Since a few years, the Image Sciences Institute (ISI) at the University of Utrecht and the Department of X-Ray Diagnostics and Pre-development at Philips Medical Systems (PMS) are partnering in a project, called Optimization of Information Extraction from Digital X-Ray Image Sequences. The scientific work fits into the extensive Innovation Targeted Research Programme (IOP), launched by the Dutch Ministry of Economic Affairs in 1996. One of the areas of interest constitutes Image Processing, a topic under which eleven additional projects reside. Last June, Dr. Wiro Niessen presented some of the aspects related to multiscale medical image analysis at ASCI'98, the annual conference of the Advanced School for Computing and Imaging, held at the Vossemeren in Lommel.
The medically inspired ISI/PMS project focuses on digital X-ray projection image sequences of human organs and especially of blood vessels but the results can be applied to alternative domains, such as for instance technical enhancement for digital television frames and image registration, suitable for video compression which is used both in teleconferencing and for efficient storage purposes. Physicians can study the dynamic behaviour and characteristics of the blood vessels by means of digital angiographic image sequences. For clinical practice, the quality of the images in the sequence has to be superior. The team of Dr. Niessen therefore has designed a set of innovative, accurate, fast and robust image processing techniques.
The current method, applied for the examination and diagnosis of blood vessels in the human body, consists of digital subtraction angiography. This type of imaging modality offers a fairly satisfying patient load and image quality but motion artefacts tend to occur because of misalignment of the consecutive two-dimensional projected images. In fact, two images are recorded, one prior and one after the opacification of the concerned vessels. The inevitable time interval allows for motion of one or several anatomical structures in the 3D scene which is being imaged, and causes the motion artefacts. The implementation of powerful image processing techniques can avoid inaccuracies of this kind.
Even the use of simple techniques, such as pixel shifting, zooming and vascular tracing, is already fit to improve the image quality substantially. In the meanwhile, the computational speed has grown inversely proportional to the system prices which enables radiologists to apply workstations of only 10,000 dollar, provided with special purpose graphics-hardware and the right software, to reach the same or even higher degree of functionality as before. The most sophisticated algorithms are introduced to automatically calculate the optimal alignment within the studied region of two angiograms.
The process can even be extended to a greater number of areas as well as to all the successive frames, in order to align complete runs without the user having to intervene. The research for the ISI/PMS project is carried out by Ph.D. students. The obtained results are used for actual development of the techniques for clinical practice to create a knowledge transfer towards the industry. As such, the optimized image processing techniques will find their way to the next generation of digital angiographic imaging systems. We refer to the University of Utrecht Web site for more information on this project.