Φιλτράρισμα και ανάλυση υφής εικόνων υπερήχων καρωτιδικής αρτηρίας

Abstract

Ischemic stroke is one of the most characteristic problems causing death, full or partial disability.The main reason for the cause of IEE is the atheromatous plaque (AP) which is bind to the carotid arteries' walls.Consequently, an early diagnose of the atheromatous plaque could prevent these consequences.The creation of an evaluation and separation of ultrasound- images system based on the filtering of multiplicative noise processes could certainly assist a doctor to define more accurately the problem and make a better diagnose so as to monitor the decease.The purpose of the following dissertation was the creation of a fully developed system in which there will be processing of ultrasound images and will apply algorithms for the removal of multiplicative noise.The removal of the multiplicative noise has as result the separation of ultrasound images (UI) into two classes, the symptomatic and the asymptomatic according to the texture features exported in them.The automatic system was implemented in a Matlab environment and a graphic interface of GUI, in order to be user-friendly and easy to use.Initially, the imported image is normalized and then the option of the algorithm filtering has to be chosen out of a choice of four different algorithms with the filtered image being presented along with the histograms of the image.There is also the choice of export of the texture features and the quality metrics within the initial image as well as in the filtered image.Furthermore, the system supports the separation of the imported and filtered image into sectors - three parts - and the export of histogram in each sector from which more information can be exported for the separation of the images into classes. The four filters were applied in a UI sample of 122 symptomatic people and 122 asymptomatic people.From the texture and quality characteristics which were exported, there was a comparison by different sectors, between the symptomatic and asymptomatic people so as to verify the characteristics which verify the separation of two different classes.Afterwards, a statistical analysis followed in order to yield the statistical differences and the separation of the two categories and the sectors through the Wilcoxon test.The research results attested the visual observation that the filters Speckle and SRAD come up with better results in the filtering process as well as in the statistical analysis of the separation of two categories.The Wilcoxon test also showed that there are some texture features which change after the filtering process using the particular filters and therefore results into the separation of the two categories.The statistical analysis of the texture features showed that in the images of symptomatic cases there are significant statistical differences in comparison to images of asymptomatic cases after the normalization and filtering processes.The biggest difference is observed when statistical analysis is applied on the three sectors where the second sector has the highest variation percentage between the two classes.The suggestion here is that a future research on various texture and quality characteristics could separate the images of the two categories as well as the application of other filters of multiplicative noise removal.Finally, the results could be used for further analysis and be applied with neural networks.