Radiologists who use monitors that are not optimized for viewing DICOM-conformant data may be at risk of missing subtle details in their displays, a research study has concluded. Proper monitor calibration can ensure that radiologists take full advantage
Radiologists who use monitors that are not optimized for viewing DICOM-conformant data may be at risk of missing subtle details in their displays, a research study has concluded. Proper monitor calibration can ensure that radiologists take full advantage of the benefits offered by the DICOM standard.
Among the findings was that one common axiom in digital display - brighter is always better - doesn't hold true. Another conclusion was that too much luminance may increase perception distortion.
"This is the first paper that really verifies in practice the principal visual concept of the DICOM display standard," said Dr. Klaus-Ruediger Peters, an adjunct professor of diagnostic imaging and therapeutics at the University of Connecticut Health Center.
To determine the best implementation of American College of Radiology recommendations on monitors that conform to DICOM display standard 3.14, Peters and colleagues measured contrast sensitivity and contrast linearity of DICOM-conformant monitors in dark room conditions using measurements of "just-noticeable differences" on special contrast test patterns.
Academic measurements in recent years have been done on monitors not optimized for DICOM display, and as a result they have produced surprising results, Peters said. These include similar diagnostic performance for both film and digital display and diagnostic performance that is independent of the luminance of film or soft-copy displays.
"Radiologists will tell you, on the other hand, that there is a big difference between conventional digital displays and film," he said.
The Connecticut study found that display luminance range and video bandwidth had a significant effect on contrast perception.
Peters and colleagues first used conventionally adjusted DICOM-conformant monitors and found that no display settings resulted in 100% contrast sensitivity or linearity.
The researchers reported that they were able to achieve high-quality images only when their displays were calibrated to a contrast resolution of one. That is, with a video bandwidth fixed at 8 bits, the researchers adjusted their monitor's luminance range so that it could accommodate exactly 256 just-noticeable differences. They then calibrated this range with the DICOM gray-scale standard display function. Using this display setup, which they termed a "natural display mode," Peters and colleagues achieved 100% contrast linearity and sensitivity. The study was published in the September 2002 issue of Academic Radiology.
The researchers also found that the appropriate monitor calibration eliminated Mach effects. Mach banding occurs with high contrast. In a high -contrast scenario, the brain can be disturbed in its perception of neighboring contrasts, Peters said. Mach effects can be so magnified by improperly calibrated monitors that primary diagnoses obtained from improperly calibrated monitors should be banned, he added.
The appropriate calibration discovered by the researchers, however, contradicts previous recommendations that called for very bright monitors.
"The kind of adjustment radiologists need to make will require dimming the screens on conventional 8-bit monitors. They will be able to see every single gray-level step of their displayed data," Peters said.
The next step in maximizing the benefit gained from the DICOM standard is to study how useful the natural perceptual information and its contrast range really are. By documenting the optimal perceptual contrast range for every diagnostic task, radiologists could take advantage of a standard range of intensity steps and could choose which level is appropriate for each task.
"This is why the DICOM display standard is so helpful - it guarantees a reproducible standard display," Peters said. " Our study showed that we can achieve 100% natural performance in display data perception every single time."