6533b838fe1ef96bd12a4d0e

RESEARCH PRODUCT

Tomographic image processing for the morphological and metrological study of Valsalva sinuses

subject

description

This Phd thesis deals with the design and the use of image processing tools in order to allow a reliable and objective study of the sinuses of Valsalva which are important cavities of the aortic root. The proposed methods can be applied on cine-MR sequences and CT examinations without any change in the settings between two examinations.Firstly, we studied the morphology of this anatomical area and its constant properties in all images of the dataset. Sinuses are one of the main bright organs with limited movements. Hence a new algorithm has been designed. It detects and characterizes each bright organ by a single trajectory. Various tools of mathematical morphology are used for this step, as for the extraction of the contour of the sinuses in each image.The segmentation step is based on the geodesic reconstruction, which is more effective and more robust than the usual active contours. The shape depicting the sinuses is simply connected and a star domain. With this assumption, a new reconstruction is proposed, called the Aurora transform. This transform limits the spread of intensities only on the radial lines and shows its results in a polar space for a better reading of edges.The relevant points of the sinuses are also detected by a study of radii and the determination of dominant points along edges. An automatic measurement of the sinuses is deduced from these points. The values are very close to the manual measures currently done according to the intra-and inter-observer variations.Some other tools are finally outlined. They includes the modeling of edges by conics, the image classification depending on the time of the cycle in sequences and the tracking of the aortic valves in these examinations.This work led to the devlopement of a diagnostic aid software based on our methods. Its interface is also presented herein.