In today's medical devices, grayscale and color images often get mixed in multimodal presentations of information. For example, physicians often benefit from seeing color-coded images—such as those from positron-emission tomography (PET) or single photon-emission computed tomography (SPECT)—superimposed on grayscale images: these might be from computed tomography, projection radiography, or mammography. In addition, color capabilities in displays can also be used to display information from detection and classification algorithms. To compare results across different types of device, however, scientists and engineers need a standard approach to characterizing colors.
Although some groups have published standards for or recommendations on how to measure features of colors in medical displays,1–5 these apply to specific kinds of technology: there is no underlying agreement on an overall approach for making color measurements.6 Consequently, we are developing a standardized method to do this. Here, we present preliminary results on different features of color—uniformity, grayscale tracking, and angular shifts—in a variety of medical displays.
Based on an earlier device to measure luminance,7,8 we developed a collimated, color probe (see Figure 1)9 that couples light from the display into an optical fiber. This, in turn, is connected to a spectrometer (MAS40, Instrument Systems, Ottawa, Canada). Unfortunately, the new design decreases the probe's sensitivity to luminance changes, and increases the amount of stray light, because of the larger diameter of the probe tip. For this reason, a new probe is currently being designed that will be closer in sensitivity to the luminance probe and still able to make accurate spectral measurements. We characterized the probe using a method similar to one we have used in the past.7
Figure 1. (a) This color probe gathers light from a display and connects it to a spectrometer via a fiber-optic adapter. (b) This probe can be used to measure light from various displays.
Figure 2 shows color-uniformity measurements taken in three ways: with our collimated probe, and with a spot colorimeter (Minolta CS100, Mahwah, NJ) in both perpendicular and rotated arrangements.9 We also used our probe to measure colors on four different displays: a one million pixel (MP) and 3MP color display, a 5MP monochrome display, and a 1MP televison (see Figure 3). Using error propagation, we computed the error bars. These figures demonstrate that the variability in measuring colors with different methods is similar to the variability found when using one method to measure colors on different kinds of displays. As a result, comparing the color performance of displays can be misleading, because the method affects the measurements.
Figure 2. Color-uniformity measurements were made with the collimated probe (S) and with a spot colorimeter in perpendicular (P) and rotated (R) arrangements.
Grayscale tracking was performed to quantify shifts in color coordinates (u',v') across the grayscale (see Figure 4). The 1MP monitor had the most change in color coordinates. Moreover, the color difference in all of the displays from black to white is larger than the maximum color difference across the screen area (see Figure 3).
Figure 3. Using the collimated probe, the color uniformity on four different displays was measured: a one million pixel (MP) television (1MP TV), 1MP and 3MP color displays, and a 5MP monochrome display.
Figure 4. Grayscale tracking with the collimated color probe shows variation among the displays.
Previous work showed angle-related changes of contrast for medical displays based on liquid-crystal technology.10,11 However, measurements of angular color shifts had not been reported. So, we oriented the probe at 0°, 15°, 30°, and 45° from the screen's normal to quantify the effects of viewing angle on color. Figure 5, for example, shows the changes in (u', v') as the viewing angle changed from 0° to 45° for the 5MP display. There is no direct correlation between the change in (u', v') and the direction of the angular shift.
Figure 5. On the 5MP display, color coordinates (u', v') change with different angles—V is vertical, H is horizontal, and D is diagonal—between the probe and the screen.
Our results suggest using caution when comparing display devices because measured parameters are not consistent across methodologies and device technologies. Factors that affect these color-measurement methods include angular dependance of the spectral emissions and—perhaps—lens flare and stray-light contamination. With new designs of the color probe, we hope to eliminate some of these problems.
Anindita Saha , Aldo Badano
Division of Imaging and Applied Mathematics, Center for Devices and Radiological Health Of ficeof Science and Engineering Laboratories
Anindita Saha received her BS in bioengineering from the University of Pittsburgh in 2005 and currently works on medical-display characterization. She is a member of the Biomedical Engineering Society and Tau Beta Pi.
Aldo Badano is the director of the Imaging Physics Laboratory in the Division of Imaging and Applied Mathematics, Office of Science and Engineering Laboratories, Center for Devices and Radiological Health, US Food and Drug Administration. He leads a program on the characterization and modeling of medical-image acquisition and image-display devices using advanced experimental and computational methods.