Medical Device Daily Washington Editor
WASHINGTON – On the last day of FDA's “Century of Science“ forum, Kyle Myers, PhD, the lab leader in the Division of Imaging and Applied Mathematics at the FDA's Center of Devices and Radiological Health, gave a description of recent developments and challenges in medical imaging.
While the pace of technological advance has been impressive, adoption of such technology faces a number of hurdles, not the least of which is the obvious one of cost, she noted.
However, the ability of providers to grapple with the increased data load is also coming under substantial pressure as the sheer volume of data continues to multiply, and studies indicate that not all physicians are equal in their abilities to correctly interpret the images they see.
Myers covered a number of subjects in her wide-ranging presentation, and even display technology came under scrutiny during the FDAer's talk.
Early in her presentation she noted that “there is an incredible amount of work being done“ to develop and improve imaging technologies for various medical uses but that these technologies are driving up the amount and quality of data that providers must find a way to accurately interpret. The increased demands on display technology fed by improved scanning technology has prompted FDA to state on its web site that computer displays are “the weakest link of the imaging chain for many applications,“ including mammography.
Perhaps because of this perceived deficit, interest in display technologies has prompted the formation of societies devoted to this area of medical science.
They include the Society for Computer Applications in Radiology (Great Falls, Virginia) and the Society for Information Display (San Diego). The former publishes the Journal of Digital Imaging, which has presented peer-review articles since 2001 covering a range of subjects, including comparisons of digital display units for clinical settings and development of processes to detect changes in magnetic resonance images of the brain.
Myers commented that cathode ray tube displays continue to lose ground to liquid crystal displays (LCD) and liquid electrode display (LED) technologies. Among these are active-matrix LCDs (AMLCDs) and organic light-emitting diode displays (OLED).
The advantages of AMLCDs are well known: brighter image and a broader viewing angle compared to passive LCD displays. The disadvantages include a phenomenon known as subpixel crosstalk, which can distort the image.
As for OLEDs, the advantages include that the organic matter can be essentially “printed“ onto a substrate, reducing the cost of manufacture, and can be printed onto a flexible substrate, allowing the user to roll up and haul away the screen. However, the organic compounds involved degrade much more quickly than the media in other types of displays and are very susceptible to damage by moisture.
CDRH commenced with an overall effort to improve medical imaging in the 1970s and, in the following decade, began working toward deploying multivariate statistical methods to assess the specificity and sensitivity of computer-aided diagnosis (CAD).
Breast cancer has been the subject of much interest as an application of CAD because of a conspicuous level of variability in diagnoses from one clinician to another. Between FDA's efforts in this area and those of private industry, a CAD system found its way through the development and regulatory thicket in 1998 in the form of the ImageChecker M1000, manufactured by R2 Technology (Los Altos, California).
However, Myers indicated that more than one device “has been approved as a second reader in mammography.“
Myers noted that the agency sees the possibility that CAD software could “possibly be so well trained“ as to eliminate the need for a human reader in evaluating Pap smears, but admitted that at present, there is “a real need for quantitative tools“ that would enable such a development.
In particular she described two studies addressing differences in how physicians approach scan evaluations that highlighted the “tremendous amount of variability in reader skill.“ However, the subject of film vs. digital mammography also came up.
Myers noted that an article appearing in the Oct. 27, 2005, edition of the New England Journal of Medicine examined how low-tech and high-tech compared in this area.
In that study the team of Pisano, et al., collected data from 33 sites in the U.S. and Canada from the screenings of more than 42,000 women largely to investigate the question of how these two technologies perform in evaluating pathologies in “radiographically dense breasts.“
The researchers concluded that while digital screening “is more accurate in women under the age of 50 years, women with radiographically dense breasts and premenopausal or perimenopausal women,“ film scored just as well in terms of “overall diagnostic accuracy.“
The authors made the case that the ease of transmission of digital images and the lower required radiation dose were meaningful advantages, but they acknowledged that “one of the major impediments to the adoption of digital mammography will be its cost,“ which the authors pegged at “1.5 to four times as much as film systems.“
As for data proliferation, Myers used a slide to depict the number of CT images processed daily at the Mayo Clinic (Rochester, Minnesota). According to the numbers she presented, Mayo processed an average of 1,500 CT scans per day in 1994, a number that ballooned to 80,000 per day this year.
This kind of increase in data, Myers argued, is driving a profound need for alternative display technologies and CAD systems.