Taking a virtual X-ray

A Teesside University lecturer has created radiography software that allows students to take a "virtual" X-ray in a safe and controlled way without getting near a patient or an X-ray machine.

For his efforts, Philip Cosson, a senior lecturer in medical imaging at the university, has been named joint winner of the 2007 Learning Technologist of the Year Award by the Association for Learning Technology.

The program he developed, which is now being used at Teesside, is also being marketed by his own university spin-out company Shaderware.

The virtual radiography software uses CT data of humans from the US National Library of Medicine's Visible Human Project from which it generates simulated radiographs based on parameters such as patient position and orientation, tube location and angulation, as well as technical factors such as tube voltage, current and exposure time - all of which are controlled by the student.

The software traces the path of a virtual x-ray beam as it travels from the tube to the sensor through the virtual human being. For a representative set of rays striking the sensor, the software calculates how the original beam intensity would have been attenuated, given its individual path through the virtual human. Based on the intensity of the attenuated beam, a grayscale image is then generated based on a simulated receptor response curve.

The number of calculations required to build the final image is enormous, and well beyond the processing capabilities of the central processor used in current consumer computers.

Fortunately, the impetus of computer gaming has driven the development of graphics processors to the point where these dedicated chips are extremely powerful parallel processing units. The Shaderware virtual radiography software uses this specialised hardware to perform the geometric, resampling and attenuation calculations in a realistic timescale.

While today's top end graphics processor aimed at the gaming market will produce a simulated image in less than a second, processors sold in 2002 can produce an acceptable image in a few seconds. This means that virtual radiography can be used on very modest hardware.