Image processing and analysis algorithms development for chronic liver disease assessment using ultrasound elastography images

• Κύριος συγγραφέας: Ζγκούρη, Μαρία
• Άλλοι συγγραφείς: Zgkouri, Maria
• Γλώσσα: English
• Έκδοση: 2023
• Θέματα: Αλγόριθμοι επεξεργασίας εικόνας; Χρόνια ηπατική νόσος; Εικόνες ελαστογραφίας; Image processing; Algorithms; Chronic liver disease assessment

The liver is an essential organ of the body that performs over 500 vital functions, such as protein, fat and carbohydrate metabolism, immunity, digestion, detoxification, bile production, vitamin and mineral storage among other functions. Although liver has the ability to react to damage, there are a lot of serious diseases that can cause even liver failure leading to death. Therefore, a variety of methods have been developed, in order to monitor liver diseases or damages, estimate the patient’s disease stages and to select the best treatment. Diagnostic procedures such as liver biopsy (LB), biochemical markers (BSMs), CT and MRI imaging, ultrasound (US) and ultrasound elastography (USE) are widely used. Although LB is considered the gold standard method, a lot of limitations are set due to its invasive technique, high cost and probable complications. Further diagnostic procedures like BSMs, CT, MRI, US or USE are non-invasive and low-cost methods but they cannot provide accurate information about chronic liver disease (CLD) stage. Imaging methods in combination with automated diagnostic systems is the state-of-the art in CLD assessment and also cost-neutral. The purpose of the present thesis is to employ a computer-aided diagnosis system that classifies chronic liver disease (CLD), by using and processing ultrasound shear wave elastography (SWE) images, and to investigate the algorithm’s accuracy in comparison to the radiologist’s accuracy for each fibrosis stage. The SWE images obtained consist of RGB (red-green-blue) scale values, with the blue color corresponding to low stiffness and the red color corresponding to high stiffness values (in units of kPa). Then we convert the color map to stiffness map with the procedure of RGB-to-stiffness inverse mapping, we split the stiffness values into 5 clusters, the maximum participation color area (MPCA) is extracted and MPCA’s mean value is calculated. Using a data set of different fibrosis stage patients, the proposed automatic algorithm classifies the patients. The algorithm’s accuracy is assessed, by ROC curves calculation, and is compared to the examiner’s accuracy for the same data set, with reference to the biopsy.