When a patient with a suspected stroke arrives in an accident and emergency unit, every minute counts. With the help of a CT angiograph (an image of the blood vessels) doctors can quickly recognize whether there is a blocked blood vessel or a hemorrhage. However, the CT angiograph does not deliver enough information on how the affected tissue is supplied with blood. For this the doctors need a CT perfusion examination: they inject a small amount of an iodine-based contrast agent into the patient’s bloodstream and record the distribution of the contrast agent through the tissue by means of several scans that for about 40 seconds. The CT machine creates numerous chronological cross-sectional images of the brain using radiography. These enable the system to calculate how long it takes the contrast agent to spread out, and from these calculations the doctors can draw conclusions regarding the flow of blood to the tissues.

A CT perfusion examination is very quick and can be carried out without any preparation. It is the best method of examining the circulation of blood to the tissues or the organs. Since the introduction of multilayer CT equipment in the late 1990s, the capacity for imaging has steadily increased, so that today whole organs can be dynamically represented. Perfusion examinations are used increasingly for the examination of tumors in the torso. They have the potential to provide better characterization of the tumor type and earlier indication of the effect of treatment. Often patients have to be examined several times. So, doctors have to decide whether the radiation exposure is justifiable, especially for children. "That's why reduction of the radiation dose is given high priority," explains Raupach. Following the completion of his studies in Cologne he obtained his doctorate in theoretical physics. "After that I was determined to work in applied research," says Raupach. Because he had long been interested in medical subjects, he found an ideal sphere of activity at the advanced development department for medical technology at Siemens. During his eight years at Siemens he has already come up with 61 inventions and been awarded 18 patents.

Almost as soon as he had begun work on this subject he came up with his basic idea. Radiation consists of photons which are directed through the body. The smaller the radiation dose, the lower the number of these light particles contained in the beam and the poorer the quality of the resulting image. Specialists use the term "noise." Raupach’s initial position was that there must be a way to reduce this noise.

The CT takes approximately 40 images per layer during the scan period so that the distribution of the contrast agent within the tissue can be observed, but the variation between one image and another is very small. "The changes in the images that are caused by the accumulation of the contrast agent in the tissue represent only very small variations from the basic image," explains Raupach. So why not use the fact that the basic image is contained in all the scans taken in order to generate a higher-quality mean result from the data obtained? "However, the difficulty was to avoid destroying the temporal definition of the contrast agent distribution, because that’s the most important information gathered from the examination," explains the physicist.

The starting point for his solution was the observation that the tissue change takes place in the area of low spatial frequency – in other words, in the blurred parts of the image – whereas the information contained in the high spatial frequency data – where the contours are sharp – only alters slowly. But image noise is found chiefly at high spatial frequency. So Raupach developed algorithms which would generate a mean image from the imaging data in the high spatial frequency areas. When this is combined with the low spatial frequency images, the relevant chronological changes are retained. The image noise is reduced at the same time. The image quality remains good, even if the radiation dose is lowered.

Because it is only the method of processing the data which changes, this system can be installed on any existing multilayer CT machine. Raupach tested his new algorithms on data sets that were already in existence and found that his theory was confirmed. "Even poor images that were actually taken with a radiation dose that was too low could be recalculated to provide good perfusion information," he says. Plans call for the algorithms to be launched on the market in early 2009 as part of the CT perfusion software.