s f
ano
cien
0903
acc
and
wor
ring
o pla
heth
of
color with minimal contamination from screen areas outside the measurement spot or from off-
normal emissions. They compare three probe designs: A modified small-spot luminance probe and
two conic probe designs based on black frusta.
Methods: To compare the three color probe designs, spectral and luminance measurements were
taken with specialized instrumentation to determine the luminance changes and color separation
abilities of the probes. The probes were characterized with a scanning slit method, veiling glare, and
a moving laser and LED arrangement. The scanning slit measurement was done using a black slit
plate over a white line on an LCD monitor. The luminance was measured in 1 mm increments from
the center of the slit to
15 mm above and below the slit at different distances between the probe
and the slit. The veiling glare setup consisted of measurements of the luminance of a black spot
pattern with a white disk of radius of 100 mm as the black spot increases in 1 mm radius incre-
ments. The moving LED and laser method consisted of a red and green light orthogonal to the
probe tip for the light to directly shine into the probe. The green light source was moved away from
the red source in 1 cm increments to measure color stray-light contamination at different probe
distances.
Results: The results of the color testing using the LED and laser methods suggest a better perfor-
mance of one of the frusta probes at shorter distances between the light sources, which translates to
less contamination. The tails of the scans indicate the magnitude of the spread in signal due to light
from areas outside the intended measurement spot. The measurements indicate a corresponding
glare factor for a large spot of 140, 500, and 2000 for probe A, B1, and B2, respectively. The
dual-laser setup suggests that color purity can be maintained up to a few tens of millimeters outside
the measurement spot.
Conclusions: The comparison shows that there are significant differences in the performance of
each probe design, and that those differences have an effect on the measured quantity used to
quantify display color. Different probe designs show different measurements of the level of light
contamination that affects the quantitative color determination. DOI: 10.1118/1.3265879
Key words: medical display, color measurements, luminance measurements
I. INTRODUCTION
Color is becoming an important aspect in the characteriza-
tion of medical imaging display devices. Display devices for-
mally utilized for grayscale applications such as computed
tomography, projection radiography, and mammography
benefit from the addition of color for the superimposition of
color-coded images from other modalities
i.e., from nuclear
medicine, PET, or SPECT studies and for the effective ad-
dition of computer aids based on detection and classification
algorithms.
Color properties of displays are becoming of great rel-
evance beyond imaging applications. For instance, in the
emerging digital pathology area, color is affected by a num-
ber of factors including the microscope light source and op-
tics, image acquisition hardware characteristics, image pro-
cessing, display calibration and physical characteristics, and
the software used for image viewing.1 To ensure accurate
representation of the imaged object for a certain clinical task,
it is imperative that consistent color management techniques
based on accurate color measurements be used across the
imaging chain. Proper color management can lead to reduced
variability in the interpretation of images, affecting indi-
vidual patients and increasing the statistical power of clinical
trials.
Several standards and recommendations for display color
measurements are available. Each of these standards are as-Accurate color measurement method
Anindita Saha, Edward F. Kelley, and Aldo Bad
Division of Imaging and Applied Mathematics, Office of S
Devices and Radiological Health, U.S. Food and Drug, 1
Silver Spring, Maryland 20993

Received 8 April 2009; revised 9 September 2009;
published 4 December 2009
Purpose: The necessity for standard instrumentation
and reproducible is the major motivation behind this
methods can yield very different results when measu
or color difference. As color increasingly comes int
color will have to be quantified in order to assess w
purposes. The authors report on the characterization74 Med. Phys. 37 „1…, January 2010 0094-2405/2010/37„or medical displays
a

ce and Engineering Laboratories, Center for
New Hampshire Avenue, WO 62-3116,
epted for publication 27 October 2009;
measurements of color that are repeatable
k. Currently, different instrumentation and
the same feature such as color uniformity
y in medical imaging diagnostics, display
er the display should be used for imaging
three novel probes for measuring display741…/74/8/$30.00

75 Saha, Kelley, and Badano: Accurate color measurement methods for medical displays 75sociated with different parts of display technology while not
necessarily agreeing upon a standard measurement
methodology.2
The necessity for standard instrumentation and measure-
ments of color that are repeatable and reproducible is the
major motivation behind this work. Currently, different in-
strumentation and methods can yield very different results
when measuring the same feature such as color uniformity or
color difference.3 As color increasingly comes into play in
medical imaging diagnostics, the major issues with displays
will have to be addressed and quantified in order to assess
whether the display should be used for imaging purposes.
One of the most significant problems with liquid crystal dis-
plays is associated with visual change in color when the
viewing angle of the display or observer shifts away from the
normal. The decreases in detectibility have been character-
ized in medical displays with respect to grayscale but not
color images.4 Current methods to characterize viewing
angle of displays including photographic techniques5 and
viewing angle measurement spheres6 require major cost and
time without necessarily being accurate and reproducible.
We have developed probes measuring the color in a small
area of the screen that are designed to be unaffected by off-
normal emissions and provide color measurements. We must
first determine how our probe designs will respond to color
under a variety of conditions that include “stress” tests with
directional light sources and tests meant to replicate medical
display emissions.
In this work, we propose three varying probe designs for
measuring color without contamination from screen areas
outside the measurement spot or from off-normal display
emissions while increasing the signal strength into the spec-
trometer. We compare the characterization to a previously
documented luminance probe used in measuring luminance
and contrast of imaging displays.7,8
II. METHODS
Our first color probe is a modification of an earlier lumi-
nance probe Fig. 1
a by adding a lens to connect the light
path to the fiber optic adapter of the spectrometer
MAS40
from Instrument Systems, Ottawa, Canada for color
measurements.9,10 The adaptation for spectral measurements
with the luminance probe did not allow enough light signal
to pass to the optical fiber for accurate measurements to be
made. Even at the highest integration times, results at the
same spot taken successively had widely variable results.
Thus, the color probe A Fig. 1
b was designed with a
larger tip to increase the light signal to the spectrometer. The
lens used for color probe A has a 19 mm focal length and
diameter of 12 mm. One major consequence to the increased
tip size is thought to be additional noise from picking up
stray light and off-normal emissions when measurements are
taken. Thus, we also decided to try an additional design in
color probe B1 Fig. 1
c, which includes a larger lens and
aperture but more baffles to ensure enough signal but colli-
mates the signal to possibly reduce the stray light and off-Medical Physics, Vol. 37, No. 1, January 2010normal emissions that can affect the color and luminance
measurements. The lens for color probes B1 and B2 have a
focal length of 50 mm and a diameter of 25 mm. Color probe
B2 is a modification of color probe B1 with an additional
baffle to decrease stray light contamination into the lens and
blacked edges of the lens, as indicated in Fig. 1
d. Color
probe B2 has an apex angle of 90° compared to the apex
angle of 30° for the other probes.
To compare the three color probe designs, spectral and
luminance measurements were taken with specialized instru-
mentation to determine the luminance changes and color
separation abilities of the probes. Luminance measurements
were done with a high-gain detector
SHD33 from Interna-
tional Light, Peabody, MA with a photopic filter because of
its low noise properties and increased sensitivity to changes
in luminance.
Spectral measurements were done with the same fiber op-
tic connecting the light path to the spectrometer.9,10 In order
to quantify color differences, we chose to represent the spec-
tral information using the u


v


metric. The choice of metric
for color comparison comes from the recommendation found
in the flat panel display measurement standard,11 as well as
in the TG18
Ref. 12 document and in other medical imag-
ing standards currently under development. Spectrometer
measurements have a lower sensitivity particularly when
measuring low gray levels. The absolute expanded uncer-
tainty with a coverage factor of 2 of any single color mea-
surement with any color probe is estimated to be 0.03 in u


and v


because each probe is not separately calibrated. How-
ever, the repeatability is estimated to be 0.001 in u


and v


,
and the uncertainty between any two different color measure-
ments using any single probe in any of our configurations is
estimated to be 0.003 in u


and v



smaller than the data
point sizes used in the graphs of the color coordinates. For
all our relative luminance measurements, two standard de-
viations is estimated to be 2%.
The ability of the probe to detect and measure contrast
and stray light is an important feature for accurate measure-
ments. The probes are characterized and compared to the
results in previous literature with the luminance probe using
the same experimental setup.9 A flat-panel display is set up
with a white line at the center of the display and a black
background on the remainder of the display. A slit plate is
positioned over the display surface with the slit placed over
the white line on the display screen. Black nonreflective pa-
per covers the rest of the display around the slit plate to
prevent reflections from the rest of the screen. The slit plate
is made of a black vinyl plastic with a thickness of approxi-
mately 2 mm. The actual slit
0.1 mm is constructed from
two razor blades tinted black. The probes are measured at
distances of 1 and 32 mm from the tip of the probe to the slit.
Luminance measurements are taken in increments of 1 mm
from the center of the slit to
15 mm above and below the
slit. The experimental setup is shown in Fig. 2.
The measurement of the veiling glare of the two color
probes is compared to that of the luminance probe9 to deter-
mine signal contamination in each of the probes based on