Diagnosis: educate doctors on imaging

The Duke University Department of Radiology networked medical-imaging system is, perhaps, one of the most sophisticated in the world (see p. 28). Images from CT and NMR scanners can be viewed almost instantly by doctors anywhere in the hospital. Despite this sophistication, however, images are merely being displayed for analysis and often output to film.

Diagnosis: educate doctors on imaging

The Duke University Department of Radiology networked medical-imaging system is, perhaps, one of the most sophisticated in the world (see p. 28). Images from CT and NMR scanners can be viewed almost instantly by doctors anywhere in the hospital. Despite this sophistication, however, images are merely being displayed for analysis and often output to film.

Just as the paperless office promised to rid journalists of piles of press releases, the filmless office promises to free doctors of mountains of expensive film. But that has not happened. One of the reasons is familiarity. Journalists often use press releases as the lead for a good story. Piles of press releases mean security.

Practicing radiologists feel similar security with film. They feel comfortable using light tables and diagnosing from film. In medicine, legal issues complicate the matter, too. If a doctor makes a wrong judgment from a JPEG corrupted digital image, will the patient`s insurance company pay? Does the insurance company even know what JPEG means? As a result of these issues, medical image-processing systems have not gained acceptance. "Even with systems like those at Duke University," says Karen Miller of Dome Imaging Systems, "radiologists want to interface with digital images the way they do film--no sophisticated image processing, just simple window level, zoom and rotate."

Progress is made

And yet, research centers around the world continue to make progress on medical applications of digital image processing. At the University of Florida, research on wavelet transforms has led to vast improvements in mammograms for diagnosing breast cancer. At the Mallincrodt Institute of Radiology in St. Louis, MO, interpolation techniques are being used to reduce the amount of reflectance from medical prostheses in patients undergoing CT and NMR scans.

But it will be a while--a long while(before such techniques are in the hands of doctors. Market acceptance of digital imaging in medicine has been slower than anticipated due to lack of education. If teaching hospitals could raise the comfort level of film-based radiology, that would be a step in the right direction. Educating such establishments on the benefits of digital image-processing techniques in diagnostic interpretation would be a giant step forward.

Programs needed now

Unfortunately, only tiny steps are being taken in this direction. What`s needed are university programs where student radiologists learn to integrate film, digital film, and digital image-processing techniques. Then, radiologists and others could reap the benefits of making the transition to digital techniques while relying on film-based methods. They would also understand the benefits that point-and-click techniques would bring to making diagnoses.

By adopting digital image-processing techniques, radiologists and others could draw on the vast experience of others. Image-processing packages that incorporate this knowledge could be developed by veteran radiologists. Even better, sophisticated neural-network programs could be trained to learn from them. In this way, medical image-processing programs could offer interpretation of images, making diagnoses more rapid and thorough. For hospitals, this would increase patient throughput and save money. For the patient, the use of sophisticated medical image processing would ensure a more complete diagnosis.

Andy Wilson Editor

andyw@pennwell.com

More in Boards & Software