Cardiac scintigraphic images segmentation techniques
Yosra Ben Fadhel, Sami Ktata, Tarek Kraiem · 2016
Segmentation is an important task in all kinds of image analysis. Especially medical image analysis segmentation has a great clinical value that improves the localization of organs or pathologies in order to raise the quality of diagnosis. The left ventricle is the most important part in radionuclide ventriculography image (VEF: Ventricular Ejection Fraction). In order to identify it, we did used two segmentation approaches which are: region-based segmentation and edge-based segmentation. The first detects homogeneities and the common features by using the active contour model (chan_vese) and the thresholding techniques. The second, seeks the existence of a transition between two related regions by using some methods like (Sobel, Canny and Perwitt). VEF image is known as a noisy image with a low contrast which make the perception of interest parts (ROI: Regions of Interest) difficult, for that in this paper the use of a preprocessing algorithm is suggested. To improve the cha-vese segmentation result, we applied a simple threshold followed by the chan-vese model (combination of two techniques).