Press play for diagnosis

A team of researchers from the University of Birmingham are developing a method to transmit medical video sequences across the telecommunications network efficiently without losing diagnostic information. If the work is successful it may permit doctors at smaller hospitals or mobile units to pass information rapidly to larger centres for expert advice.

The Engineering and Physical Sciences Research Council are funding the team, made up of Dr Sandra Woolley, Dr Mike Spann and Dr John Townend.

The project is focusing on angiography, a procedure where a dye is injected into a patient’s bloodstream and an X-ray video sequence digitally recorded, which shows features such as narrowing of the arteries around the heart.

On average, a video sequence lasts around 5 minutes and uses approximately one gigabyte of data. ‘It would take around 40 hours to transmit this over a normal telephone line’ said Dave Gibson, a research fellow on the project.

To make the data more manageable the research team is investigating how to ‘compress’ it. To do this the data must be slimmed down by mathematical manipulation to enable it to be transmitted efficiently, then restored at its destination.

‘Most transmission of medical images uses a so-called lossless approach,’ said Dr Woolley. ‘This means that what you retrieve at the other end is identical to the information that was sent. However, to achieve this the data can be compressed only in a limited way, typically two- or three-fold.’

This would be insufficient for angiogram video sequences. Instead a ‘lossy’ approach must be taken, whereby some data is lost on decompression. ‘This allows much greater compression,’ said Dr Spann. ‘The trick is to ensure that the data which is lost is not important – it must be diagnostically lossless.’

The research team has examined conventional lossy approaches to compressing the data, similar to those used for sending digital television images, but these methods have proved inadequate.

Instead the researchers are investigating a relatively new technique called wavelet-based image compression. In this a mathematical operation – the wavelet transform – is performed on the image data. This restructures the information content of the images to make efficient compression possible.

‘By performing the inverse operation the data can be reconstructed,’ said contract research fellow Dr George Tsibidis. ‘The result is that some of the image data is lost, but importantly the most significant features of the image are conserved.’

By using this technique the Birmingham team has compressed angiogram video data by a factor of ten and it is currently being assessed for its diagnostic accuracy.