| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Full Paper |
Paediatric Cardiology and Biomedical Engineering, University of Kiel, Schwanenweg 20, 24105 Kiel, Germany
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
In addition, the radiation risk increases progressively with younger age groups. Nevertheless, the available angiographic systems are supplied with X-ray exposure controls configured for adult patients. In this study we try to lower X-ray body dose and improve image quality in paediatric cardiac angiography. For a commercial angiography system, the essential factors that can be controlled are X-ray beam filtering, the tube voltage and the use of an anti-scatter grid.
Guidelines for equipping cardiac catheterization laboratories [10] recommend providing for 0.1 mm Cu filtering and a removable grid. In the literature [11–14], there is an on-going controversy concerning the use of grids for moderately sized paediatric patients as well as the type of filtering [4, 13, 15–17]. Technical advances in the development of high output X-ray tubes with liquid bearing systems allow examinations to be performed at higher currents and enable changes to be made in the X-ray spectrum by using heavy copper filtration [15, 17, 18]. Barkhausen et al [15] and Geijer et al [16] demonstrated that optimization of beam filtering reduces the radiation dose without compromising image quality in adults. Seifert et al [19] also observed significantly decreased radiation dose in fluoroscopic and radiographic procedures due to low-energy filters. In phantom experiments they estimated the change of image quality by analysing the change of absolute brightness values for steps of different X-ray density. However, a digital system allows the user easily to adjust the brightness scale at the display stage. Brown et al [14] reported on phantom experiments in different fluoroscopy modes and adapted the automatic exposure control (AEC) of their fluoroscope, which was initially configured with adult fluoroscopy settings, to the requirements in paediatrics. They described image quality by visually comparing images of contrast mesh objects.
Thicker filters within the X-ray beam of systems using an AEC generally increase the tube voltage. From the literature, it is not clear to what extent the reported dose reduction is linked to the modified pre-filtration and how much to the raised voltage. As a higher voltage leads to reduced image contrast, a technique using copper filters is not always adopted in clinical routine [20].
In order to estimate optimal filtration, tube voltage and grid usage for patients of different sizes, Tapiovaara et al [13] used a Monte Carlo computational model of the fluoroscopic image chain. The image quality was quantified by the observer's signal-to-noise ratio (SNR) of a low-contrast image. They came to the remarkable conclusion that the tube voltage should be adjusted as low as 50 kV in order to get the highest image quality with the lowest radiation dose. Recently Fenner et al [17] confirmed Tapiovaara's results in phantom experiments. They assessed image quality by counting the number of discs visible on standard contrast sensitivity phantoms.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
gives a typical example of biplane recorded frames stored with a resolution of 512 x 512 x 8 bits. The detection of small structures, which is limited by noise, is equally important in areas of high and low radiation absorption. The utilized brightness range is approximately 130.
|
With the Philips system for cardiovascular imaging the following settings can be configured or selected:
Experimental set-up
According to the acceptance certificate of the system, the AEC circuit keeps the detector dose rate constant at 0.12 µGy per frame for the 23 cm XRII field of view used in cinepulse acquisition mode. We measured values between 0.17 and 0.22 µGy. Other selectable fields of view (17 cm and 13 cm) were not used in these measurements. The acquisition frame rate was set to 12.5 frames s–1. The phantom was positioned in the isocentre of the gantry with a distance of 76.5 cm from the X-ray tube. The tube to XRII distance was 100 cm.
Six AEC programs, P1 to P6, were configured using two copper filters (P1, P3, P5: 0.4 mm Cu; and P2, P4, P6: 0.2 mm Cu) and different kV/mA curves (Figure 2). For the Philips system the minimum current, the maximum power and the voltages that determine the linear transition between these two domains define these curves. The pulse times were 6 ms for P1 to P4 and 4 ms for P5 and P6. The nominal focal spot size was 0.8 mm for the programs P2 and P3 and 0.5 mm for the remaining ones. All other parameters of the image chain were kept constant. During the first seven pulses of each run the tube current and peak voltage changed along the selected curve to reach the pre-programmed XRII entrance dose. For each recording the adjusted values of the X-ray peak voltage and tube current were entered into the log. It turned out that with thick phantoms the mA–kV product attained only about 80% of the configured maximum power. This may be due to some pre-programmed power reserve to avoid excessive heat at long runs.
|
depicts the irradiation geometry. Four rectangular PMMA objects (25 x 25 cm wide and 8, 11, 15.5 and 18.5 cm thick) were used and image sequences with and without an anti-scatter grid were recorded. Contrast structures were a piece of lead (black value, BV), a 40 mm broad copper step wedge (level thickness 0.35 mm, steps S0 to S5) and a 10 mm aluminium strip (thickness 2 mm, step S5+) (Figure 4). Each acquisition was taken within the control range of the AEC.
|
|
). The distance between focus and radiation detector was 69 cm. ED is the measured dose divided by the number of pictures (approximately 50) of the respective run.
Image evaluations
The analysis of one digital image (512 x 512 x 8 bit) taken from each run was done without any pre-processing. For a total of 48 pictures (four object sizes, with and without a grid and with six AEC programs, three with 0.2 mm and three with 0.4 mm Cu filtering) the mean grey levels and their variances were measured in eight regions of interest (ROIs) on a digital workstation. They were classified as S0 to S5, S5+ and BV as illustrated in Figure 4. Each ROI was 48 x 48 pixel in size and all ROIs were placed individually for each image.
The lookup table optimized by Philips for image display was not changed. A dark region corresponds to an area of high radiation attenuation. The brightness range (BR) was calculated as the difference between the brightest region S5 and the black value BV in the unprocessed image.
Two differential SNRs (or inherent contrast-to-noise ratios) were defined on the basis of the measured grey levels. They were derived for a dark (SNRd) and a bright (SNRb) region of the radiographs as the ratios of the difference of two adjacent grey levels divided by the square root of the sum of the variances 
2 in those regions:
|
|
The effect of the grid on image quality was analysed quantitatively by using the signal-to-noise improvement factor SIF [21] for the dark and bright region calculated as
|
|
The quality of the image of the thin wires in the radiation detector of the dosemeter was judged subjectively (see Figure 5, bottom right corner of each picture).
|
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
illustrates two images of the 11 cm phantom recorded with 56 kV and 0.2 mm Cu (left image) and 76 kV and 0.4 mm Cu (right image), respectively. The quality of both exposures was acceptable, while ED of the left image was three times as high as with the right one. As presented in Table 1, SNRd increased from left to right, whereas SNRb decreased.
|
presents the rapid decay of ED with rising voltage on a semilog scale. For the same object, ED varied approximately by a factor of 3, depending on the selection of beam filtering and tube voltage. Tables 2 and 3 summarize all the results for the four phantoms measured with and without a grid.
|
|
|
Using Tables 2 and 3, several isovoltage parameter pairs (±1 kV) can be extracted for different filter settings. On average, the ratio of the doses with 0.4 to 0.2 mm Cu filtering was 0.714 (±0.136 standard deviation) with a grid (n=6) and 0.739 (±0.022) without a grid (n=8).
Image quality
BR decreased slowly with the object thickness (Tables 2 and 3). By omitting the grid, BR was reduced by 35% (±10%, n=24) on average. It can easily be re-scaled by grey level windowing. SNRd and SNRb did not change thereby.
SNRb and SNRd showed different dependences on X-ray voltage (Figure 7). SNRb (dashed curve) was optimal for low voltages and decreased rapidly with rising voltages. SNRd became small at low voltages, and the signal difference between the darkest Cu steps (S0, S1) vanished in the high amplitude quantum noise. In our experiments this was the case for the 0.2 mm Cu filter and small objects. SNRd (solid curve in Figure 7) had its maximum at approximately 79 kV.
|
) and without a grid (Table 3). These differences are small and can hardly be seen in Figure 7.
When all the SNRb data for the two filters, 0.2 and 0.4 mm Cu, were averaged it seemed that using the thinner filter leads to a higher SNR in the area of low attenuation. However, there is a simultaneous shift of mean voltage for these two sets, which is responsible for this difference. Conversely, SNRd, measured in the high attenuation (i.e. dark) area of the image, seemed to increase by the transition from the 0.2 to the 0.4 mm Cu filter, when the mean values of all measurements from Tables 2 or 3 were compared. In this case the associated voltage shift seems to favour the data for the thicker filter. It is therefore important to separate the effects of filter thickening and voltage change on image quality.
Effect of an anti-scatter grid
The increase of patient dose caused by the grid can be retrieved from the Tables 2 and 3. The average BF of all measurements with the same object and AEC program was 1.86 (±0.33 standard deviation). BF showed only a weak correlation with object thickness (r=0.745). However, measurements at different voltages were mixed for this analysis. For equal object thickness and filtration, the AEC selected a lower voltage when no grid was used.
In 10 measurements the voltages were equal or nearly equal (±1 kV) with and without a grid. Results are summarized in Table 4. The average BF was 2.58 (±0.74). Dose increase was smallest for the thin object (factor 1.9) and largest for the thickest object (about 3.5).
|
). The average SIFs were SIFd=1.12 and SIFb=1.05 (n=24). Selecting only those 10 measurements where the voltages were equal (Table 4), gave SIFd=1.11 and SIFb=1.27 on average. A dependency on object thickness and filter thickness was not recognized. |
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
The availability of high heat load X-ray tubes allows the spectrum to be changed by additional filtration, reducing the peak voltage and increasing the tube current [15]. Thus, the contribution of low-energy photons which do not penetrate the patient to form the image is reduced. While the positive effect has been demonstrated for fluoroscopy in adults [4, 15, 16, 18, 19] and children [1, 14, 17], it was not clear how the tube voltage should be chosen when large areas of the images are opacified by iodine contrast material as in cardiac angiography.
Conventional X-ray units allow only a limited change of the radiation settings [14]. Reproducible adjustments of X-ray voltage, current and pulse length are difficult to make. However, the parameters used to configure different AEC programs for objects of different sizes can be assessed on the basis of the physical and technical knowledge about generation, interaction and detection of X-rays [22]. The AEC curves (Figure 2) used in this study were configured on the basis of some rough assessments in order to get a broad variety of radiographic performance.
For the final application of a certain radiographic settings, phantom measurements are indispensable. We selected simple structured test phantoms, which showed similar attenuation and contrast values as clinical angiograms of children of all age groups (Figures 1 and 5). Under these "natural" conditions, ED and image quality were compared, while the XRII entrance dose was kept constant at a rather low level.
Image quality versus dose optimization is always task dependent. In contrast to chest radiography, where anatomic noise limits the detection of subtle lung nodules [23], in cardiac angiography the detector dose per frame is so small that quantum noise is limiting the recognizability of small structures.
Radiation dose quantities
In clinical routine, the patient's dose can easily be measured by recording the DAP, from which the entrance surface dose (ESD) for a given geometry can be calculated. In the selected experimental set-up (Figure 3), which closely meets the clinical setting, the ED is measured directly. Although there is a gap of about 9 cm between the phantom and the dosemeter, some backscatter is recorded beside the primary radiation, the more the higher the voltage. On the other hand, the semiconductor dosemeter is less sensitive to higher energies. Overall, we regard ED as a useful estimate of the patient's entrance dose for the purpose of optimization.
DAP and ESD, however, do not give good measures of relative risk of stochastic and deterministic effects. Also, these values cannot be directly compared when measured with different X-ray beam qualities, organs and projections. The effective dose E, which is a weighted sum of dose equivalents received by each organ of interest, is the appropriate radiation dose quantity. However, provided that the detected dose at the image entrance field is constant, a minimal ED value is directly related to a minimal E. Therefore, for optimizing the radiographic parameters, the measured relative ED can be used as an appropriate quantity.
In recent years there has been some effort directed to converting DAP and ESD to E [5, 6, 9, 24–26]. There is still limited information on the sensitivity of the conversion factors E/DAP and E/ESD to variation of tube voltage and beam filtration. According to Wise et al [24] for chest views of adults, a 10% increase in tube voltage leads to 5% increase of E/ESD. Schmidt et al [25] found similar results measured with paediatric phantoms. In addition, they report on an increase of E/DAP of approximately 30% by adding 0.1 mm Cu. In conclusion, for a given ESD or ED, respectively, E increases with rising voltage and filter thickness. For a high ED value (low voltages, moderate filtration) there is a steep exponential decrease of dose within the phantom (or patient) to reach the required XRII dose value. Conversely, for a relatively low measured ED (high voltages, strong filtration), the dose within the patient decreases slowly towards the fixed image dose value. Consequently, the saving of E may be up to 40% smaller than expected from the ED values presented in Figure 6.
Descriptors of image quality changes
In high quality X-ray equipment for cardiac angiography, image noise is mostly due to the limited number of detected X-ray quanta. In this domain spatial resolution can only be enhanced by a better detector efficiency or higher entrance doses at the XRII. Improvements can be demonstrated by visualization of bar-pattern or mesh objects [9, 14, 16]. If the XRII dose is kept constant as in our experiments, image quality must be quantified differently. Contrast resolution can be measured with contrast sensitivity phantoms like the Leeds TOR and N2 phantoms [4, 17] or the CDRAD phantom [27]. However, small changes of the grey level distribution can hardly be analysed by such subjective methods.
The SNR [13, 21, 26, 28], which is proportional to the square root of the XRII dose, allows improved comparison of changes in image quality. Due to the strong dependence between attenuation and the energy of the X-ray quanta, any change of the tube voltage or beam filtration changes the grey level distribution of a structured radiograph. As image noise is signal dependent and different in areas of high and low attenuation, we measured the SNR in regions of low and high brightness, in order to analyse changes of the image characteristics. In adult and paediatric cardiology, the recognition of small brightness differences in darker regions is often as important as in the bright areas (Figure 1). Only for voltages of at least 55 kV the SNR is high enough to enable an observer to detect small objects within the opacified ventricle.
As in our experiments most components of the radiographic system did not vary, only relative comparisons were needed, which reflect the changes of attenuation and scattering reproducibly. For this task the differential SNR was an adequate measure, which was observed in regions of interest of both low and high attenuation. The simple copper step wedge that we used fulfilled all the requirements. For the observer the most striking fact is the change of the BR or contrast of the images acquired under different conditions. BR depends on the thickness of the object, the X-ray voltage and most of all, the use of the grid (Tables 2 and 3). Compared with the findings of Seifert et al [19], BR did not significantly decrease because of using a thicker filter. A change was only observed when tube voltages altered. Contrary to analogue imaging, a small BR is not a genuine problem as one can easily adjust the contrast at the display stage. However, this does not improve the underlying SNR. The parameters SNRd and SNRb did not depend on linear grey level windowing. This demonstrates that these ratios (Equation 1) are robust descriptors of the quality of image generation and detection.
We separated the effects of changes in tube voltage and in filtering on dose reduction and image quality. For equal voltages, the dose saving was about 29% by switching from 0.2 to 0.4 mm Cu. As seen from Figure 7, there was no significant change in the SNRs ratios for these two filters, neither for SNRd, nor for SNRb.
Tapiovaara et al [13] determined the observer's SNR for low contrast objects and concluded that a voltage as low as 50 kV is optimal for imaging with diluted iodine contrast material. But the restriction to low contrast objects does not seem to be sufficient when beam filtering or voltages are altered. Our measurements show that, for such a low voltage, SNRb is indeed maximized (Figure 7, dashed curve). However, at 50 kV the steps S0, S1 and S2 (Figure 4) were all black. In the image presented in the left pane of Figure 5, taken at 56 kV, the border between steps S0 and S1 was already undetectable. No information can be retrieved from darker regions in a cardiac angiogram.
The SNRs could also be measured at the intermediate steps of the phantom. The curves shown in Figure 7 would fade into each other, yielding an almost horizontal curve for the SNR determined with steps S2 and S3.
The use of a thicker copper filter results in significant dose reduction without impairing the image quality. This holds as long the voltage is not increased simultaneously, which is usually the case when an additional filter is put into the beam. Therefore, to keep the image quality at a high level, the AEC has to be adapted to make use of the high output of advanced tubes. The AEC programs should be configured in a way that the X-ray voltage does not fall below 55 kV. On the other hand, approximately 77 kV should not be exceeded to avoid the decrease of SNRb. With adult patients the stronger filtering may necessitate larger focal spots (nominal 0.8 mm or larger instead of 0.5 mm), which leads to a reduced spatial resolution. This observation is an argument against using even thicker filters.
Use of an anti-scatter grid
As the use of a grid for moderately sized paediatric patients is judged differently [10–14], we also investigated its effect on dose and image quality. ED rose by an average factor of 2.6 when the voltage was kept constant. For the smallest phantom, equivalent to a newborn, it increased by a factor of about 1.9. Regarding the image quality improvement as quantified by SIFb and SIFd, the grid should be used for all patient sizes. For equal voltages the SNR improves within the bright (less absorbing) area of the image by approximately 27% and in the dark area by 11% (Table 4). This result is in agreement with visual comparisons of optimal scaled images recorded with and without a grid.
The grid must be inserted or removed before the patient is put on the table. Often there is a large variation of imaging geometry even during the same cardiac examination, especially in the lateral view. Therefore, a prior decision for or against using the grid is needed. In our experiments with the smallest phantom the air-gap between the distal side of the phantom and the XRII was relatively large. As a result, the relative contribution of scattered radiation was smaller than for the thicker phantoms. This may explain the low signal-to-noise improvement of SNRb for larger voltages and small objects (Table 4).
In view of the variable conditions we conclude that in accordance with Tapiovaara et al [13] and Brown et al [14] the rise of ED should be accepted in view of the substantial image improvement. This recommendation stands in contrast to the national and international guidelines [10–12] for paediatric radiography.
Potential limitations
The transfer of our results to clinical settings is limited by the fact that they were obtained with static phantom experiments at a fixed frame rate of 12.5 frames s–1 and a field of view of 23 cm. In addition, we used copper absorbers instead of iodine contrast material as applied in angiography. The different attenuation characteristics of iodine may modulate the curves depicted in Figure 7. However, other structures as ribs, heart muscle, diaphragm and interventional devices furnish further disturbing patterns. The ranges of X-ray attenuation of the radiographs were similar to cardiac angiography (Figures 1 and 5). The applied AECs have also been used routinely during the last year in our cardiac catheterization laboratory, where the AEC programs with a thicker copper filter (P3 and P5) were acceptable. Although our experiments did not prove that the results could be transferred to higher frame rates (25 and 50 frames s–1) and other fields of view, we see no reason why changes should occur.
|
Conclusions |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
SNRd and SNRb were functions of tube voltage and to a large degree independent of object size and beam filtration (Figure 7). The consequence is that the filtration selected should be as high as possible provided that the power of the X-ray tube allows the voltage to be kept in the range of high image quality. The AEC should be adapted when filtration is increased in order to keep the voltage low. Considering only low contrast structures, this means about 50 kV would be optimal. However, when structures in darker areas must also be detectable, as in cardiac angiography, the optimal voltage should be raised to increase their SNR.
In terms of the ALARA principle for cardiac angiography our results allow the following recommendations:
|
Acknowledgments |
|---|
Received for publication July 11, 2003. Revision received October 24, 2003. Accepted for publication November 25, 2003.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  V. Tsapaki, S. Kottou, S. Korniotis, N. Nikolaki, S. Rammos, and S. C. Apostolopoulou RADIATION DOSES IN PAEDIATRIC INTERVENTIONAL CARDIOLOGY PROCEDURES Radiat Prot Dosimetry, January 2, 2009; (2009) ncn313v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D G W Onnasch, F K Schroder, G Fischer, and H-H Kramer Diagnostic reference levels and effective dose in paediatric cardiac catheterization Br. J. Radiol., March 1, 2007; 80(951): 177 - 185. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J Partridge, G McGahan, S Causton, M Bowers, M Mason, M Dalby, and A Mitchell Radiation dose reduction without compromise of image quality in cardiac angiography and intervention with the use of a flat panel detector without an antiscatter grid Heart, April 1, 2006; 92(4): 507 - 510. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
