This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Quality Control of Cathode-Ray Tube Monitors for MedicalImaging Using a Simple Photometer
David M. Parsons, Yongmin Kim, and David R. Haynor
KEY WORDS: Quality control, picture archiving andcommunication system workstations, cathode-ray tubemonitors, medical image display.
M ANY MEDICAL facilities use computermonitors for consultation and, in some
instances, diagnosis. Therefore, monitors shouldbe required to meet quality-control standards atleast as stringent as those applied to film-basedradiography. For example, a simple smudge onthe glass of the monitor caused by someonetouching the screen could be mistaken for anabnormality. Poor focus of the electron beaminside the cathode-ray tube (CRT) could blurdetails to the point where a hairline bonefracture could be missed. Smudges on the screenand poor electron beam focus are both commonoccurrences that can be fixed, if detected. Thus,a quality-control protocol is necessary to detectproblems before they affect medical diagnoses.This paper addresses this problem by taking the
Fromthe Departments ofElectrical Engineering and Radiology, University ofWashington, Seattle, WA.
Supportedin part by a grantfrom the USArmy.Address reprint requests to YongminKim, PhD, Department
of Electrical Engineering, FT-10, University of Washington,Seattle, WA 98195.
first step toward developing a quality-controlprocedure for gray-scale CRT monitors used forviewing medical images.
Small changes in the characteristics of amonitor over a long period of time are impossible to detect by simple observation. Periodic,quantitative tests are required to detect when amonitor is out of adjustment. Previous research 1,2 has suggested several tests that measure factors that are generally accepted asaffecting image quality. These include gammavalue, maximum luminance, and spatial resolution. We combined these tests into a preliminary quality-control protocol, which we thenevaluated on 10 different monitors that are partof the Medical Diagnostic Imaging Support(MDIS) System at Madigan Army MedicalCenter (MAMC).3
The overall goal of a quality-control programis to ensure optimal display of medically significant information. A quality-control protocolmust also be clinically practical. The tests mustonly require inexpensive equipment and musttake a short time to perform. A few laboratoriesin the country are equipped to fully test CRTmonitors. Full testing is appropriate for monitorcharacterization, but is too complicated andexpensive for routine quality assurance. Theother consideration is the time involved toperform the tests. MAMC is eventually planning on installing over 200 CRT monitors byearly 1995. If each monitor required 20 minutesto test each week, more than one full-timeperson would be required just to perform theCRT quality control measurements. Becausethis is not practical, the tests were designed tobe performed quickly.
Basics ofCRT Operation
This paper is concerned with gray-scale CRTmonitors. Although color monitors are becoming more common for viewing ultrasound andnuclear medicine images, and research is beingconducted into using color for computed tomography and magnetic resonance images, the majority of medical images are still gray scale. Thelower luminance and spatial resolution of color
monitors are further reasons why gray-scalemonitors are preferred. Another competingtechnology is that of flat-panel displays such asliquid-crystal displays. Although flat-panel di~
plays are becoming more popular for use III
laptop and notebook computers, their luminance and spatial resolution are currently muchinferior to those of CRT monitors. CRT monitors will remain as main devices for displayingmedical images in the foreseeable future.
CRT monitors consist of several parts, including an electron source, control grid, acceleration electrodes, focusing and deflection sections, a phosphor screen, and a glass envelopefor containing a high vacuum.' The display isthe portion of the screen that is illuminated bythe image. The neck of the glass tube is wherethe electron beam is generated, accelerated,deflected, and focused at a particular spot onthe screen. The control grid is used to controlthe flow of electrons. Adjusting the voltageapplied to the control grid affects the brightnessof the pixels on the screen. The inside of theface has a phosphor coating that converts theelectron beam energy into visible light. The lightis then transmitted through the glass on thefront of the screen. To increase the contrast ofthe display and reduce the effects of ambientlight, many monitor screens are etched or havean anti reflective coating. This typically reducesthe maximum luminance and spatial resolutionof the monitor.
To approximate the experience of viewingimages on a light box, monitors used for viewingmedical images should be relatively large andflat.' This causes problems with focusing theelectron beam because the corners of the screenare farther away from the neck of the tube thanthe middle of the screen. Advanced monitorsattempt to overcome this problem by dynamically adjusting the focus of the beam dependingon its current location. All of these factors addcomplexity to the operation of the monitor,further emphasizing the need for qualitycontrol tests to detect monitor malfunctions andmisadjustments.
An additional need for quality control arisesfrom the fact that some picture archiving andcommunication system (PACS) workstationshave multiple (two to eight) monitors. Thisallows a radiologist to view different images of
11
the same patient at one time and compare oldand new images if necessary. This also addscomplexity to the quality-control procedure,because an image should look the same independent of the monitor used to display it. If animage looks different on one monitor versusanother, then the ability to compare imagescould be hampered.
Current Research
Most research in the area of CRT testing hasbeen directed at identifying the tests required toaccurately characterize a monitor.v" Characterizing a monitor is a one-time procedure that isintended to definitively measure a monitor'sperformance on many different tests, such asinput impedance and luminance as a function ofCRT beam current. The tests performed formonitor characterization are typically time consuming and require sophisticated analysis toolssuch as a vibration-free bench, video testpattern generator, and charge-coupled devicecamera. Some of the tests require disassemblingthe monitor to measure internal characteristics.This requires special training because there aredangerously high voltages present inside a CRTmonitor. On the other hand, quality-controltests should be simple, so they can be performedfrequently by less highly trained people usinginexpensive equipment.
Other papers have suggested tests that shouldbe performed for CRT quality control.vP butdid not say how often the tests should beperformed or what kind of results to expect. Anearly paper" discusses the use of the Society ofMotion Picture and Television Engineers(SMPTE) test pattern for adjusting the brightness and contrast of displays used for MRI, butdoes not cover other aspects of monitor imagequality. Other researchers'S" concentrate onthe display function (luminance versus grayscale curve) of a monitor, but do not cover testsfor other factors of monitor image quality.
MATERIALS AND METHODS
In this research, tests were performed on 10 CRTmonitors over a period of several months. The bulk of thetests were based on reports from two leading laboratoriesstudying CRT monitor performance: the National Information Display Laboratory at the David Sarnoff ResearchCenter, I and the Monitor Characterization Laboratory at
12
the University of Arizona." This section describes theequipment used and the tests performed.
The 10 monitors that were tested were located in a singleviewing room with controllable overhead lighting in theMAMC Radiology Department. The manufacturer, modelnumber, and addressable resolution for each monitor areshown in Table 1.The monitor types are referred to as A, B,or C in the discussion that follows. The 10 monitors areconnected to three different workstations, with the 4 Amonitors connected to one workstation, the 4 B monitorsconnected to a second, and the 2 C monitors connected tothe remaining workstation. A number after the letterindicates which monitor is being referred to. For example,B-2 refers to the second from the left of the 4 B monitors.
EquipmentThe main equipment used was a 117Luma Color (Tektro
nix Inc, Beaverton, OR) photometer with a 11803 luminance sensor head. The hand-held unit displays the digitalluminance reading to within a tenth of a foot-Lambert (fL;IfL = 3.426 cd/rn- or nit). The attached sensor with anambient-light shield to block external light was placedagainst the monitor screen. Use of a sensor with an ambientlight shield is a distinct advantage in hospital settings suchas an emergency department or intensive care unit where itisnot possible to control or eliminate ambient light. Measurements of a dark monitor screen taken with the overheadlights off and on at full brightness showed a difference ofless than 0.1 fL. A disadvantage of this meter is that it is noteasy to determine exactly what area of the image is beingmeasured.
Figure 1 shows the image, developed by the SMPTEspecificallyfor monitor testing that was used for most of thetests. ll •l7 It has several useful features, including a set of 11different areas of gray-scale intensity ranging from 0% to100% luminance in increments of 10%. There are also 5%and 95% areas located on top of the 0% and 100% areas,respectively. These are used for determining the contrastpresent at the two ends of the luminance scale. Spatialresolution test patterns are located in the four corners andthe center of the image. These consist of six horizontal andvertical modulation gratings. Three of these have a 100%modulation grating at three different spatial frequencies.The other three gratings share the same frequency, but haveintensity modulations of 1%,3%, and 5%, respectively. Thetest pattern also has white on black and black on whitewindows for testing the unit response of the monitor.Finally, a grid that covers the entire background of the testimage can be used to detect spatial nonuniformities. Normally, the pattern is presented so that it just fills the displayof the monitor. However, the workstations at MAMC canalso zoom and pan the pattern.
Most of the tests can be done with the SMPTE test
Table 1. Monitors Used for Measurements
AddressableLabel Manufacturer Model Resolution Number
A Tektronix GMA212 1,536 x 2,048 4
B Image Systems M21P 1,152 x 1,536 4
C Image Systems M24LMAX 1,024 x 832 2
PARSONS, KIM, AND HAYNOR
Fig 1. SMPTE test pattern.
pattern. However, some of the tests can be performed morequickly and accurately using special test patterns that aredescribed in the sections where they are used.
Monitor Preparation
To prepare the monitors for testing, fingerprints and dustwere removed with a glass cleaner and soft cloth. Measurements at MAMC show that these smudges decrease luminance output by as much as 10%. The static charge presenton the screens of the monitors also attracts dust, whichadversely affects image quality.
When a monitor is first turned on after being left off for afew hours, its luminance output can vary. The luminancestabilizes once the monitor has warmed up, which typicallytakes 2 to 6 hours. I Because the monitors at MAMC werealwayson, they did not require any warm-up during testing.
Brightness and Contrast AdjustmentThe first procedure to perform is to adjust the brightness
and contrast of the monitors. According to several papers,9,1l,14.15 the best method for adjusting these values is touse the 5%/0% and 95%/100% areas of the SMPTE testpattern. When both of these patches are just discerniblefrom their background patches, contrast is good across thewhole range of gray-scale values. Although these patchesonly display the contrast at the extremes, the contrast inbetween these two extremes is usually a linear function ofgray scale. Thus, the contrast in the middle should also begood.
At MAMC, all brightness and contrast adjustments aredone by the PACS vendor. During the period when measurements were being taken, the brightness controls of themonitors were periodically adjusted by vendor personnel tomeet specific maximum-luminance requirements. A log waskept of when adjustments were performed so that a correlation to the measurements for this paper could be made.Unfortunately, the brightness and contrast of the monitors
QUALITY CONTROL OF CRT MONITORS 13
(1)
Gamma Value
The luminance emitted by a pixel on a CRT monitor isnot directly proportional to its gray-scale value.l" Brightness(B) is related to the gray-scale value or gray-scale percentage (G) by
were not adjusted using the SMPTE test-pattern methoddescribed above. Instead, the entire display was set at 100%luminance and the brightness was adjusted to meet specificmaximum-luminance requirements established for eachmonitor. In most cases, no adjustments were made to thecontrast of the monitor.
The spatial resolution of a monitor determines how muchdetail a monitor can display. Discerning small details in animage requires high spatial resolution. This is often different from addressable resolution, which is the number oflocations in the video display buffer that can be individuallyaddressed and displayed. Because ofthe Gaussian profile ofthe light emitted by a single pixel on a CRT screen, an idealspacing between neighboring pixels is difficult to determine.When pixels are placed close together, the number ofresolvable pixels is often less than the addressable resolution.
The modulation transfer function, originally developedfor photography-" can be used to quantify the resolution ofa monitor.v'? However, it has drawbacks to be used inquality control. It requires measuring the luminance of themonitor in small steps, which is time-consuming and requires complicated and expensive equipment, including aslit photometer and vibration-free bench. An alternativemethod is to use the frequency response of the human eye toestimate the spatial noise and square wave response (contrast transfer) function of a monitor." This requires severalmeasurements from several observers, which would not bepractical for routine quality assurance. Because of thesedrawbacks, a qualitative method was chosen for evaluatingthe monitors.
To assess the focus of each monitor, the spatial resolutiontest patterns in the corners and middle of the SMPTE testpattern were examined and compared. The sharpness of thegrids was qualitatively assessed and recorded. If the transitions between the dark and light areas in the spatialresolution grid were discernible, the grid was judged to besharp, otherwise it was considered to be blurry.
Temporal Luminance Stability
The purpose of the temporal luminance stability test is todetermine how much the luminance output of a monitor
Geometry can be accurately measured with sophisticatedmonitor characterization equipment that detects the position of grid lines displayed on the monitor through a timeconsuming process. This equipment can quantitatively detect distortions that are not visible to the human eye.Instead, we used a qualitative test combined with a simplequantitative test because of the additional expense and timerequired for complex quantitative tests.
To detect distortion, a special test pattern consisting of agrid of evenly spaced white lines was displayed on themonitor. At a distance of at least one meter, the test patternwas examined to determine if the outside edges werebowing inward or outward. To detect disproportionatehorizontal or vertical stretching, the full height and width ofthe test pattern at the midsection were measured with aflexible, transparent, plastic ruler. This was not a straightforward operation because the actual image is a few millimeters behind on the back surface of the glass. Incorrectsighting of the edge of the test pattern can cause errors inthe measurements. Local distortions in the image are alsopossible. This can be caused by nearby magnetic devices orstatic buildup on the screen. We examined the grid testpattern and looked for areas that did not appear square.
Spatial Resolution
(2)10g(B) = '( 10g(G) + C,
which is equivalent to
where C is a constant and '{ is the gamma value of themonitor that characterizes the relationship between brightness and gray-scale value. The gamma value can be found bymeasuring the luminance output for several different grayscale values, and then computing the best linear fit of thelogarithm of the luminance values to the logarithm of thegray-scale inputs. The slope of the linear fit is the gammavalue.
To measure the luminance output, the SMPTE testpattern was used. However, readings from the darkerpatches were affected by veiling glare. This is caused by lightscattering in the glass of the screen." To reduce this effect,the SMPTE test pattern was zoomed as large as possibleuntil a single square nearly filled the display area. Toincrease the accuracy of the measurements, they were alltaken from the same location on the screen. Thus, thephotometer was kept stationary for each reading. TheSMPTE test pattern was panned to read the differentgray-level patches at the same physical location on thescreen. This reduced the effect of spatial nonuniformities onthe readings.
Maximum Luminance
The maximum luminance output from a typical x-ray lightbox is around 500 fL.8 Most monitors are currently capableof emitting less than 100 fL. This means that less contrastinformation can be conveyed through a monitor. Thus, it isimportant to keep a monitor as bright as possible. However,if a monitor is too bright, spatial resolution will be reducedand intense localized heating may occur resulting in damageto the screen." We used the luminance at the center of thedisplay of the 100% square of the SMPTE test pattern,measured in the gamma test, for the maximum-luminancetest.
Geometry
Geometry refers to the positioning of pixels on the screen.If pixels are not displayed in their proper position, theimage will be distorted. There are many types of geometricdistortion with varying causes. The most well known are pincushion and barrel distortion. Pin-cushion distortion occurswhen one or more sides of the image are bowed inward, andbarrel distortion occurs when the sides are bowed outward.
14 PARSONS, KIM, AND HAYNOR
changes with respect to time after an image is first displayedfollowing a long period of displaying a blank screen.Immediately after displaying a zoomed version of the 100%square of the SMPTE test pattern, measurements of theluminance output were taken every 5 seconds for the first 30seconds. During the next 30 seconds, measurements weretaken every 10 seconds.
SpatialUniformity ofLuminance
Spatial nonuniformity of luminance refers to the variationin luminance output of the screen as a function of location.Because of unavoidable variations in the phosphor coatingon the screen, there will always be small variations inluminance.'
The 100% gray-level square of the SMPTE test patternwas zoomed so that it filled the display area. The screen wasdivided into nine squares, and measurements of the luminance output were taken at the center of each of thesesquares. Next, the screen was examined to determine ifthere were any obviously light or dark areas. Additionalmeasurements were taken at these locations.
Another important characteristic of a monitor is highfrequency noise.9,21,22 Characterizing the noise of a monitorrequires special measuring equipment which is too expensive and time consuming for a practical quality-controlprotocol. For this reason, it is neither covered by this papernor included in our tests.
quantify the results. All luminance readingswere taken in units of foot-Lamberts.
Gamma
The gamma values for the four A monitorsover 11 weeks are plotted in Fig 2. The resultsfor the Band C monitors were similar. Over thisperiod, the brightness and contrast controls ofthe monitors were not adjusted except betweenAugust 31 and September 10. The large drop ingamma value for the A-4 monitor on September22 was caused by a change in the method usedfor taking measurements. The luminance of theA-4 monitor drops steadily during the first fewminutes after an image is displayed. Before thisdate, the monitor was not allowed to fullystabilize when an image was first presented.This was a quicker method of collecting data,but the results were not repeatable. Figure 3shows how the display functions vary when thegamma value is different. To produce similarluminance outputs at all gray levels, neighboring monitors should have similar gamma valuesas well as similar maximum luminance output.
(3)
CO) CO) CO) CO) CO) CO) CO) CO) CO) CO)
Q? Q? Q? Q? Q? Q? Q? Q? ~ Q?C\l Ol !!1 C\l Ol ~ 0 I!) ~~ ~
~ >:? ~co 00 00 Oi Oi0
I'- I'- co Ol
Fig 2. Plot of gamma values for A monitors. I_I, A·1; (el,A·2; (AI,A·3; (+), A·4.
.. ----o: -- - ./. - --~ --- -- -- ....
\
".-~
3
2.8
2.6
2.4Ql
~ 2.2>~ 2~ 1.8CJ
1.6
1.4
1.2
Maximum Luminance
Figure 4 shows the maximum luminance readings taken from the A monitors. The suddendrop on September 22 of the A-4 monitor'smaximum luminance is again caused by the factthat beginning on this date, the monitor wasallowed to stabilize after the test pattern wasdisplayed. This graph shows that the monitorsare not well matched. When the monitors were
RESULTS
This section presents the data collected fromthe tests of the previous section. It describeshow the data was analyzed and discusses theresults for each test. Where possible, a singlenumber was computed for each monitor to
where L o is the luminance of the 0% square, L JOO is theluminance of the 100% square, and Lblank is the luminanceof the blank screen. Another useful measurement is practical dynamic range which is computed by dividing L /00 by Lo.
Veiling Glare
When light is emitted from the phosphor layer, it entersthe glass of the screen. Although most of the light istransmitted straight through the glass, some is scatteredcausing the spatial resolution to be lowered. Veiling glare isa measure of how much light is scattered by the glass of theCRT screen.' Because spatial resolution is an importantfactor for viewing most radiologic images, it is important tomeasure the veiling glare of the monitor. While a nonzoomed SMPTE test pattern was displayed, luminance readings were taken at the 100% and 0% gray-level squares.Next, the luminance of a completely blank screen wasmeasured.
The veiling glare data is expressed as a percentage usingthe following formula:
VG = Lv - Lblank X 100,L IOO - Lblank
QUALITY CONTROL OF CRT MONITORS 15
luminance of the A-2 and A-3 monitors of 8.1%and 7.6%, respectively. Over the course of 11weeks, this change was not perceptible. Thisshows that without periodic testing, over severalmonths a monitor may emit a much lower orhigher luminance than desired.
Spatial Resolution
Qualitative evaluation with the SMPTE testpattern showed that of the three types ofmonitors, the B monitors were the most out offocus, with the focus at the screen corners being
Screen Geometry
Data was collected for all three screen geometry tests. None of the monitors were found tohave significant pin-cushion, barrel, or localdistortion. Height and width measurements weretaken on all 10 monitors at two different timeswith a 3-week separation. The height versuswidth ratios are plotted in Fig 5. Two observations can be made from this graph. The first isthat all but one (B-1) of the monitors fall within10% of a one-to-one ratio. The second observation is that the ratio of height to width does notchange significantly over a period of 3 weeks.
Comparing text displayed on monitors B-1and B-2, two observers were able to detect adifference in the ratio of height to width. Although this does not necessarily indicate thatdiagnostic image quality is being compromised,it does show that a 10% variation in the heightversus width ratio is noticeable.
II :j
) r;AV
~V /
... rI""'". ~ ......
:".
60
50
~Ql 40oc:ctlc:
30'E:::l
--l
20
10
oo 10 20 30 40 50 60 70 80 90 100
Percent gray level input
70
Fig 3. Plot of luminance versus gray-level input for Amonitors on October 8. I_I, A-1; leI, A-2; (.1,A-3; 1+1, A-4.
compared by simply viewing them, there was anoticeable difference in luminance from the A-4monitor, but the others appeared about thesame even though they varied by up to 15 fl.
As described later, there are variations acrossthe surface of the screen in the luminanceproduced. Thus, to get exactly repeatable results, the luminance must be read from thesame position on the screen each time. Becausethis is impractical in a clinical setting, somevariation in the data is expected.
It can also be seen from this chart that themaximum luminance of a monitor, in the absence of adjustments, is a relatively stable value.However, there is a small, steady decline in the
Fig 4. Maximum luminance of A monltore. I_I, A-1; leI,A-2; (.1, A-3; (+), A·4. Fig 5. Screen geometry. I_I, 9/24/93; (ffill,10/15/93.
16
worse than that in the center. All four wereblurry at the top of the screen, and B-2 was alsoblurry at the bottom. The A monitors were allsharp except for A-4, which was very blurry overthe entire display. The C monitors were bothsharp. The spatial resolution did not changenoticeably over several weeks.
Temporal Luminance Stability
Figure 6 shows the results of the temporalluminance stability test. The readings are normalized so that the measured value at time zerois 100%. Notice that all of the monitors remainwithin 5% of their original luminance level after60 seconds, except for the A-4 monitor. Theluminance of this particular monitor decreasesfor over 4 minutes from an initial value of 63.2fL to around 20 fL. This is clearly unacceptableperformance because the monitor is specified todisplay 60 fL of luminance.
SpatialUniformity ofLuminance
The coefficient of variation (standard deviation of the nine measurements divided by themean) gives an approximate value for theamount of luminance variation present in thedisplay. The ideal value is 0%. Figure 7 showsthe coefficient of variation for each monitor onthree separate dates. All of the monitors performed better than 10%, but even in the worstcase these nonuniformities are not detectableby mere visual inspection. From our experience,a coefficient of variation of 10% corresponds
PARSONS, KIM, AND HAYNOR
10
9
8c0 7~.~
6>15 5'EQ)
4'(3
:EQ) 30o
2
oA-1 A-2 A-3 A-4 B-1 B-2 B-3 B-4 C-1 C-2
Monitor
Fig 7. Spatial uniformity coefficient of variation. Readingswere not taken for the A-4 monitor on September 15 becauseof the instability of the luminance with respect to time. I_I,9/15/93; (lillI,9/22/93; ([!jl, 10/13/93.
approximately to a 25% range in luminance of amonitor. Figure 7 also shows that the coefficientof variation does not change appreciably over al-rnonth time period.
Veiling Glare
Figure 8 shows the veiling glare percentagescomputed with equation 3 for all 10 monitors ontwo separate dates. Similar model monitorsproduce similar results. This is expected because the veiling glare is caused by the glass ofthe screen, and each type of monitor has different screen characteristics. The A monitors did
_ .I.or;
I":<,
<,<,
<,<,
<,
110
105OJg 100
'".~ 95~
--l 90Cii°a 858 8015'E 75OJ~OJ 70lL
65
60o 10 20 30 40
Time (sees)50 60
2.5--.----------------
Q) 2+------J...'if-------------1OJ~cQ)
~ 1.5a,
e'"aOJ,§'iii> 0.5
oA-1 A-2 A-3 A-4 B-1 B-2 B-3 B-4 C-1 C-2
MonitorFig 6. Temporal luminance stability, percentage of original
not perform as well compared with the B or Cmonitors. Between September 22 and October15, all of the values increased slightly. Theaverage practical dynamic range measurementsfor the A, B, and C monitors were 52.9, 349.8,and 82.8, respectively.
A QUALITY-CONTROL PROTOCOL
This section presents our initial recommendation toward a quality-control protocol for medical displays. The procedure for performing eachtest is presented along with the method foranalyzing the results. The recommendations fortolerance levels in the measurements are notspecific, because the tests have only been performed on three types of monitors. The recommended frequency values are based on theResults section.
A few recommendations apply to all of thetests. Each time the test pattern is changed, theluminance output should be observed for a fewseconds to make sure the monitor has stabilizedbefore taking any readings. For each test, asingle method should be chosen and adhered toas much as possible. Many photometers haveserial output capability. This can be used todirectly enter the data into a spreadsheet via aportable computer, significantly reducing thetime required to record and analyze the measurements.
Equipment
Because of its ease of use and relatively lowcost, the most important piece of equipment forthese tests is a photometer. The sensor shouldbe shielded from ambient light. The anglebetween the sensor and the screen should benearly perpendicular when readings are taken.Most of the tests use a SMPTE test pattern.Some tests can be performed more quickly andaccurately with additional test patterns. Theseare described below in the test in which they areused.
Monitor Preparation
Before performing any tests, the monitorshould be cleaned with a glass cleaner and softcloth. For frequently used monitors, this shouldbe done as frequently as once per week. Fingerprints are easier to see if a light image isdisplayed on the monitor. The monitor should
17
be turned on for more than two hours, with orwithout displaying an image, before the testsare performed to ensure that it is properlywarmed up.'
Brightness and Contrast Adjustment
The brightness and contrast controls shouldbe adjusted so that the 5% and 95% areas of theSMPTE test pattern are discernible from theirbackground patches. When properly adjusted,the luminance of the 100% square should notexceed the maximum luminance specified forthe monitor, the dark area surrounding the testpattern should not be easily distinguishablefrom the outer edge of the screen, and brightareas should not be out of focus. If any of theseoccurs, the brightness and possibly the contrastsettings need to be reduced.
The brightness and contrast adjustments arethe two most easily changed values of a monitor,and they also have the most profound effect onimage quality.11 Thus, adjustment should beperformed weekly.
Gamma
For this test, the SMPTE test pattern can beused. It should be zoomed so that one squarefills the display area. The pattern can then bepanned to read each square. More accurateresults will be obtained by using 11 differentuniform fields set to gray levels that are 0% to100% of maximum by increments of 10%. Aluminance meter is used to measure the luminance output from each gray-scale value from0% through 100%. The gamma is then computed from these measurements.
Once the gamma value is determined, itshould be compared with previous values obtained for the monitor in addition to the gammavalues of the other monitors at the same workstation. Large changes in the gamma value fromweek to week or from monitor to monitorshould be investigated by checking the brightness and contrast settings.
Maximum Luminance
A luminance meter is used to measure theoutput at the center of the display from thezoomed 100% square of the SMPTE test pattern or from a uniform field at 100% gray level.If the measured luminance value deviates from
18
the specification of the monitor by a largeamount, or if the luminance emitted by differentmonitors at the same workstation differs, thebrightness adjustment should be checked. If theexternal brightness control is not sufficient foradjusting the maximum luminance, there isusually a beam cutoff control inside the monitorthat can be adjusted by a technician. A monthlytest will detect any long-term degradation inperformance.
Geometry
To check the geometry of the display, a gridof evenly spaced lines should be displayed.Alternatively, the background grid of theSMPTE test pattern can be used. While viewingthe display from a distance of one meter, thegrid should be examined to see if the linesappear to be straight. If the grid is unusuallydistorted, the area around the monitor shouldbe checked for magnetic objects. If the sides ofthe outside border appear bowed inward oroutward, the monitor needs adjustment. Highquality monitors typically have an internal control for this effect. While still displaying the testpattern, the width and height of the outsideborder should be measured. Dividing one measurement by the other, the resulting value shouldbe close to 1.0. This distortion can be controlledby a control on the inside or sometimes on theoutside of the monitor.
SpatialResolution
While displaying the SMPTE test pattern, thebar patterns in the middle and corners of thedisplayshould be examined. If they are unacceptably blurry, the focus of the monitor may needto be adjusted. This is usually an internaladjustment. It is usually difficult to have thecorners and the middle of the screen perfectlyfocused at the same time. If this is the case, anintermediate setting should be chosen thatslightly favors the center of the screen, becausethis is where the area of interest is usuallylocated. Because the focus does not appear tochange noticeably over a few weeks, this isrecommended as a monthly test.
Temporal Luminance Stability
On a monitor that is already warmed up buthas been blank for over 10 minutes, a large
PARSONS, KIM, AND HAYNOR
white area should be displayed on the screen,such as the 100% square of the SMPTE testpattern. Because the monitor must be blank fora period of time, this test should be performedbefore the others. A luminance reading shouldbe taken immediately after the white area isdisplayed, and then again after 30 seconds. Ifthe luminance increases or decreases by a largeamount, the monitor should be checked. Thisattribute does not typically change month tomonth, so it is recommended as a quarterly test.
SpatialUniformity ofLuminance
A solid test pattern that covers the area of thedisplay is required for this test. A gray level of100% will show the most variations because ithas the highest luminance output, but a graylevel of 50% represents a more typical viewingcondition. The important characteristic is thatthe entire display is the same gray level. Thedisplay is preferably divided into sixteen squares,four across and four down for a larger numberof data points. The luminance at the center ofeach square is then measured. Additionally, anyareas that appear light or dark should be measured. The coefficient of variation is computedby dividing the standard deviation of thesemeasurements by the mean. Large values shouldbe investigated. A quarterly check is sufficient,as this characteristic does not change much withtime.
Veiling Glare
If the ability to create special test patterns isavailable, two images should be created. Thefirst pattern should consist of a small blacksquare, 1 em per side, in the middle of thescreen, surrounded by a 100% square, 7 em perside, surrounded by a black border. The luminance of the center small black square is measured (Lo). The other test pattern is the sameexcept that the center black square is changedto the 100% gray level. The luminance outputfrom the center of the 100% square is measured(L 100) . If special test patterns are not available,the luminance output of the 0% and 100%squares of a regular (unzoomed) SMPTE testpattern can be substituted for Lo and L 100,
respectively. Finally, the luminance output of ablank screen is measured (Lblank)' The veilingglare percentage (VG) is computed using equa-
QUALITY CONTROL OF CRT MONITORS 19
Table 2. Frequency and Measurement Times (per monitor) forQuality Control Tests
CONCLUSION
Like most analog equipment, the characteristics of a CRT monitor vary with time. BecauseCRTs are being used to view medical images,they must be tested periodically. Our long-termgoal is to develop a clinically important yetpractical quality-control protocol for maintain-
tion 3. Typical values for this measurementrange between 0% and 3%. Larger values mayindicate a problem with the monitor. Becauseveiling glare is caused mainly by the glass of thecomputer screen.? then this value should notchange much over time. Thus, this test can beperformed quarterly. Another useful measurement which can be derived from this test is thepractical dynamic range. This is computed bydividing L]()o by L;
Measurement Time
Table 2 presents the approximate time required to perform the measurements for eachtest. All of the weekly tests combined requireabout 4.5 minutes per monitor, the monthlytests take 1 minute per monitor, and all of thequarterly tests combined require 3 minutes permonitor. Analyzing the results and performingany required corrective actions will take additional time. These tests can be arranged suchthat the monthly and quarterly tests are performed on the monitors on a rotating basis.Assuming this, all of the tests combined willtake approximately 5 minutes per monitor perweek, excluding data analysis.
ing high image quality from gray-scale CRTmonitors used in medical imaging. As a startingpoint, tests from leading monitor characterization laboratories were performed and refined ina clinical setting over a period of 5 months. Thetests use quantitative measurements, exceptwhere expensive equipment or time-consumingprocesses would have been required. In thesecases, qualitative tests were developed to makethe protocol more practical.
To be implementable in a medical centersetting, our tests were designed to be performedby personnel without specialized training inmonitor testing and to use inexpensive equipment. Because PACS workstations may havemore than one monitor, another importantconsideration was intermonitor variations incharacteristics such as gamma and maximumluminance. Based on the results from performing the tests, a preliminary quality-control protocol was presented with recommendations forhow frequently to perform the tests. Characteristics that change rapidly and that most affectimage quality should be tested weekly. Similarly, the other tests have been categorized asmonthly or quarterly tests. The time required toperform all of the tests combined is about fiveminutes per monitor per week.
FutureResearch
There are several areas where more workcould be done in finalizing the quality controlprocedure. The most useful addition to thiswork would be to further apply and evaluate thequality control tests. Additional studies andexperience applying these tests to many different types of monitors would allow definitivetolerance levels to be established for each test.Also, the recommended test frequencies couldbe adjusted to lower the time required toperform the tests without compromising displayquality.
To decrease the amount of time required fortesting, a more efficient means of recording andanalyzing the data could be developed. TheTektronix 117 photometer has a serial outputport that could be connected directly to acomputer to record the measurements. Customsoftware could then analyze and track the datafor each test. The user would be alerted when amonitor performed outside of prescribed toler-
4515
60
30
60165
15
270
30
9060
180
Time (sec)
Weekly tests
Monitor preparation
Brightness and contrast adjustment
GammaMaximum luminance
Total
Monthly tests
Geometry
Spatial resolution
Total
Quarterly tests
Temporal luminance stability
Spatial uniformity of luminance
Veiling glare
Total
Procedure
20
ance levels. Another possibility for future workis to better quantify the spatial resolution test.This could be done by developing an inexpensive test instrument to quickly measure pixelspot profile or modulation pattern output. Thisinstrument would need to be relatively portableand easy to set up so that it could be carriedaround to each workstation on a regular basis.
PARSONS, KIM, AND HAYNOR
Further testing would need to be performed tojustify the additional expense of the equipment.
ACKNOWLEDGMENT
The authors would like to thank LTC John C. Weiser ofthe MDIS Project Management Office and Mr. Jon Carterof MAMC, Department of Radiology for providing helpfulinput on quality control measurements.
REFERENCES
1. National Information Display Laboratory, David Sarnoff Research Center: Test Procedures for Evaluation ofCRT Display Monitors: Version 2.0. Princeton, NJ, DavidSarnoff Research Center, 1991. (technical report)
2. Blume H, Ji TL, Roehrig H: Physical PerformanceData of the Sony DDM-2802F Monitor. Department ofRadiology and Optical Sciences Center, University of Arizona, Tucson, April 1992. (technical report)
3. Smith DV, Smith S, Bender GN, et al: Evaluation ofthe medical diagnostic imaging support (MDIS) systembased on two years of clinical experience. J Digit Imaging1995 (in press)
5. Haynor DR, Smith DV, Park HW, et al: Hardware andsoftware requirements for a picture archiving and communication system's diagnostic workstations. J Digit Imaging5:107-117,1992
6. Roehrig H, Dallas W, Ji TL, et al: Physical evaluationof CRTs for use in digital radiography. SPIE Med Imaging1091:262-278,1989
7. Roehrig H, Ji TL, Browne M, et al: Signal-to-noiseratio and maximum information content of images displayed by a CRT. SPIE Med Imaging 1232:115-133,1990
8. Blume H, Roehrig H, Browne M, et al: Comparison ofthe physical performance of high resolution CRT displaysand films recorded by laser image printers and displayed onlight-boxes and the need for a display standard. SPIE MedImaging 1232:97-114,1990
9. Roehrig H, Blume H, Ji TL, et al: Performance testsand quality control of cathode ray tube displays. J DigitImaging 3:134-145,1990
10. Gray JE, Stears J, Wondrow M: Quality control ofvideo components and display devices. SPIE Med Imaging486:64-71, 1984
11. Gray JE: Use of the SMPTE test pattern in picture
archiving and communication systems. J Digit Imaging5:54-58,1992
12. Reiker GG, Gohel N, Muka E, et al: Quality monitoring of soft-copy displays for medical radiography. J DigitImaging 5:161-167, 1992
13. Dwyer SJ III, Stewart BK, Sayre JW, et al: PACSmini refresher course: Performance characteristics andimage fidelity of gray-scale monitors. Radiographies 12:765772,1992
14. Bronskill MJ: Experience with the SMPTE test pattern in quality control of magnetic resonance images. SPIEMed Imaging 486:180-184, 1984
15. Nawfel RD, Chan KH, Wagenaar DJ, et al: Evaluation of video gray-scale display. Med Physics 19:561-567,1992
16. Davros WJ, Gaskil JW, Mun SK: Quality assurancefor image display devices in a hospital-wide IMACS network. SPIE Med Imaging 1091:301-304, 1989
17. Society of Motion Picture and Television Engineers:SMPTE recommended practice, RP 133-1986, specifications for medical diagnostic imaging test pattern for television monitors and hard-copy recording cameras. SMPTEJournal 95:693-695, 1986
18. Roehrig H, Seeley GW, Dallas WJ, et al: PhysicalEvaluation of CRT Displays. SPIE Med Imaging 767:717725,1987