Technique could reduce CT scan radiation exposure

GE Global Research and the University of Michigan have received $1.9m (£1.2m) from the US National Institutes of Health to conduct research into improving the quality of lower-dose CT scans.

CT (computed tomography) scans X-ray the body in slices and assemble the data into a three-dimensional picture. The scans are used to find tumours, complex fractures, blood clots, emphysema and clogged coronary arteries, among numerous other conditions.

Modern scanners with multiple detector rows can collect full data sets in seconds. Increased use of CT imaging has, however, led to increased radiation exposure.

Jeff Fessler, a professor in the Department of Electrical Engineering and Computer Science, said: ‘You want to expose people to as little radiation as possible. This is especially true for sick people who need repeated scans.’

Fessler is the principal investigator on the project, and will collaborate with radiologists in the U-M Health System and scientists at GE’s Global Research Center in Niskayuna, New York.

The project will focus on algorithms that can deliver more information out of the X-ray data.

A conventional CT scan takes pictures of 64 cross-sections of the body at a time, each less than 1mm thick. A typical scan of the head would produce 6,000 images.

‘It’s an enormous amount of data,’ said Fessler. ‘And the raw data is uninterpretable by doctors. We need image reconstruction software under the hood of the scanner, so to speak, to produce meaningful pictures.’

CT scans were invented in the 1970s, but the mathematics used now in their image processing algorithms dates back to 1917.

‘These algorithms are used because they are relatively simple and fast,’ Fessler said. ‘But they don’t extract the most from the data that is theoretically possible.’

The collaborating teams have demonstrated that their techniques have the potential to generate comparable image quality with a quarter of the present radiation exposure. In this project, they will conduct broader trials and work on speeding up the processing to make it practical for busy clinics.

This method could also enable higher-resolution images, illuminating the fine details of very small lung airways. While this project focuses on lung scans, these techniques could be expanded to all CT scans.