1992, The British Journal of Radiology, 65, 686-690
An alternative approach to contrast-detail testing of X-ray image intensifier systems By C. J. Kotre, MSc, ACGI, N. W. Marshall, BSc, ARCS, and K. Faulkner, PhD, ARCS Regional Medical Physics Department, Newcastle General Hospital, Newcastle-upon-Tyne NE4 6BE, UK (Received 26 June 1991 and in revised form 14 November 1991, accepted 14 January 1992) Keywords: Fluoroscopy, Image quality, Quality assurance
Abstract. The difficulties of making the results of threshold contrast-detail diameter tests on X-ray image intensifier systems consistent with published performance standards are discussed. The current approach to contrast-detail testing is described and an alternative method intended to give greater consistency for all image intensifier input field diameters proposed. The current and alternative test conditions are compared on two image intensifier systems. The results obtained show that the contrast-detail curves for image intensifier systems with a wide range of input field diameters can be effectively normalized to be directly comparable to a common reference standard by applying the proposed alternative test conditions. The implications of this result on the interpretation of the contrast-detail test are discussed.
Threshold contrast-detail diameter test objects (specifically the "Leeds" test objects) have been a valuable tool for the non-invasive assessment of the image quality of X-ray image intensifier/TV systems since their introduction in the early 1960s (Hay, 1964). Although intended to provide the means for performing serial quality assurance measurements on individual systems, the use of these test objects has often been extended to the distinctly separate task of making measurements of image quality over a wide range of systems. This extension of function, although discouraged in the instructions supplied with the Leeds test objects (Hay et al, 1979), has been necessitated by the need to provide assessments of image quality as part of the acceptance/ commissioning tests on new image intensifiers (HPA, 1981; DHSS, 1982; Henshaw, 1989) and to give advice on the replacement of old ones. The current approach adopted to assess image quality in these situations is to perform contrast-detail tests at pre-determined dose rates, X-ray beam quality and viewing conditions, and to make a comparison between the resulting contrast-detail curves and one of a number of standard curves obtained from surveys of well maintained equipment. The Department of Health and Social Security (now Department of Health) document STB/7/82 (DHSS, 1982) shows the results of two such surveys (for intensifiers with 22-25 cm input field diameters), one for new equipment and one for equipment of average age 5 years, both at an input dose-rate of 0.26 /iGy/s. It recommends that the performance level described by the latter curve be used as a reasonable guideline for the performance of new image intensifier systems. This advice was repeated in the subsequent second (1985) and third (1986) editions of this document together with an additional section concerning the application of the Leeds test objects to image intensifiers with larger input field diameters (up to 35 cm) than the 25 cm diameter maximum for which the objects were 686
originally designed. The dose rates recommended for these larger field diameters are, however, manufacturerdependent and no expected performance curve for these larger intensifiers is given. The document also includes the recommendation from a group of equipment manufacturers that, where performance assessments of larger input field diameter intensifiers are to be performed, the magnified field giving the closest to a 23 cm effective input field diameter should be used for comparison with the standard survey result. This approach, however, does not solve the problem of comparing the performances of systems whose maximum field diameters are less than 23 cm. A curve describing the average performance of smaller field-of-view image intensifiers at 0.87 juGy/s has been published in the Leeds test objects instruction manual, and this is often used as a guideline for acceptable performance of these smaller diameter systems, although the DHSS documents do not recommend this. The Hospital Physicists' Association publication TGR32 (HPA, 1981) gives a range of typical results for input dose-rates of 0.26 nGy/s and 0.87 //Gy/s but does not state for what input field diameter the results apply. The present approach to contrast-detail testing
The approach to contrast-detail testing embodied in the design and instructions for use of the Leeds test objects is to test the image intensifier system under viewing conditions similar to those used by the radiologist in the clinical situation and at clinically realistic dose-rates. When most X-ray image intensifiers were either of single-field 15 cm or dual-field 23/12 cm designs, it was possible to deal with most equipment encountered in two categories: input field diameters of 23-25 cm tested at 0.26• [iGy/s, and 12-15 cm tested at 0.87 fiGy/s. The recommended viewing distance for both categories is four times the diameter of the TV monitor display. At this distance the visual bias of the The British Journal of Radiology, August 1992
An alternative approach to contrast-detail testing
contrast decreases with viewing distance as spatial averaging over the sampling aperture of the observer acts to reduce the perceived level of noise, but for small diameter details, the threshold contrast increases with viewing distance owing to the decreasing ability of the eye to resolve them. There is, therefore, no single viewing distance that is optimal for all detail sizes. It is important, however, that a fixed proportionality between the displayed size of the test object and the viewing distance is maintained so that the psychooptical bias inherent in the contrast-detail test is also kept constant. The conventional viewing distance of four times the diameter of the TV display was originally chosen so that a 625-line TV raster is just not perceptible to a person with normal eyesight (Hay et aJ, 1985) but this restriction is probably now less important An alternative approach If the rationale behind the choice of test conditions is because of improvements in TV monitor design and the altered from an attempt to simulate clinical conditions increasing use of high-resolution 1000-line TV systems. to an attempt to obtain a quantitative measurement of Whilst the limit of visibility of TV raster lines with image intensifier/TV performance, an alternative test distance has been investigated (Siedband, 1981) it is not protocol can be formulated. Under this protocol it is clear that their perception or otherwise has a significant proposed that the input dose-rate and the visual para- effect on the detection of low-contrast disks in noise. meters affecting the observer's ability to detect the test The display magnification (i.e. the magnification details in noise are maintained constant for all input factor between the size of the image displayed on the TV field diameters. This offers the advantage that the results monitor and the original size of the object) and TV can be compared directly with the DHSS guideline, but screen diameter reported (DHSS, 1981) for a new GEC suffers from the disadvantage that the dose-rate used for Kompact 9/5 intensifier/TV system, typical of the equipthe test may be different from that being used in clinical ment on which the 1982 DHSS guideline survey was practice. Thus, if the constrast-detail measurement is originally performed, were used to establish a suitable intended to form part of a quality assurance pro- standard viewing distance. The reported display magnigramme, the conventional test protocol should be used fication measured at the centre of the 23 cm field at clinically realistic dose-rates. (1.025:1) was related to the conventionally recomThe X-ray quantum noise present in the image is mended viewing distance of four times the display circle governed by the dose-rate at the intensifier input plane. diameter (23 cm) to establish a standard viewing To make this comparable to the DHSS guideline it distance for an unmagnified display of 90 cm. The proshould be set to 0.26 ^Gy/s. Since the gain of an image cedure for setting the viewing distance is therefore to intensifier is directly proportional to the ratio of its measure the display magnification for the field size input and output phosphor areas, an implication of the under test and multiply this factor by 90 cm. The use of a fixed dose-rate for all intensifies is that the display magnification can be obtained by measuring the automatic gain control of the video system must be distance on the TV monitor screen across a suitable capable of maintaining the image brightness and video number of squares on the Leeds M1 matrix test object modulation at this dose-rate without increasing the and dividing by the actual distance as defined by the contribution of electronic noise to the extent that the object. This method of setting the viewing distance has the additional advantage that any geometrical scaling quantum noise no longer dominates. To minimize variations in the psycho-optical para- caused by the test object not being at the image intensimeters involved in the contrast-detail detection task, the fier input plane will be automatically corrected for. critical variable to be considered is the viewing distance. Other important factors include the ambient room Experimental measurements lighting and TV monitor contrast and brightness adjustThreshold contrast-detail diameter tests with the ment. These should, however, be reasonably constant Leeds test objects N3 and TO 10 were carried out on two for the task of contrast-detail testing where the room new image intensifer/TV systems to compare the lighting is normally dimmed in line with clinical practice conventional test protocol with the alternative protocol and the TV monitor controls are pre-set using the Leeds proposed above. In all cases a measured tube potential grey-scale test object GS1. (The appearance of the GS1 of 70 kVp and an additional filtration of 1 mm of image at 0.26 ^Gy/s can also be used to establish that copper were used. These parameters cannot be altered the dynamic range of the automatic video gain system without invalidating the calibration of the contrasts has not been exceeded.) given by the test object. The systems had input field As the viewing distance is changed, the threshold diameters of 30, 23 and 17 cm and 36, 25 and 15 cm. In contrast for various sized details will also change. In each case visual assessments of the contrast-detail test general, for large diameter details, the threshold objects were made by two experienced observers. observer is expected to be similar to that operating in clinical practice. Cowan et al (1987) suggest that this is an advantage of the conventional test protocol, but while this may be so when the, desired result is a simulation of clinical conditions, it will be shown that where the desired result is a quantitative measurement of performance, variations in visual bias with input field diameter can produce misleading results. The limitations of the present approach emerge when attempts are made to compare the results obtained with the DHSS reference standard. Such comparisons are invalid unless the same test conditions (input field diameter of 23 cm and input dose-rate of 0.26 /^Gy/s) have been used.
Vol. 65, No. 776
687
C. J. Kotre, N. W. Marshall and K. Faultier
To test the assumption that the threshold contrast is quantum noise dominated at the proposed standard dose rate of 0.26 fiGy/s over a range of intensifier sizes, threshold contrast was measured using the N3 test object over a range of input dose-rates for all field sizes on the two systems. The viewing distance was set in proportion to the display magnification as described above. To compare the contrast-detail results obtained using the conventional and alternative test protocols, the TO 10 test object was used with all field sizes on the two systems at a constant input dose-rate of 0.26 fiGy/s both with fixed viewing distance and with viewing distance proportional to the display magnification.
«
3
In addition, measurements of limiting spatial resolution for each field size were made using a Huttner type 18 test pattern placed at 45° to the direction of the TV raster lines and viewed from close range. Low tube voltage and high tube current were used to produce a high-contrast, low-noise image for this test. Results
Figure 1 shows the change in threshold contrast with input dose-rate for 11.1 mm diameter details. The figures show individual threshold contrast measurements for the three field sizes on each intensifier as symbols, and a curve fitted to the mean of the three
CTT
3 i
>
)
I
>
( )
i
0.1
0.1 dose rate [pGys"1]
1.0 dose rate [pGys"1] (b)
(a) Figure 1. Threshold contrasts measured using the N3 test object over a range of input dose-rates for 30 (x), 23 (O) and 17 (A) cm input field diameters (a) and 36 (x), 25 (O) and 15 (A) cm input field diameters (b) of two image intensifier systems tested using viewing distances proportional to the display magnification.
100 detail diameter [mm] detail diameter [mm] (b) (a) Figure 2. Contrast-detail curves for 30 (x), 23 (O) and 17 (A) cm input field diameters (a) and 36 (x), 25 (O) and 15 (A) cm input field diameters (b) of two image intensifier systems tested using the conventional test approach. The example error bar represents a range of + 1 standard deviation. 688
The British Journal of Radiology, August 1992
An alternative approach to contrast-detail testing
detail diameter [mm]
detail diameter [mm]
(a)
(b)
Figure 3. Contrast-detail curves for the same two systems and field sizes as Fig. 2, i.e. 30 ( x ) , 23 (O) and 17 (A) cm input field diameters (a) and 36 ( x ) , 25 (O) and 15 (A) cm input field diameters (b), but tested using the alternative approach described in this paper. The example error bar represents a range of + 1 standard deviation.
measurements at each dose-rate. The error bars represent a range of +1 standard deviation from the mean assuming a relative standard deviation of 0.15 (Hay et al (1985) quote a relative standard deviation of the order of 0.1 to 0.2). The experimental points lie within this error range except for the highest dose rate in Fig. lb where the image was starting to saturate. Both systems exhibit threshold contrast performance consistent with quantum noise domination up to about 0.7 /iGy/s. Figures 2a and 2b show the contrast-detail curves obtained under the conventional test conditions for the two image intensifiers. Figures 3a and 3b show the results using the alternative test conditions. The results of the limiting spatial resolution measurements are given in Table I. In Fig. 2, the contrast-detail curves for both systems would seem to be consistent with each field size having a different level of performance, with the largest input field being superior at large detail sizes, and the smallest input field being superior at small detail sizes. Such a marked change in low-contrast performance between field sizes would seem to be unlikely since the signal-to-noise ratio for a constant dose-rate should be similar for allfieldsizes on the same image intensifier at Table I. Measured limiting spatial resolutions 30 cm diameter system
36 cm diamater system
Input field diameter (cm)
Limiting resolution (Ip/mm)
Input field diameter (cm)
Limiting resolution (lp/mm)
30 23 17
1.3 1.6 2.0
36 25 15
1.3 1.6 2.2
Vol. 65, No. 776
dose-rates where quantum noise dominates. When the visual bias of the observer is kept constant the resulting contrast-detail curves show an effectively identical performance for all field sizes (Fig. 3). Since the dose-rate and relation between viewing distance and display magnification have been set to be in agreement with those typically operative for the DHSS reference performance survey, the results for any size of image intensifier should be directly comparable to the reference standard using this approach. Discussion
The limiting spatial resolution of an image intensifier/TV system is directly related to the system point spread function (PSF), which is a combination of the individual PSFs of the components of the system. The major contributions to the system PSF are normally made by the TV camera tube and monitor, which therefore limit the spatial resolution of the system (DHSS, 1982). The PSF of the TV system is unchanged for magnified fields, so it might be expected that changing the viewing distance in proportion to the magnification would alter the shape of the contrast-detail curves simply because a closer observer will perceive the system PSF as being wider. Since the shape of the normalized contrast-detail curves are not significantly changed over a wide range of field sizes (Fig. 3), it can be concluded that the shape of the contrast-detail curve is only weakly affected by the resolution of the system. This conclusion is reinforced by the limiting spatial resolution results (Table I), which show that there was a difference in the limiting spatial resolution measured for the various field sizes that was not picked up by the contrast-detail test when the bias of the visual system was maintained constant. Hay et al (1985) have shown the contrastdetail test to be affected by system defocusing for 689
C. J. Kotre, N. W. Marshall and K. Faulkner
limiting spatial resolutions in the range 0.375-1.45 linepairs/mm, but the results of Fig. 3 suggest that, for the higher resolutions typical of modern multifield intensifier/TV systems, the sampling aperture of the observer can be dominant over the PSF of the system under test.
Conclusion
It has been demonstrated that the contrast-detail curves for image intensifiers with widely differing input field diameters can be normalized to be directly comparable to a common reference standard by maintaining constant the visual bias of the observer. This is achieved by using test conditions under which the input dose-rate and the relation between the magnification of the displayed image and the viewing distance are held constant. A comparison of contrast-detail curves measured under the conventional and alternative test conditions shows that features of the curves usually taken as being indicative of differences in the limiting spatial resolution of the system do not appear when changes in the visual bias of the observer are minimized.
Acknowledgments We would like to thank Dr R. M. Harrison for his comments on the manuscript and Professor K. Boddy for his support and encouragement. This work was partially funded by the European Communities Radiation Protection Programme, contract number B17-0014.
690
References COWAN, A. R., HAYWOOD, J. M., WORKMAN, A. & CLARKE,
O. F., 1987. A set of X-ray test objects for image quality control in digital subtraction fluorography. 1: Design considerations. British Journal of Radiology, 60, 1001-1009. DHSS, 1981. Leeds and Kcare Assessment of the GEC Kompact 9 HD Series 2 Image Intensifier System. Scientific and Technical Branch Report STB/12/81 (Department of Health, London). 1982 (2nd edn 1985, 3rd edn 1986). The Testing of Image Intensifier-Television Systems. Working Group Report STB/7/82 (Department of Health, London). HAY, G. A., 1964. A physical assessment of the cinelix electrooptical image intensifier in television fluoroscopy. Radiology, 83, 86-91. HAY, G. A., CLARKE, O. F., COLEMAN, N. J., COWAN, A. R. &
CRAVEN, D. M., 1979. Instructions for the use of the Leeds test objects. (Supplied with each set of test objects). HAY, G. A., CLARKE, O. F., COLEMAN, N. J. & COWAN, A.
R.,
1985. A set of X-ray test objects for quality control in television fluoroscopy. British Journal of Radiology, 58, 335-344. HENSHAW, E. T., 1989. Quality control measurements of X-ray image intensifier television chain systems. In Technical and Physical Parameters for Quality Assurance in Medical Diagnostic Radiology, BIR Report 18 (British Institute of Radiology, London), pp. 120-22. HPA, 1981. Measurement of the Performance Characteristics of Diagnostic X-ray Systems used in Medicine. Part II, X-ray Image Intensifier Television Systems, TGR 32, Part 2 (Hospital Physicists' Association, London). SIEDBAND, M. P., 1981. Fluoroscopic imaging. In The Physical Basis of Medical Imaging, ed. by C M . Coulam, J. J. Erickson, F. D. Rollo and A. Everette James, Jr (PrenticeHall, London), pp. 75-92.
The British Journal of Radiology, August 1992
An alternative approach to contrast-detail testing of X-ray image intensifier systems By C. J. Kotre, MSc, ACGI, N. W. Marshall, BSc, ARCS, and K. Faulkner, PhD, ARCS Regional Medical Physics Department, Newcastle General Hospital, Newcastle-upon-Tyne NE4 6BE, UK (Received 26 June 1991 and in revised form 14 November 1991, accepted 14 January 1992) Keywords: Fluoroscopy, Image quality, Quality assurance
Abstract. The difficulties of making the results of threshold contrast-detail diameter tests on X-ray image intensifier systems consistent with published performance standards are discussed. The current approach to contrast-detail testing is described and an alternative method intended to give greater consistency for all image intensifier input field diameters proposed. The current and alternative test conditions are compared on two image intensifier systems. The results obtained show that the contrast-detail curves for image intensifier systems with a wide range of input field diameters can be effectively normalized to be directly comparable to a common reference standard by applying the proposed alternative test conditions. The implications of this result on the interpretation of the contrast-detail test are discussed.
Threshold contrast-detail diameter test objects (specifically the "Leeds" test objects) have been a valuable tool for the non-invasive assessment of the image quality of X-ray image intensifier/TV systems since their introduction in the early 1960s (Hay, 1964). Although intended to provide the means for performing serial quality assurance measurements on individual systems, the use of these test objects has often been extended to the distinctly separate task of making measurements of image quality over a wide range of systems. This extension of function, although discouraged in the instructions supplied with the Leeds test objects (Hay et al, 1979), has been necessitated by the need to provide assessments of image quality as part of the acceptance/ commissioning tests on new image intensifiers (HPA, 1981; DHSS, 1982; Henshaw, 1989) and to give advice on the replacement of old ones. The current approach adopted to assess image quality in these situations is to perform contrast-detail tests at pre-determined dose rates, X-ray beam quality and viewing conditions, and to make a comparison between the resulting contrast-detail curves and one of a number of standard curves obtained from surveys of well maintained equipment. The Department of Health and Social Security (now Department of Health) document STB/7/82 (DHSS, 1982) shows the results of two such surveys (for intensifiers with 22-25 cm input field diameters), one for new equipment and one for equipment of average age 5 years, both at an input dose-rate of 0.26 /iGy/s. It recommends that the performance level described by the latter curve be used as a reasonable guideline for the performance of new image intensifier systems. This advice was repeated in the subsequent second (1985) and third (1986) editions of this document together with an additional section concerning the application of the Leeds test objects to image intensifiers with larger input field diameters (up to 35 cm) than the 25 cm diameter maximum for which the objects were 686
originally designed. The dose rates recommended for these larger field diameters are, however, manufacturerdependent and no expected performance curve for these larger intensifiers is given. The document also includes the recommendation from a group of equipment manufacturers that, where performance assessments of larger input field diameter intensifiers are to be performed, the magnified field giving the closest to a 23 cm effective input field diameter should be used for comparison with the standard survey result. This approach, however, does not solve the problem of comparing the performances of systems whose maximum field diameters are less than 23 cm. A curve describing the average performance of smaller field-of-view image intensifiers at 0.87 juGy/s has been published in the Leeds test objects instruction manual, and this is often used as a guideline for acceptable performance of these smaller diameter systems, although the DHSS documents do not recommend this. The Hospital Physicists' Association publication TGR32 (HPA, 1981) gives a range of typical results for input dose-rates of 0.26 nGy/s and 0.87 //Gy/s but does not state for what input field diameter the results apply. The present approach to contrast-detail testing
The approach to contrast-detail testing embodied in the design and instructions for use of the Leeds test objects is to test the image intensifier system under viewing conditions similar to those used by the radiologist in the clinical situation and at clinically realistic dose-rates. When most X-ray image intensifiers were either of single-field 15 cm or dual-field 23/12 cm designs, it was possible to deal with most equipment encountered in two categories: input field diameters of 23-25 cm tested at 0.26• [iGy/s, and 12-15 cm tested at 0.87 fiGy/s. The recommended viewing distance for both categories is four times the diameter of the TV monitor display. At this distance the visual bias of the The British Journal of Radiology, August 1992
An alternative approach to contrast-detail testing
contrast decreases with viewing distance as spatial averaging over the sampling aperture of the observer acts to reduce the perceived level of noise, but for small diameter details, the threshold contrast increases with viewing distance owing to the decreasing ability of the eye to resolve them. There is, therefore, no single viewing distance that is optimal for all detail sizes. It is important, however, that a fixed proportionality between the displayed size of the test object and the viewing distance is maintained so that the psychooptical bias inherent in the contrast-detail test is also kept constant. The conventional viewing distance of four times the diameter of the TV display was originally chosen so that a 625-line TV raster is just not perceptible to a person with normal eyesight (Hay et aJ, 1985) but this restriction is probably now less important An alternative approach If the rationale behind the choice of test conditions is because of improvements in TV monitor design and the altered from an attempt to simulate clinical conditions increasing use of high-resolution 1000-line TV systems. to an attempt to obtain a quantitative measurement of Whilst the limit of visibility of TV raster lines with image intensifier/TV performance, an alternative test distance has been investigated (Siedband, 1981) it is not protocol can be formulated. Under this protocol it is clear that their perception or otherwise has a significant proposed that the input dose-rate and the visual para- effect on the detection of low-contrast disks in noise. meters affecting the observer's ability to detect the test The display magnification (i.e. the magnification details in noise are maintained constant for all input factor between the size of the image displayed on the TV field diameters. This offers the advantage that the results monitor and the original size of the object) and TV can be compared directly with the DHSS guideline, but screen diameter reported (DHSS, 1981) for a new GEC suffers from the disadvantage that the dose-rate used for Kompact 9/5 intensifier/TV system, typical of the equipthe test may be different from that being used in clinical ment on which the 1982 DHSS guideline survey was practice. Thus, if the constrast-detail measurement is originally performed, were used to establish a suitable intended to form part of a quality assurance pro- standard viewing distance. The reported display magnigramme, the conventional test protocol should be used fication measured at the centre of the 23 cm field at clinically realistic dose-rates. (1.025:1) was related to the conventionally recomThe X-ray quantum noise present in the image is mended viewing distance of four times the display circle governed by the dose-rate at the intensifier input plane. diameter (23 cm) to establish a standard viewing To make this comparable to the DHSS guideline it distance for an unmagnified display of 90 cm. The proshould be set to 0.26 ^Gy/s. Since the gain of an image cedure for setting the viewing distance is therefore to intensifier is directly proportional to the ratio of its measure the display magnification for the field size input and output phosphor areas, an implication of the under test and multiply this factor by 90 cm. The use of a fixed dose-rate for all intensifies is that the display magnification can be obtained by measuring the automatic gain control of the video system must be distance on the TV monitor screen across a suitable capable of maintaining the image brightness and video number of squares on the Leeds M1 matrix test object modulation at this dose-rate without increasing the and dividing by the actual distance as defined by the contribution of electronic noise to the extent that the object. This method of setting the viewing distance has the additional advantage that any geometrical scaling quantum noise no longer dominates. To minimize variations in the psycho-optical para- caused by the test object not being at the image intensimeters involved in the contrast-detail detection task, the fier input plane will be automatically corrected for. critical variable to be considered is the viewing distance. Other important factors include the ambient room Experimental measurements lighting and TV monitor contrast and brightness adjustThreshold contrast-detail diameter tests with the ment. These should, however, be reasonably constant Leeds test objects N3 and TO 10 were carried out on two for the task of contrast-detail testing where the room new image intensifer/TV systems to compare the lighting is normally dimmed in line with clinical practice conventional test protocol with the alternative protocol and the TV monitor controls are pre-set using the Leeds proposed above. In all cases a measured tube potential grey-scale test object GS1. (The appearance of the GS1 of 70 kVp and an additional filtration of 1 mm of image at 0.26 ^Gy/s can also be used to establish that copper were used. These parameters cannot be altered the dynamic range of the automatic video gain system without invalidating the calibration of the contrasts has not been exceeded.) given by the test object. The systems had input field As the viewing distance is changed, the threshold diameters of 30, 23 and 17 cm and 36, 25 and 15 cm. In contrast for various sized details will also change. In each case visual assessments of the contrast-detail test general, for large diameter details, the threshold objects were made by two experienced observers. observer is expected to be similar to that operating in clinical practice. Cowan et al (1987) suggest that this is an advantage of the conventional test protocol, but while this may be so when the, desired result is a simulation of clinical conditions, it will be shown that where the desired result is a quantitative measurement of performance, variations in visual bias with input field diameter can produce misleading results. The limitations of the present approach emerge when attempts are made to compare the results obtained with the DHSS reference standard. Such comparisons are invalid unless the same test conditions (input field diameter of 23 cm and input dose-rate of 0.26 /^Gy/s) have been used.
Vol. 65, No. 776
687
C. J. Kotre, N. W. Marshall and K. Faultier
To test the assumption that the threshold contrast is quantum noise dominated at the proposed standard dose rate of 0.26 fiGy/s over a range of intensifier sizes, threshold contrast was measured using the N3 test object over a range of input dose-rates for all field sizes on the two systems. The viewing distance was set in proportion to the display magnification as described above. To compare the contrast-detail results obtained using the conventional and alternative test protocols, the TO 10 test object was used with all field sizes on the two systems at a constant input dose-rate of 0.26 fiGy/s both with fixed viewing distance and with viewing distance proportional to the display magnification.
«
3
In addition, measurements of limiting spatial resolution for each field size were made using a Huttner type 18 test pattern placed at 45° to the direction of the TV raster lines and viewed from close range. Low tube voltage and high tube current were used to produce a high-contrast, low-noise image for this test. Results
Figure 1 shows the change in threshold contrast with input dose-rate for 11.1 mm diameter details. The figures show individual threshold contrast measurements for the three field sizes on each intensifier as symbols, and a curve fitted to the mean of the three
CTT
3 i
>
)
I
>
( )
i
0.1
0.1 dose rate [pGys"1]
1.0 dose rate [pGys"1] (b)
(a) Figure 1. Threshold contrasts measured using the N3 test object over a range of input dose-rates for 30 (x), 23 (O) and 17 (A) cm input field diameters (a) and 36 (x), 25 (O) and 15 (A) cm input field diameters (b) of two image intensifier systems tested using viewing distances proportional to the display magnification.
100 detail diameter [mm] detail diameter [mm] (b) (a) Figure 2. Contrast-detail curves for 30 (x), 23 (O) and 17 (A) cm input field diameters (a) and 36 (x), 25 (O) and 15 (A) cm input field diameters (b) of two image intensifier systems tested using the conventional test approach. The example error bar represents a range of + 1 standard deviation. 688
The British Journal of Radiology, August 1992
An alternative approach to contrast-detail testing
detail diameter [mm]
detail diameter [mm]
(a)
(b)
Figure 3. Contrast-detail curves for the same two systems and field sizes as Fig. 2, i.e. 30 ( x ) , 23 (O) and 17 (A) cm input field diameters (a) and 36 ( x ) , 25 (O) and 15 (A) cm input field diameters (b), but tested using the alternative approach described in this paper. The example error bar represents a range of + 1 standard deviation.
measurements at each dose-rate. The error bars represent a range of +1 standard deviation from the mean assuming a relative standard deviation of 0.15 (Hay et al (1985) quote a relative standard deviation of the order of 0.1 to 0.2). The experimental points lie within this error range except for the highest dose rate in Fig. lb where the image was starting to saturate. Both systems exhibit threshold contrast performance consistent with quantum noise domination up to about 0.7 /iGy/s. Figures 2a and 2b show the contrast-detail curves obtained under the conventional test conditions for the two image intensifiers. Figures 3a and 3b show the results using the alternative test conditions. The results of the limiting spatial resolution measurements are given in Table I. In Fig. 2, the contrast-detail curves for both systems would seem to be consistent with each field size having a different level of performance, with the largest input field being superior at large detail sizes, and the smallest input field being superior at small detail sizes. Such a marked change in low-contrast performance between field sizes would seem to be unlikely since the signal-to-noise ratio for a constant dose-rate should be similar for allfieldsizes on the same image intensifier at Table I. Measured limiting spatial resolutions 30 cm diameter system
36 cm diamater system
Input field diameter (cm)
Limiting resolution (Ip/mm)
Input field diameter (cm)
Limiting resolution (lp/mm)
30 23 17
1.3 1.6 2.0
36 25 15
1.3 1.6 2.2
Vol. 65, No. 776
dose-rates where quantum noise dominates. When the visual bias of the observer is kept constant the resulting contrast-detail curves show an effectively identical performance for all field sizes (Fig. 3). Since the dose-rate and relation between viewing distance and display magnification have been set to be in agreement with those typically operative for the DHSS reference performance survey, the results for any size of image intensifier should be directly comparable to the reference standard using this approach. Discussion
The limiting spatial resolution of an image intensifier/TV system is directly related to the system point spread function (PSF), which is a combination of the individual PSFs of the components of the system. The major contributions to the system PSF are normally made by the TV camera tube and monitor, which therefore limit the spatial resolution of the system (DHSS, 1982). The PSF of the TV system is unchanged for magnified fields, so it might be expected that changing the viewing distance in proportion to the magnification would alter the shape of the contrast-detail curves simply because a closer observer will perceive the system PSF as being wider. Since the shape of the normalized contrast-detail curves are not significantly changed over a wide range of field sizes (Fig. 3), it can be concluded that the shape of the contrast-detail curve is only weakly affected by the resolution of the system. This conclusion is reinforced by the limiting spatial resolution results (Table I), which show that there was a difference in the limiting spatial resolution measured for the various field sizes that was not picked up by the contrast-detail test when the bias of the visual system was maintained constant. Hay et al (1985) have shown the contrastdetail test to be affected by system defocusing for 689
C. J. Kotre, N. W. Marshall and K. Faulkner
limiting spatial resolutions in the range 0.375-1.45 linepairs/mm, but the results of Fig. 3 suggest that, for the higher resolutions typical of modern multifield intensifier/TV systems, the sampling aperture of the observer can be dominant over the PSF of the system under test.
Conclusion
It has been demonstrated that the contrast-detail curves for image intensifiers with widely differing input field diameters can be normalized to be directly comparable to a common reference standard by maintaining constant the visual bias of the observer. This is achieved by using test conditions under which the input dose-rate and the relation between the magnification of the displayed image and the viewing distance are held constant. A comparison of contrast-detail curves measured under the conventional and alternative test conditions shows that features of the curves usually taken as being indicative of differences in the limiting spatial resolution of the system do not appear when changes in the visual bias of the observer are minimized.
Acknowledgments We would like to thank Dr R. M. Harrison for his comments on the manuscript and Professor K. Boddy for his support and encouragement. This work was partially funded by the European Communities Radiation Protection Programme, contract number B17-0014.
690
References COWAN, A. R., HAYWOOD, J. M., WORKMAN, A. & CLARKE,
O. F., 1987. A set of X-ray test objects for image quality control in digital subtraction fluorography. 1: Design considerations. British Journal of Radiology, 60, 1001-1009. DHSS, 1981. Leeds and Kcare Assessment of the GEC Kompact 9 HD Series 2 Image Intensifier System. Scientific and Technical Branch Report STB/12/81 (Department of Health, London). 1982 (2nd edn 1985, 3rd edn 1986). The Testing of Image Intensifier-Television Systems. Working Group Report STB/7/82 (Department of Health, London). HAY, G. A., 1964. A physical assessment of the cinelix electrooptical image intensifier in television fluoroscopy. Radiology, 83, 86-91. HAY, G. A., CLARKE, O. F., COLEMAN, N. J., COWAN, A. R. &
CRAVEN, D. M., 1979. Instructions for the use of the Leeds test objects. (Supplied with each set of test objects). HAY, G. A., CLARKE, O. F., COLEMAN, N. J. & COWAN, A.
R.,
1985. A set of X-ray test objects for quality control in television fluoroscopy. British Journal of Radiology, 58, 335-344. HENSHAW, E. T., 1989. Quality control measurements of X-ray image intensifier television chain systems. In Technical and Physical Parameters for Quality Assurance in Medical Diagnostic Radiology, BIR Report 18 (British Institute of Radiology, London), pp. 120-22. HPA, 1981. Measurement of the Performance Characteristics of Diagnostic X-ray Systems used in Medicine. Part II, X-ray Image Intensifier Television Systems, TGR 32, Part 2 (Hospital Physicists' Association, London). SIEDBAND, M. P., 1981. Fluoroscopic imaging. In The Physical Basis of Medical Imaging, ed. by C M . Coulam, J. J. Erickson, F. D. Rollo and A. Everette James, Jr (PrenticeHall, London), pp. 75-92.
The British Journal of Radiology, August 1992
