British Journal of Radiology 74 (2001),841-846 © 2001 The British Institute of Radiology
A protocol-based evaluation of medical image digitizers
E P Efstathopoulos, MSc,
L Costaridou, PhD,
O Kocsis, MSc and
G Panayiotakis, PhD
Department of Medical Physics, School of Medicine, University of Patras, GR 26500 Patras, Greece
Correspondence: George Panayiotakis
 |
Abstract
|
|---|
Medical film digitizers play an important transitory role as digital-to-analogue bridges in radiology. Their use requires performance evaluation to assure medical image quality. A complete quality control protocol is presented, based on a set of test objects adaptable to the specification of various digitizers. The protocol includes parameters such as uniformity, input–output response, noise, geometric distortion, spatial resolution, low contrast discrimination, film slippage and light leakage, as well as associated measurement methods. The applicability of the protocol is demonstrated with two types of medical film digitizers; a charge-coupled device (CCD) digitizer and a laser digitizer. The potential value of the protocol is also discussed.
|
Introduction
|
|---|
Although digital imaging represents the current trend in radiology, most of the workload in a radiology department is still performed by analogue systems. Medical image digitizers have been introduced to enable the transformation of analogue radiological images to digital images. There are three main categories of digitizers used clinically: video cameras, charge-coupled device (CCD) based digitizers (either flatbed scanners or portable cameras) and laser scanners. Their role in a radiology department is two-fold. They may contribute to the development of a picture archiving and communication system (PACS) in the department until native digital X-ray capture technologies (e.g. computed radiography) become more common and they may also aid in primary clinical diagnosis through image processing and image quantification. In both cases it is important to ensure that the digitization process does not cause loss of clinically useful information and, therefore, like all medical imaging devices and procedures, the development of a quality control (QC) protocol for scanners is required. Published work regarding evaluation of medical image digitizers has proposed a limited set of parameters to be tested. The most common acceptable parameter is spatial resolution [1–12], usually expressed by modulation transfer function (MTF) [2, 6–12]. Other common parameters considered include input–output response curves (also called characteristic curves), low contrast discrimination [1–6, 8, 9, 13–15], noise [1–3, 6, 10, 12, 15] and geometric distortion [1, 6, 9, 10]. A complete protocol was proposed based on a test film object [16], however, this test object is not adapted to the specifications of the various film scanners.
In this paper, a complete protocol is presented for performing QC, as well as measurement methods, based on a set of film test objects adaptable to the specifications of various film digitizers. The applicability of the protocol is demonstrated with two types of medical film digitizer, a CCD digitizer and a laser digitizer.
|
Materials and methods
|
|---|
Quality control parameters, test film objects and measurements
Selection of the parameters proposed has been guided by relevant literature for QC of digitizers, as well as existing protocols for other medical imaging modalities such as gamma cameras [17] and fluoroscopic equipment [18–20]. As the main objective of this protocol is to assess the degradation of image quality introduced by the digitizer, the following parameters have been selected: uniformity; input–output response; noise; geometric distortion; spatial resolution; low contrast discrimination; film slippage; and light leakage.
Implementation of the protocol is based on two test films. Objects contained in the test films were digitally created using a software tool developed in our department for designing digital test objects for module performance evaluation in medical digital imaging [21]. The objects correspond to ideal mathematical objects. The final quality of these objects on the test film was dependent on the quality of film printing. Test films were printed using an AGFA Drystar 2000 camera (Agfa Gevaert, Antwerp, Belgium). Film printing was tested with respect to input–output response curves and spatial resolution and was found to operate within the manufacturer's specifications [22]. Uniformity and geometric distortion of the printed test object were measured and found to be greater than 99% and less than 0.25%, respectively. The MTF of the printer was not assessed.
Optical density (OD) measurements of films were performed using a densitometer (X-Rite 331; X-Rite, Grandville, MI). The first film had a uniform OD of 1.00 and was used for assessment of uniformity. The second film (Figure 1
) was used to measure all other proposed parameters. The specific test objects embodied into this film for evaluating these parameters are shown in Figure 1.
Test films were digitized and measurements regarding QC parameters were carried out using a software tool developed in our department appropriate for the display and processing of medical images [23].
Uniformity
To assess uniformity of the digitization area, the uniform test film was used. Measurement of uniformity is based on pixel values over 95% of the whole digitization area, considered as the useful field of view (UFOV) and expressed as integral uniformity and differential uniformity. Integral uniformity is the percentage of the difference between maximum and minimum pixel values over the UFOV, divided by the sum of these values. Differential uniformity is the percentage of maximum change of pixel values over a range of five pixels, in both x and y directions, divided by the sum of maximum and minimum pixel values of these five consecutive pixels.
Input–output response
To assess the relationship between OD of the film and pixel values of the digitized image, the step pattern of the test film was used. The test pattern consists of 16 small squares of different ODs. Corresponding OD values are presented in Table 1. The mean pixel value of each step was measured at a rectangular region of interest (ROI) of size 20 pixels x 20 pixels.
Noise
Noise dependence on OD was assessed using the ROIs defined over the previously used step pattern. Standard deviation of pixel values was used to estimate the noise.
Geometric distortion
To assess geometric distortion, a rectangular grid pattern of 2 cm cell size was embodied into the test film (Figure 1
). A central line profile, perpendicular to the grid lines, is drawn across seven consecutive grid cells (Figure 2a). The number of pixels corresponding to 2 cm at the centre, as well as the number of pixels corresponding to 14 cm, is measured. Geometric distortion isexpressed as a percentage of the absolute difference of unity (i.e. 1), minus the ratio of 7 times the number of pixels corresponding to 2 cm divided by the number of pixels corresponding to 14 cm. The measurement is performed in both directions.
View larger version (23K):
[in this window]
[in a new window]
|
Figure 2. Line profile used (a) to measure geometric distortion, (b) to derive square wave response and (c) for low contrast discrimination.
|
|
Spatial resolution
Spatial resolution was expressed by two parameters; pixel size, and MTF(0.05), i.e. the spatial frequency corresponding to 5% of MTF. Pixel size was estimated as the number of pixels between two points of known distance (2 cm), divided by this distance. MTF derivation is based on the square wave response (SWR) using Coltman's equation [24–26]. SWR is derived using a line profile perpendicular to a resolution pattern (Figure 2b). The spatial frequency values corresponding to this pattern are presented in Table 1.
Low contrast discrimination
Low contrast discrimination is assessed using two low contrast discrimination patterns, one containing lighter objects than the background and the other containing darker objects than the background. Each pattern consists of 10 columns with 13 circular objects embedded in each of them. Within the same column the contrast is constant, while contrast changes between columns. Object sizes and contrast are presented in Table 1. For contrast measurements, line profiles across each column of objects are used (Figure 2c). From each line profile plot, the smallest detectable object size is identified, according to a profile amplitude criterion [16]. According to this criterion and for the purposes of this work, a detectable object is considered as an object having at least 10% difference in pixel value from its background.
Film slippage
Film slippage is used to assess uniformity of motion of moving digitizer parts during the digitization process. Film slippage is evaluated by visual inspection of diagonal lines introduced into the test film. All lines should appear straight. Any inhomogeneity in movement during the digitization process will be detected as a non-linear segment in the straight lines of the test pattern.
Light leakage
Light leakage was assessed by measuring the width of the black border of a known nominal width (1 cm) surrounding the test image. This was carried out by taking a line profile perpendicular to the border. The difference between measured and nominal width, expressed as a percentage, is used to express light leakage.
The digitizers
The presented protocol was applied for the QC of two film digitizers for medical applications, a CCD digitizer (Lumiscan 20; Lumisys, Sunnyvale, CA) and a laser digitizer (Lumiscan 75; Lumisys) [27, 28]. The film size that can be digitized by these scanners varies from 7–14 inches in width and from 7–36 inches in length. The standard scan rate is 80 lines s-1 for Lumiscan 20 and 115 liness-1 for Lumiscan 75. According to manufacturer specifications, Lumiscan 75 is of superior performance.
|
Results
|
|---|
A summary of the QC of both digitizers is given in Table 2. According to measurements, Lumiscan 75 is characterized by superior performance regarding uniformity (integral and differential), range of linear response, noise, spatial resolution (expressed both as pixel size and MTF(0.05)) and light leakage. Similar performance is observed regarding geometric distortion, low contrast discrimination and film slippage.
The behaviour of some of the above parameters as a function of input values is presented in Figures 3–5. Figure 3 presents the input–output response of both digitizers. Lumiscan 75 is characterized by superior clinical and linear range; 0.2–3.6 and 0.2–3.2 OD values, respectively. For clinical range we refer to the range of OD values quoted by the manufacturer; the linear range is the part of that range for which the input–output response curve of the digitizer is linear. Figure 4 presents the noise dependency on OD. Both digitizers are characterized by similar behaviour up to 2.4 OD. Above this value, the noise of Lumiscan 20 increases considerably. Figure 5 presents the image transfer function characteristics of the digitizers regarding spatial frequency, expressed by MTF. As the results were not compensated for the non-ideal performance of the printer, the reported MTF is the cascaded function of the printer with that of the digitizer. Image transfer function characteristics are superior for Lumiscan 75 for all spatial frequencies examined.
View larger version (10K):
[in this window]
[in a new window]
|
Figure 4. Noise dependence on optical densities for the Lumiscan 20 ( ) and Lumiscan 75 (•). s.d., standard deviation.
|
|
View larger version (10K):
[in this window]
[in a new window]
|
Figure 5. Spatial resolution (expressed as modulation transfer function (MTF)) of the Lumiscan 20 () and Lumiscan 75(•).
|
|
Measurement of light leakage showed a significant difference between the two scanners, with Lumiscan 75 having superior results.
|
Discussion
|
|---|
The main advantage of the proposed protocol is its completeness and its adaptability to digitizer specifications by means of the proposed test objects. In addition, it is easily applicable if suitable software for digital image display and measurement is available.
This protocol can be utilized for quality assurance purposes. Parameters such as input–output response, noise dependence on OD, and image transfer characteristics, expressed as MTF, offer valuable information not usually provided by manufacturers. Use of such QC parameters in combination with adaptation of parameter ranges in the test object can better guide the selection of a digitizer regarding specific clinical needs. For example, a digitizer with unacceptable linearity of input–output response at high OD values cannot be used for visualization of lung parenchyma in chest radiography, air in abdominal radiography or breast periphery in mammography. This protocol can also be used for evaluation and comparison purposes if limitations arising from the non-ideal performance of the film printer are dealt with.
A limitation of the proposed protocol is a lack of tolerance limits regarding parameter values, as clinical evaluation studies are required for their definition. These tolerance limits are expected to be different depending on the intended clinical use of a digitizer.
As medical image film digitizers still play an important transitory role as analogue-to-digital bridges in radiology, their QC is critical. A major step towards assuring digitized image quality would be provided by the development and adoption of an official digitizer protocol as part of a quality assurance program in radiology. Efforts such as the one presented in this paper could contribute towards this goal.
Received for publication December 14, 1999.
Accepted for publication April 18, 2001.
|
References
|
|---|
-
Esser PD, Halpern EJ, Amis ES. Quality assurance of picture archiving communications systems with laser film digitizers. J Digit Imaging 1991;4:248–50.[Medline]
-
Yin F-F, Giger ML, Doi K, Yoshimura H, Xu X-W, Nishikawa RM. Evaluation of imaging properties of a laser film digitizer. Phys Med Biol 1992;1:273–80.
-
Meeder RJJ, Jaffray DA, Munro P. Tests for evaluating laser film digitizers. Med Phys 1995;22:635–42.[Medline]
-
Teslow TN. The laser film digitizer: density, contrast, and resolution. J Digit Imaging 1997;3(Suppl. 1):128–32.
-
Dawkins TA, Freedman MT, Lo S-C, Mun SK. High resolution (35 micron) CCD-based film digitizer for mammography. Proc SPIE 1993;1897:511–7.
-
Shrout MK, Potter BJ, Yurgalavage HM, Hildebolt CF, Vannier MW. 35-mm film scanner as an intraoral dental radiograph digitizer. I: a quantitative evaluation. Oral Surg Oral Med Oral Pathol 1993;76:502–9.[Medline]
-
Shrout MK, Potter BJ, Yurgalavage HM, Hildebolt CF, Vannier MW. 35-mm film scanner as an intraoral dental radiograph digitizer. II: effects of brightness and contrast adjustments. Oral Surg Oral Med Oral Pathol 1993;76:510–18.[Medline]
-
Döler W, Rassow S, Jäger A, Vosshenrich R. Investigation of the imaging properties of an X-ray film scanner. Phys Med Biol 1994;39:917–22.
-
Lo S-CB, Butson P, Lin J-S, Li H, Freedman MT, Mun SK. Performance characteristics of high resolution charged-coupled device film digitizers. Proc SPIE 1995;2432:470–5.
-
O'Hare NJ, Wallis F, Kennedy JMT, Hickey E, McDermott GJ, Dowling A, et al. Specification and initial evaluation of a multiple application teleradiology system. Br J Radiol 1996;69:735–42.[Abstract]
-
Yin F-F, Giger ML, Doi K. Measurement of the presampling modulation transfer function of film digitizers using a curve fitting technique. Med Phys 1990;17:962–6.[Medline]
-
Davies DH. Technical note: digital mammography — the comparative evaluation of film digitizers. Br J Radiol 1993;66:930–3.[Medline]
-
Lo S-CB, Gaskill JW, Mun SK, Krasner BH. Contrast information of digital imaging in laser film digitizer and display monitor. J Digit Imaging 1990;2:119–23.
-
Siegel EL, Templeton AW, Cook LT, Eckard DA, Harrison LA, Dwyer SJ. Image calibration of laserdigitizers, printers, and gray scale displays. Radiographics 1992;12:329–35.[Abstract]
-
Hangiandreou NJ, O'Connor TJ, Felmlee JP. An evaluation of the signal and noise characteristics of four CCD-based film digitizers. Med Phys 1998;25:2020–6.[Medline]
-
Halpern EJ. A test pattern for quality control of laser scanner and charge-coupled device film digitizers. J Digit Imaging 1995;8:3–9.[Medline]
-
National Electrical Manufacturers Association. Performance measurements of scintillation cameras, NEMA NUI. Rosslyn, VA: NEMA, 1986.
-
Moores BM, Stieve FE, Eriskat H, Schibilla H, editors. Technical and physical parameters for quality assurance in medical diagnostic radiology. Tolerances, limiting values and appropriate measuring methods. Report 18: Proceedings of the 1988 workshop organized by the Commission of the European Communities Directorate-General for Science, Research and Development (Radiation Protection) and Directorate-General for Employment, Social Affairs and Education (Health and Safety); 1988 February 23–25; Brussels. London: British Institute of Radiology, 1989.
-
Moores BM, Wall BF, Eriskat H, Schibilla H, editors. Optimization of image quality and patient exposure in diagnostic radiology. Report 20: Proceedings of the 1988 workshop organized jointly by the Commission of the European Communities and National Radiological Protection Board; 1988 September 27–29; Oxford. London: British Institute of Radiology, 1989.
-
Hendee WR, Rossi RP. Quality assurance for fluoroscopic X-ray units and associated equipment. Rockville: Bureau of Radiological Health, US Department of Health, Education and Welfare, Public Health Service, Food and Drug Administration, 1979.
-
Kocsis O, Costaridou L, Efstathopoulos EP, Lymberopoulos D, Panayiotakis G. A tool for designing digital test objects for module performance evaluation in medical digital imaging. Med Inf 1999;24:291–308.
-
Agfa Drystar 2000–technical specifications. http://www.agfamedical.com/products_services/hardcopy/drystar/drystar_2000_tech.asp
-
Sakellaropoulos P, Costaridou L, Panayiotakis G. An image visualization tool in mammography. Med Inf 1999;24:53–73.
-
Doi K. Basic imaging properties of radiographic systems and their measurement. In: CG Orton, editor. Progress in medical radiation physics, Vol. 2. New York: Plenum Press, 1985: 194–8.
-
Coltman JW. The specification of imaging properties by response to a sine wave input. J Opt Soc Am 1954;44:468–71.
-
Weiss JP. Notes on determining modulation transferdata for X-ray film–screen combinations. Journal of Applied Photographic Engineering 1978;4:82–4.
-
Lumiscan 20, system specifications. Lumisys Inc. 1998. http://www.lumisys.com/support/manuals.html
-
Lumiscan 75, system specifications. Lumisys Inc. 1998. http://www.lumisys.com/support/manuals.html
This article has been cited by other articles:
|
|
|
|
 
I BONIATIS, L COSTARIDOU, D CAVOURAS, I KALATZIS, E PANAGIOTOPOULOS, and G PANAYIOTAKIS
A morphological index for assessing hip osteoarthritis severity from radiographic images
Br. J. Radiol.,
February 1, 2008;
81(962):
129 - 136.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
P. Peloschek, G. Langs, M. Weber, J. Sailer, M. Reisegger, H. Imhof, H. Bischof, and F. Kainberger
An Automatic Model-based System for Joint Space Measurements on Hand Radiographs: Initial Experience
Radiology,
December 1, 2007;
245(3):
855 - 862.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
I S Boniatis, L I Costaridou, D A Cavouras, E C Panagiotopoulos, and G S Panayiotakis
Quantitative assessment of hip osteoarthritis based on image texture analysis.
Br. J. Radiol.,
March 1, 2006;
79(939):
232 - 238.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
