Medical Imaging

Contact: Twan Maintz

Description
Projects
Student projects
Publications

Description
Our research in the area of medical imaging focuses on image processing and image analysis techniques for the tasks of image segmentation and image registration. Currently, specific projects involve the accurate segmentation of the aortic lumen and aortic thrombus using a model-based approach, and the improvement and extension of current medical image registration techniques based on an information-theoretic approach.

Projects
Segmentation for support of AAA-treatment
In minimally invasive surgical procedures, instruments are introduced into the patient's body through small incisions. These procedures are expected to have several advantages compared to traditional surgery, such as a quick recovery of the patient, shorter hospital stay, and the possibility to use local anesthesia. An example of such a minimally invasive surgical procedure is the socalled TEAM-procedure used for endovascular repair of an Abdominal Aortic Aneurysm (AAA). TEAM stands for Transfemoral Endovascular Aneurysm Management. In this procedure, a prosthesis is inserted into the abdominal aorta via the femoral arteries. Insertion is done using a delivery system, which contains a folded prosthesis. If the prosthesis is at the right position in the aorta, it can be deployed, using the handles and knobs on the delivery system. Balloons are used to fix the prosthesis to the vessel wall. The whole procedure is done under fluoroscopic image control. To get an indication of the severity of the aneurysm and to check the result of the operation in the long term, the volume of the thrombus is determined. In current clinical practice, the volume is obtained by manual delineation of the boundaries of the thrombus in every slice of a CTA image, thus obtaining a segmentation of the thrombus.
CTA picture CTA picture
This segmentation as well as others -such as the segmentation of the prosthesis- is highly subjective. In this project we want to automate the process of these segmentations, thus gaining a more objective basis for important clinical measures. This is done using physical as well as statisical models for shape. Please refer to the publication list (search for e.g. 'CTA') for more details.

Coregistration of 3D images
Coregistration -the bringing into the same coordinate reference frame- and integration of images of the same scene but of a different modality is an area of research that has received much attention in the past 15 years. Demand for accurate coregistration is coming especially from the medical community, where use of multimodal imagery is commonplace.

The past decennium has seen the emergence of many automated methods for the coregistration of multimodal images, the most succesful of which are based on statistical paradigmata that do not assume a specific relationship between the images.

In this project, we research the use of coregistration based on mutual information; a statistical measure (also known as the Kullback-Leibler distance) that has been shown to be accurate for a number of applications. The mutual information is computed from the two-dimensional histogram of the images involved, see the images below for an example. For more details, please refer to our publication list (search e.g. for 'registration').
CT MRandHisto

Student projects

An incomplete list of master and other projects (in Dutch) can be found here . Contact Twan Maintz for more information.


webmaster: Twan Maintz