|
|
Advanced Algorithms for 3D Medical Image Data Fusion in Specific Medical Problems
Malínský, Miloš ; Flusser,, Jan (oponent) ; Kozubek,, Michal (oponent) ; Jan, Jiří (vedoucí práce)
Image fusion is one of today´s most common and still challenging tasks in medical imaging and it plays crucial role in all areas of medical care such as diagnosis, treatment and surgery. Three projects crucially dependent on image fusion are introduced in this thesis. The first project deals with the 3D CT subtraction angiography of lower limbs. It combines pre-contrast and contrast enhanced data to extract the blood vessel tree. The second project fuses the DTI and T1-weighted MRI brain data. The aim of this project is to combine the brain structural and functional information that purvey improved knowledge about intrinsic brain connectivity. The third project deals with the time series of CT spine data where the metastases occur. In this project the progression of metastases within the vertebrae is studied based on fusion of the successive elements of the image series. This thesis introduces new methodology of classifying metastatic tissue. All the projects mentioned in this thesis have been solved by the medical image analysis group led by Prof. Jiří Jan. This dissertation concerns primarily the registration part of the first project and the classification part of the third project. The second project is described completely. The other parts of the first and third project, including the specific preprocessing of the data, are introduced in detail in the dissertation thesis of my colleague Roman Peter, M.Sc.
|
|
|
Fusion and Analysis of Multidimensional Medical Image Data
Peter, Roman ; Flusser,, Jan (oponent) ; Kozubek,, Michal (oponent) ; Jan, Jiří (vedoucí práce)
Analysis of medical images has been subject of basic research for many years. Many research papers have been published in the field related to image analysis and focused on partial aspects such as reconstruction, restoration, segmentation and classification, registration (spatial alignment) and fusion. Besides the introduction of related general concepts used in medical image processing, this thesis deals with two specific medical problems formulated in cooperation with Philips Netherland BV, Philips Healthcare division. The first topic is focused on subtraction angiography in patients’ lower legs utilizing image data from X-Ray computed tomography (CT). CT subtraction angiography (CTSA) is typically used for indication of the Peripheral Artery Occlusive Disease (PAOD) and for examination of acute injuries of lower legs such as acute fractures, etc. Current methods in clinical praxis are not sufficient regarding the pre-processing such as masking of patient desk, cover, splint, etc. The subtraction of blood vessels adjacent to neighboring bones in lower legs is of low accuracy due to the Partial Volume artifact. Masking of calcifications and detection of tiny blood vessels complementing necessary information about the alternative blood supply in lower legs in case of obstruction in main arteries is also not reliable for fully automated process presently. Therefore, the main aim regarding this topic was to develop an automated framework that could overcome current shortcomings in CTSA examination. The second topic is oriented on the identification and evaluation of pathologic changes in human spine, focusing on osteolytic and osteoblastic lesions in individual vertebrae in CT images. Such changes occur typically as a consequence of metastasizing process of cancerous disease. For the detection of pathologic changes, an identification and segmentation of individual vertebrae is necessary. Moreover, the analysis of individual lesions in vertebrae depends also on correct identification of vertebral body and posterior segments of each vertebra, and on segmentation of their trabecular centers. Patients are typically examined more than once during their therapy. Then, the evaluation of possible tumorous progression is based on accurate detection of pathologies in individual vertebrae in the base-line and corresponding follow-up images. Since the corresponding vertebrae are in mutually different positions in the follow-up images, their fusion leading to the analysis of the lesion progression is complicated. The main aim regarding this topic is to develop a complex framework for detection of pathologic lesions on spine, with the main focus on osteoblastic and osteolystic lesions. Such system has to provide not only reliable segmentation of individual vertebrae and detection of their main regions but also the masking of their cortical bone, detection of their pathologic changes and their evaluation. Although this dissertation thesis is primarily oriented at the experimental part of medical image processing considering both the above mentioned topics, it deals with all necessary processing steps, i.e. preprocessing, image registration, post-processing and evaluation of results, leading to the future use of both frameworks in clinical practice. Since both frameworks were developed in a team, there are some chapters referring to the dissertation thesis of Milos Malinsky.
|
host ::
RSS.