Advanced Image Reconstruction for Limited View Cone-Beam CT

Thumbnail Image
Publication date
Defense date
Journal Title
Journal ISSN
Volume Title
Google Scholar
Research Projects
Organizational Units
Journal Issue
In a standard CT acquisition, a high number of projections is obtained around the sample, generally covering an angular span of 360º. However, complexities may arise in some clinical scenarios such as surgery and emergency rooms or Intensive Care Units (ICUs) when the accessibility to the patient is limited due to the monitoring equipment attached. X-ray systems used in these cases are usually C-arms that only enable the acquisition of planar images within a limited angular range. Obtaining 3D images in these scenarios could be extremely interesting for diagnosis or image guided surgery. This would be based on the acquisition of a small number of projections within a limited angular span. Reconstruction of these limited-view data with conventional algorithms such as FDK result in streak artifacts and shape distortion deteriorating the image quality. In order to reduce these artifacts, advanced reconstruction methods can be used to compensate the lack of data by the incorporation of prior information. This bachelor thesis is framed on one of the lines of research carried out by the Biomedical Imaging and Instrumentation group from the Bioengineering and Aerospace Department of Universidad Carlos III de Madrid working jointly with the Hospital General Universitario Gregorio Marañón through its Instituto de Investigación Sanitaria. This line of research is carried out in collaboration with the company SEDECAL, which enables the direct transfer to the industry. Previous work showed that a new iterative reconstruction method proposed by the group, SCoLD, is able to restore the altered contour of the object, suppress greatly the streak artifacts and recover to some extend the image quality by restricting the space of search with a surface constraint. However, the evaluation was only carried out using a simulated mask that described the shape of the object obtained by thresholding a previous CT image of the sample, which is generally not available in real scenarios. The general objective of this thesis is the designing of a complete workflow to implement SCoLD in real scenarios. For that purpose, the 3D scanner Artec Eva was chosen to acquire the surface information of the sample, which was then transformed to be usable as prior information for SCoLD method. The evaluation done in a rodent study showed high similarity between the mask obtained from real data and the ideal mask obtained from a CT. Distortions in shape and streak artifacts in the limited-view FDK reconstruction were greatly reduced when using the real mask with the SCoLD reconstruction and the image quality was highly improved demonstrating the feasibility of the proposal.
Angular span, X-ray imaging, 3D Scanner, Computed Tomography (CT), Biomedical Imaging, SCoLD
Bibliographic citation