:||Since many years the use of medical imaging techniques has increased significantly. Medical imaging has driven the development of treatments and diagnosis to increase the efficiency and the precision of the physicians. This thesis proposes two methods to help the segmentation of anatomical structures in medical images. The first technique creates semi-automatic segmentation for the opening of the aortic valve. This method combines active contours (snakes) and a priori information from the electrocardiogram for guiding the segmentation. This association is the major contribution of this approach. This method has been tested on six patients. The curve of the area of the opening of the valve produced by the algorithm is very similar to the same curve obtained with manual segmentation. The second technique extracts a segmentation of the prostate on CT images using ultrasound data. The aim of this tool is to suggest initial contours to the physician in order to reduce the variability in his delineation of the prostate volume. The major contribution of this technique is to project planning ultrasound contours on the CT images. After the projection, the contours are directly adapted to the CT image with a deformation process. A clinical survey has been led to assess that this tool can help to reduce the intra and inter-observer variability in his delineation of the prostate volume. The result of this study shows that it is possible to reduce the inter-observer variability by 6% on the complete volume. It is also possible to reduce the intra observer variability by 12%. The time for delineation of the prostate was also a factor that was measured in the clinical study. It was found that it is possible to reduce the time to draw contours as much as 46% when initial contours are suggested to the physician.|
|Document Type:||Thèse de doctorat|
|Open Access Date:||16 April 2018|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.