| Περίληψη: | The liver is one of the most important glands in the human body as it has a wide range of functions including detoxification, protein synthesis and the production of biochemicals necessary for the digestion of food.
Liver failure is a disease which, if it reaches its final stage, can result in the death of the patient. For this reason, various methods have been developed where we can locate the situation. These are biopsy, blood tests and also imaging techniques, namely ultrasound and MRI scans. Although biopsy is the most valid method of diagnosis, it is an invasive method which makes it to some extent a deterrent due to its cost and also due to complications that may arise. Although blood tests are an invasive and non-costly method, the results of blood tests cannot reflect the degree of fibrosis of the liver. Imaging methods in combination with automated diagnostic systems are the most cost-effective and cost-neutral method at the present time.
The aim of this work is first to process elastography images and then to use them in a deep learning network in order to categorize the stage of fibrosis the patient is in and to compare this network with the physician's estimates. The elastographic image obtained consists of RGB values where values close to red indicate high stiffness while values close to blue indicate low stiffness. The first step was to map these values and convert them into values expressing the stiffness in kPa. Then after having collected a sufficient sample of patients (200) in different stages of chronic hepatopathy, then a deep learning network is used to categorize the patients. The accuracy of the system is evaluated and then the results are compared with the physician's diagnosis using the biopsy as a reference.
|
|---|
