Medical Device Daily Washington Editor
GAITHERSBURG, Maryland – There is no novelty to the assertion that technology drives FDA's regulations, so the fact that FDA would convene a two-day session with the hematology and pathology devices advisory committee to discuss how to reliably employ digital images of pathology slides in diagnostic workups is not a surprise.
Also not surprising is the impression that the agency is not "guidance-ready" for this application of imaging technology, a fact which may be driven largely by FDA's perception that there is a lot of variance in how modern computer monitors display images, a variance that one speaker suggested industry will have to find a way to control in clinical trials for their equipment.
Speaking for FDA on this subject was Aldo Badano, PhD, of the Office of Science and Engineering Labs at FDA's Center for Devices and Radiological Health. "Displays are components in the imaging chain," he noted, a chain which includes image acquisition, processing, storage and transmission. However, he informed the panel – and implicitly industry – that FDA is of the view that "the display can be the weakest link in the chain."
Badano said that among the problems FDA sees in the variety of displays is "poor display quality," which he said often leads to "incorrect or inconsistent diagnostic decisions." Display quality can also lead to increased variability in reads and longer image read times. He suggested that an application for a digital imaging system would have to specify the type of display with which the imaging system would be used.
"We can think of having a regulatory path that will go by component," Badano said, a paradigm that "allows us to model system improvement" via plug-and-play of the various component types. While such an approach "raises some questions of inter-operator variability," this approach might nonetheless be useful but is "only possible if standard characterization methods are used to demonstrate the performance of the components," he said.
Badano reminded the panel that monitors based on cathode ray tubes were still in wide use in 2000, but that things have changed in the nine years since then. Monitors offering a resolution of as much as eight million megapixels are now available, offering a much larger color palette that could be a confounder for diagnosis. "If a system comes in with a different technology, we have to be able to reconsider the testing methods," Badano warned.
Most displays are based on liquid crystal technology, Badano remarked, and the image quality "depends on what comes out of the backlight and what is differentially absorbed by the layers" of materials just behind the front of the screen. He made note of a comprehensive measurement test for LCDs promulgated by the American College of Radiology (ACR; Reston, Virginia), a standard he described as a "simplified version of requirements for practice in radiology." This is in reference to an ACR document published by ACR in the Journal of the American College of Radiology in August 2006.
Another standard, this one offered by the American Association of Physicists in Medicine (AAPM; College Park, Maryland), is a "very long document of testing procedures for characterizing display performance in radiology," Badano said. This document, written by AAPM's task group 18, "has a variety of test patterns" for initial checks of display quality, but "also a few clinical images" for checks of display quality "before reading sessions."
A standard that is en route will be published in December by the International Electrotechnical Commission (IEC; Geneva), which Badano said will include visual tools with test patterns. "A good performing display for radiology" should be able to qualify under this standard, he said, but Badano also noted that all the current standards "only apply to grayscale displays."
While these standards deal with imaging issues such as pixel defects, "what we do not have at the moment is an extension [of those standards] to color displays," Badano observed, which may be helped by standards from other applications of imaging technology. Among these, he said, is ISO 13655, a graphic arts standard, as well as the ASTM standards of E1336 and E1455. All the same, Badano noted, "we do not have a practical standard for measuring the physical characteristics" of some color displays.
Other issues that have yet to be resolved include that image capture technology may be outrunning display technology. "We have detectors that could acquire two to nine megapixels, which are much larger" than display capacities, which often run no higher than five megapixels. He also warned that he is of the impression that "display products with lower image quality are being considered based primarily on physical measurements."
"Another aspect that is well understood is ambient light," Badano said, noting digital radiology reading rooms often require the use of only indirect lighting. Consequently, direct light exposure from one display to another is seen as problematic, something to bear in mind during bench testing and clinical trial design.
Badano recommended that when the panel is faced with a PMA for such a system, it should tread carefully where color gamut is concerned. "One of the questions ... is what fraction of color that exists in pathology slides would be outside of the camera and display gamut," he said. Some compression schemes could eliminate information because of inability to fully or accurately display certain colors.
Most displays "are designed to have a color intent for either television or photography," Badano remarked, and hence are biased. "Skin color, sky blue and ocean blue" are favored by displays geared for use in such systems and consequently may distort images of biopsy samples, he said.
The viewing angle issue for LCDs introduces an odd twist to the color shift issue, Badano observed. "The image you see changes as you move around," because while each color tends to shift, "not all the colors move in the same direction, which means it is possible to have color inversions" at different viewing angles, he said.
Mark McCarty, 703-268-5690