| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Full Paper |
Department of Medical Physics & Engineering, Leeds Teaching Hospitals NHS Trust, Leeds General Infirmary, Leeds LS1 3EX, UK
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionConclusionsReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionConclusionsReferences |
|---|
The human body varies in composition both along its length and in the transverse plane at any given point along the body. This produces variations in X-ray attenuation due to both the external dimensions of the body and its internal composition. In CT scanning, as the X-ray tube and detectors rotate around the body, the attenuation can change by two orders of magnitude [4]. These differences in attenuation are most significant in the regions of the shoulder and pelvis, where large thicknesses of bone are found in the lateral projections, but a much smaller thickness of bone is present in the anterior–posterior projections. It is these examinations which provide the greatest challenges, in terms of the dose–image quality balance. As a result, using a constant tube current (mA) for each scan angle within a given rotation may result in either photon starvation artefacts on the high attenuation projections or overdosing in the lower attenuation projections.
In the CARE Dose system, during each rotation of the tube and detector assembly around the patient, a small number of the central detector channels provide attenuation information, which is dependent upon the patient cross section and scan angle, to the X-ray generating system [3]. The information provided by these detector channels is used to determine to what extent the mA can be modulated, with respect to an initial tube current setting, without adversely affecting the image quality. As a result the tube current is modulated dynamically with a delay of one rotation relative to the attenuation measurement.
The first patient based assessment by Greess et al [6] showed that, when CARE Dose is used, a dose reduction of approximately 25% (in terms of total mAs reduction) is possible in pelvic scanning "with no significant decrease" in subjective assessments of image quality. Similar percentage dose reductions have been demonstrated in other clinical work [7] and these showed good agreement with phantom based data [2, 5]. Most of the published work has used image noise and/or subjective image assessment to quantify image quality. A small number of papers [8, 9] have used standard deviations from regions of interest (ROIs) to yield a more objective assessment of image noise.
Claims that the image quality was not affected by the CARE Dose system were queried by local users. Having used CARE Dose for a period of time, they perceived that the quality of the images for pelvis scans was subjectively worse when CARE Dose was used and this raised concerns that it may have a detrimental effect on the accuracy of diagnosis. This is despite the manufacturer's recommendation that CARE Dose is used for all clinical situations other than for extremely large patients. This discrepancy between the reported claims and local experience prompted this investigation into the relationships between patient dose, image percentage noise and the signal to noise ratio (SNR) as an indicator of diagnostic detectability. The objective was to clarify whether the CARE Dose system can yield significant dose reduction for no loss of image quality in pelvic scanning.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionConclusionsReferences |
|---|
. The thickness of the phantom is 14.5 cm in the z-axis. The hip phantom does not contain any bony structure and therefore the similarity of the phantom to the pelvic region is geometric only. As such it may be expected that the magnitude of the tube current modulation in clinical practice may be different from that found for this phantom.
|
. Reconstruction kernel B40s was used.
|
Measurements were made which investigated how four different parameters changed with the application of CARE Dose. The four parameters that were investigated were (i) absorbed dose to air, measured in the phantom, (ii) image percentage noise, (iii) CT number for water and polytetrafluoroethylene (PTFE), and (iv) the SNR for both of these materials. These parameters were investigated for each pitch setting, both with and without CARE Dose. Water was chosen to represent low density abscess and PTFE to represent bony structure.
Consistency tests
Prior to the testing, all of the test equipment was placed in the scanning room for at least 4 h in order for the temperature of the phantom and test equipment to stabilize with the room temperature. At the start of each visit the scanner was air-calibrated using the software on the scanner. A short series of consistency tests were performed immediately after the air-calibration which, on all subsequent visits, enabled us to verify that the performance of the scanner had not changed from the previous visit.
On each occasion the hip phantom was positioned 15 cm from the end of the couch on top of the mattress in order to maintain consistent scattering conditions. The phantom was aligned using the laser lights on the scanner and with a spirit level. The set up is shown in Figure 2.
|
The first test was a measurement of the absorbed dose to air. A scan was performed at the standard exposure factors at a pitch of 1.25 with a calibrated 3 cm3 pencil ionization chamber (Capintec Inc., Ramsey, NJ), having an active length of 100 mm, in the central position. The chamber was connected to a Keithley 35050A Dosimeter (Keithley Instruments Inc., Cleveland, OH). The absorbed dose to air was recorded and the mean PMMA CT number and standard deviation (
) were measured adjacent to each of the five possible chamber positions (see Figure 1
), on the CT slice closest to the centre of the phantom, using the region of interest (ROI) tool on the scanner. The size of the ROI that was used was kept constant throughout all of the measurements. The mean CT number of air was also measured at a standard position outside the phantom using a ROI of the same size.
This scan and measurement procedure was then repeated with the ion chamber in the right lateral measurement position. On each occasion the ambient air temperature and pressure were measured, in addition to the phantom temperature, so that an air density correction could be applied to the dose measurements.
The ion chamber was then removed from the phantom and a PTFE rod was inserted into the central measurement position. The scan was repeated and the mean CT number and of the PTFE rod were recorded in addition to the measurements described above. Again this was repeated with the PTFE rod in the right lateral position.
Absorbed dose to air and noise measurements
Absorbed dose to air measurements were made for each of the five chamber positions both with and without CARE Dose at each of the three pitch settings. For each scan the measured dose and total mAs were recorded. On the central slice the mean PMMA CT number and were recorded at each measurement position and the mean air CT number was also recorded. For each measurement position the image percentage noise was calculated using Equation (1) [11]:
|
|
is the standard deviation and CT is the mean CT number (Hounsfield Unit) of the indicated material.
The absorbed dose to air was corrected for ambient temperature and pressure and the ion chamber calibration factor was applied. The volume averaged CT dose index (CTDIvol) was then calculated for the scans with and without CARE Dose, using Equation (2) [11]. This was performed for each pitch setting:
|
|
CTDIvol is actually defined for a cylindrical phantom and as such it is not strictly applicable to the hip phantom that was used in this study. However, the CTDIvol method is an accepted way of accounting for the distribution of dose within a phantom. Since, in this case, it is the comparison between the CTDIvol values for two different scanning situations, rather than the absolute value that was of most importance, the CTDIvol was used simply as an indicator of the relative change in absorbed dose to air. As such, the term "CTDIvol" is used for all calculations that relate to the hip phantom.
The effect of the CARE Dose system on the percentage dose reduction was also evaluated over a range of initial effective mAs settings (50–200 mAs).
Signal measurements
Two sets of signal to noise measurements were made, for the water and PTFE inserts.
For the water measurements thin rubber sheaths were inserted into each of the five holes in the phantom and distilled water was inserted into each of the sheaths and the ends were secured with plastic clips. The sheaths were similar in diameter to the holes in the phantom which made it possible to almost completely fill the holes with water. The set up of the phantom for the water measurements is shown in Figure 3.
|
For each set of measurements the phantom was scanned five times at each pitch setting with and without CARE Dose. For each scan the mean CT number and of the water/PTFE and PMMA were recorded at each measurement position on the central slice in addition to the mean CT number of air at the standard position.
Measurements were also repeated 10 times on one scan of the PTFE rods in order to establish the repeatability of the measurements.
The SNR for the inserts was calculated using Equation (3).
|
|
is the standard deviation. The modulus was used as the mean CT number for water was sometimes below zero.
SNR calculations were performed for each measurement point for each pitch setting. Error propagation was performed for all of the parameters of interest and the calculated values are shown with the results.
The pooled standard deviation of the SNRs was calculated for each pitch setting for the water measurements and this result was used to power the study. The powering process showed that for a result to be statistically significant at the 95% level 25 measurements were required (both with and without CARE Dose).
As a result, a further set of SNR measurements were made for both water and PTFE. The phantom was set up as described earlier and 25 scans were performed both with and without CARE Dose. For each scan the mean CT number and of the insert (PTFE/water) was measured in the central position in addition to the mean CT number and of the PMMA adjacent to the central insert. SNRs were calculated from these measurements and errors were calculated as for the previous measurements.
Statistical analysis was performed on these results (Kruskal–Wallis non parametric test) to determine whether the SNRs of water and PTFE changed significantly for the scans with CARE Dose.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionConclusionsReferences |
|---|
For clarity all the results for the 1.25 pitch setting are shown with summary results for the other pitches.
Dose measurements
For the scans without CARE Dose (i.e. constant mA) the absorbed doses to air were significantly higher in the top and bottom positions than in the lateral positions. For the scans with CARE Dose there was a significant decrease in the absorbed dose to air in each position. The reductions were approximately 42% in the central position, 42% in the top and bottom positions and 8% in the lateral positions. These results were as expected and are shown in Figure 4.
|
The percentage reduction in "CTDIvol" was independent of pitch to within 0.5% over the pitch range of 1–1.25, as shown in Figure 5. The error bars that are shown in Figure 5 represent one standard deviation about the mean.
|
. The reduction in absorbed dose to air is approximately 40% for mAs settings between 50 mAs and 165 mAs. However, this reduction in absorbed dose to air rises to 50% at 200 mAs. It was not possible to obtain results for scans with mAs settings above 200 mAs as this would have exceeded the maximum tube loading at this pitch setting.
|
, across the five positions showed that, in general, there was not a large difference between the measured values with and without CARE Dose, other than for the top and bottom positions, as shown in Figure 7. These discrepancies between the noise values in the top and bottom positions were not found for the other pitch settings and are thought to be anomalous results, relative to the other pitch settings. As expected the noise values in the top and bottom were slightly different from those found in the lateral positions.
|
There was a general reduction in the image percentage noise as the pitch setting was increased (Figure 8), for both CARE Dose on and off, although this was not greater than the experimental uncertainties.
|
along with the calculated p-values.
|
SNR measurements
The measurements for the water and PTFE inserts were used to calculate the SNR for each material (Equation (3)) and these results, for a pitch of 1.25, are shown in Figures 9 and 10. These results from the initial tests show that the SNRs, for both water and PTFE, are lowest in the centre of the phantom both with and without CARE Dose. This is as expected as the noise values were highest in the centre of the phantom. The differences seen between the PTFE SNR values with and without CARE Dose, were generally within the experimental uncertainties. There appears to be a general decrease in the water SNR at each position for the scans with CARE Dose, which is an undesirable trend. However, the differences in water SNR were also within the experimental uncertainties.
|
|
show that there are positional variations in the SNR within the phantom. As such it is not valid to average the SNR for the five different positions as this will mask the positional variations and will result in large uncertainties in the results. For a pitch of 1.25, the water SNR values measured with CARE Dose were lower than those without CARE Dose. This trend was observed for the other pitch values for the water scans but was not observed for the PTFE scans.
For the 25 additional scans the signal and noise values were measured for PTFE and water in the central position in the phantom. From these results the SNR for both inserts were calculated as for the original scans. The differences between the values of CT number and SNR for the scans with CARE Dose on and off were tested for significance using the Kruskal–Wallis test. The mean and standard deviations of the CT numbers and SNRs and the resulting p-values are shown on Table 3. These results show that there were no significant differences, at the 95% level, in CT number or SNR between the scans with and without CARE Dose despite the SNRs generally being decreased when CARE Dose was used.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionConclusionsReferences |
|---|
). The largest reductions, up to 42%, were found in the top, bottom and central positions as these positions lie on the lowest attenuation paths through the phantom and therefore experience the largest tube current modulation and reduction in absorbed dose to air. The dose reductions in the lateral positions, around 8%, are much smaller in magnitude as the attenuation is at its highest in these positions which means the tube current will be at its maximum value. These dose reductions were smaller than those shown by Kalender et al [5] who found a 45% "average" dose reduction via direct dose measurement. Kalender used a scan time of 1 s (compared with a 0.5 s scan time in our work) which allowed for a larger modulation amplitude and therefore a greater dose reduction than in this study. We also found that the reduction in the effective mAs values were lower than those found by Kalender et al, at approximately 25% compared with 40%. However, Kalender's work showed an associated increase in image noise of approximately 10%, which was not found in the first part of this study. Kalender's work was performed with a prototype version of the CARE Dose system which may also explain some of the differences between those initial results and the results of this study. Kalender et al [5] measured an average dose reduction of 45% in the hip phantom with a 3 cm3 ionization chamber similar to that used in this study. This 45% reduction in dose, however, was a straightforward average of the five measurement points rather than a volume average ("CTDIvol") which was calculated here. A straightforward average of our results yields a dose reduction of 30%. The reduction in "CTDIvol" of approximately 32% was in good agreement with the relative dose reduction found by Gies et al [4], who found dose reductions of approximately 38%, for computer simulations using the hip phantom. The large reduction in absorbed dose to air in the central position is of importance as most of the more radiosensitive organs lie centrally. These results imply that the reduction in an individual organ dose (with an associated change in the effective dose) may be larger than the reduction in the values of "CTDIvol" shown here. These results have implications for calculating effective doses in CT as the current Monte Carlo data sets that are used do not reflect the distribution of dose within the patient when a tube current modulation system is used. The large dose reduction in the centre of the phantom also has significant implications for pelvic scans of pregnant patients. If CARE Dose was used for these patients the risk to the fetus may be significantly reduced relative to scans performed with a constant tube current.
Tack et al [10] showed that when using CARE Dose, the percentage dose reduction was independent of the initial effective mAs setting. They used six different mAs settings between 20 mAs and 100 mAs for chest and abdomen CT scans. Our results (Figure 6) show that the percentage dose reduction is approximately constant at a value of around 40% for initial effective mAs values up to 165 mAs. Above this value the percentage dose reduction increases, to approximately 50% at 200 mAs. This occurred as the mAs setting approached the maximum tube current rating for the tube. The Manufacturers recommend that for extremely large patients, where the mAs setting may be close to the tube limit, CARE Dose is not used. No measurements were made to determine whether or not the tube output varied linearly with mAs so we cannot exclude poor output linearity with mAs as a possible cause of the results shown in Figure 6.
The image percentage noise level was not significantly affected by the application of CARE Dose, as shown in Figure 7, for the initial set of noise measurements. The reduction in dose of approximately 8% in the left and right positions occurs as a result of the integration of the reduction in tube current over all scan angles as there is no reduction in the tube current setting in the lateral projections. Given that there has been a general reduction in dose across the phantom there should have been an associated increase in the image percentage noise. No such increase in image percentage noise was found. Combining these results and those for the dose measurements shows that the reductions in absorbed dose to air that were calculated are net dose savings, i.e. they come with no significant noise penalty. Previous work [3, 5–7] showed that dose reductions of 23–45% were possible in the pelvis region with no significant difference in subjective assessments of image quality.
The slight decrease in the image percentage noise with pitch setting, for both CARE Dose on and off is thought to be due to the combined effect of setting a constant effective mAs value and the magnitude of the over-scan which is necessary in helical scanning.
For the additional scans with the water inserts there was an increase in the noise level of approximately 10% for the scans with CARE Dose on relative to the scans with CARE Dose off. This was not found to be significant at the 95% level (p=0.099). This 10% increase in noise agrees well Kalender's work [5]. A similar change in noise was not found for the scans of the PTFE inserts (p=0.727). As the PTFE provides much greater X-ray attenuation than water there is less scope for modulation of the tube current when the PTFE inserts are scanned. As a result the slightly larger reduction in the reported tube current that was found when the water inserts were scanned results in a larger percentage change in noise relative to the scans with CARE Dose off.
Figures 9 and 10 show that there were differences between the SNRs calculated for the scans with CARE Dose on and off. These figures also show that the SNR varied with position within the phantom. The highest values of image percentage noise and the lowest values of SNR were found in the central position which is as expected from photon path length and reconstruction theories. For PTFE the SNRs for the scans with CARE Dose on showed no distinct trend relative to the SNRs for the scans without CARE Dose. This is in contrast to the situation for water where the SNRs for the scans with CARE Dose on were lower than those for the scans with CARE Dose off for 80% of the total number of scans. This shows that there is a trend towards decreased SNR for water when CARE Dose is used.
The larger set of SNR measurements showed a difference in the SNRs of approximately 10% for water whilst there was no difference for the PTFE measurements. This is attributable to the similar percentage change in the noise which was found (Table 2). Statistical analysis showed that there was no statistically significant difference in the SNRs for PTFE and water between the situations with and without CARE Dose (p=0.197 for water, p=0.764 for PTFE). Table 3 shows that, at the 95% level, there was also no significant change in the CT numbers for water and PTFE for the scans with and without CARE Dose. Since the SNR depends on both the signal and noise, neither of which showed a significant change at the 95% level, there was no associated significant change in the calculated SNRs for both water and PTFE. This does not provide an explanation for the users' subjective opinions that the images acquired with CARE Dose, for imaging pelvic abscess, were unsatisfactory.
When the SNR values for water are error corrected (mean value minus uncertainty), the average SNR for the scans with CARE Dose is only just above the detectability threshold of 5 as defined by Rose [12]. Water has an inherently low SNR relative to the PMMA background, but this is further reduced by 10% when CARE Dose is activated. The worse case SNR (i.e. the lowest value of SNR taking into account the calculated uncertainties) was below the threshold value of 5 for 25% of the measurements with CARE Dose off and for 40% of the measurements with CARE Dose on. Although these differences may not be statistically significant they may be detectable by the person viewing the image and are therefore important differences.
The X-ray attenuation path in clinical scanning is non-homogeneous and the human pelvis may have an even more asymmetric attenuation pattern than this phantom. This may introduce a larger modulation in the tube current which would affect the noise and serve to further worsen the SNR situation. This may therefore reduce the confidence with which the viewer of the image can detect tissues which have subtle differences in SNRs. This combination of the decrease in the water SNR and the non-homogeneous attenuation path may therefore explain why subjectively the images that were acquired with CARE Dose had been reported as unsatisfactory for pelvic abscess imaging.
There were large uncertainties in the results of this study. However, the reductions in the SNRs that were found were repeatable over a large number of scans and are therefore considered to be a true representation of the performance of the system. The main explanation for the large uncertainties was that the ROIs that were used for the water and PTFE measurements were small – these were limited by the size of the inserts which were, in turn, limited by the construction of the phantom. If measurements were made too close to the edge of the insert then the mean CT number would have been skewed by the presence of any air around the insert or by the background material itself. It was not possible to make any changes to the phantom design. If it had been possible to use larger inserts (and therefore larger ROIs) it may have been possible to obtain results which were less error dominated. We would recommend that any future studies should consider using larger inserts and ROIs to improve the noise statistics and to ensure homogeneity in the measurements taken within the signal areas. However, it should be noted that at 12 mm in diameter the size of the water inserts were representative of abscesses which are found in the pelvis.
Some differences were found between the results for the left and right lateral positions in the phantom, in terms of absorbed dose to air, noise and SNR for both inserts. Further tests showed that the central alignment laser was inaccurate by approximately 3 mm which resulted in a relative difference between the left and right measurement positions of around 6 mm and that the differences were not due to the performance of the CARE Dose system.
|
Conclusions |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionConclusionsReferences |
|---|
There appears to be a trend towards decreased SNRs for both water and PTFE when CARE Dose was used although no significant differences were found at the 95% level. These changes in SNR were mostly due to changes in the image percentage noise values. The largest decreases in SNR were found for water and were as large as 14%. Since the water inserts were representative of low-density abscess this suggests that the use of CARE Dose may decrease the visibility of low-density structures relative to the background. Therefore using CARE Dose in situations where subtle differences in low CT number tissue pathology are of interest may not be advisable.
|
Acknowledgments |
|---|
Received for publication October 18, 2005. Revision received May 24, 2005. Accepted for publication May 31, 2005.
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionConclusionsReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  T. A. Jaffe, T. T. Yoshizumi, G. I. Toncheva, G. Nguyen, L. M. Hurwitz, and R. C. Nelson Early First-Trimester Fetal Radiation Dose Estimation in 16-MDCT Without and With Automated Tube Current Modulation Am. J. Roentgenol., April 1, 2008; 190(4): 860 - 864. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
