British Journal of Radiology 74 (2001),932-937 © 2001 The British Institute of Radiology
MAPD—an objective way to select mAs for paediatric brain CT
E T H Wong, BSc1,
S K Yu, PhD2,
M Lai, MSc1,
Y C Wong, FRCR1 and
P C Lau, FRCR1
1 Department of Diagnostic Radiology
2 Medical Physics Division, Tuen Mun Hospital, Hong Kong
 |
Abstract
|
|---|
CT is an advanced imaging modality, but the imaging parameters are normally selected subjectively. For standard head examinations, most of the parameters used are consistent amongst different centres, with the exception of large variations in the selection of the tube current–exposure time product (mAs). As a result, CT images may contain unacceptable levels of noise, or the patient may receive excessive radiation. In this study, the maximum anteroposterior diameter (MAPD) was shown to be a good criterion for mAs selection, and could be measured in a pilot view. 200 paediatric brain CT studies were randomly selected to determine the MAPD at the mid brain level. With knowledge of MAPD distribution, a phantom study was performed to determine the relationship between MAPD and the mAs required for consistent and acceptable image noise. It was found that the required mAs increased linearly with MAPD. Assuming the manufacturer's recommended value is "appropriate" for the average MAPD, the appropriate mAs value could be estimated. Using this method, appropriate mAs values were calculated retrospectively for a group of 240 randomly selected paediatric brain CT studies and compared with the actual mAs subjectively determined by the radiographer. Although their average values were similar, the difference between the calculated and actual values deviated markedly in some cases. When the actual mAs was smaller than the calculated value, higher image noise was observed. However, reduction of image noise was barely observed when the applied mAs was larger than the calculated value. Thus, this method is more objective and appropriate for determination of the mAs value for paediatric brain CT than the traditional subjective method.
|
Introduction
|
|---|
CT offers high diagnostic capability but is alsoa major contributor to collective dose in diagnostic radiology practice [1]. The increasing use of CT in the diagnosis and assessment of cancer and other pathological conditions has made a substantial impact on population exposure from medical X-rays [2].
Patient dose received during CT examinations is determined by a number of factors, including applied voltage (kV), tube current (mA), scan time (s), filtration, section thickness, section spacing and scanning range, etc. Most factors, such as applied voltage, are consistent among different centres. However, the selection of the tube current–exposure time product (mAs) varies significantly even for the same make and model of CT system [3]. This may be due to the fact that CT is a digital technique, where acquisition and display are not interlinked with radiation dose in the same way as conventional radiography. In conventional radiography, overexposed or underexposed film is easy to see. The effect is minimal in a CT image whose brightness is affected mainly by change of window level and width. The most important effect of radiation dose on CT images is an alteration of signal-to-noise ratio.
If all other imaging parameters are fixed, radiation dose increases linearly with increasing mAs. Ideally, mAs should be chosen to produce an image with an acceptable level of noise. That is, an increase in mAs and/or a reduction in noise level does not provide additional diagnostic information. This acceptable noise level should also be relatively constant among all patients to maintain consistent image quality. Unfortunately, selection of mAs often relies on subjective decisions and predictions. For most systems, the minimum and maximum values selected for mAs differ by up to a factor of 3–4 [3]. In other words, images will be unacceptably noisy if the chosen mAs is less than optimal. Alternatively, and perhaps more likely, patients may receive a higher radiation dose because of the observation that a larger mAs always produces a superior image.
Therefore, an objective method of selecting mAs for CT scanning is needed. This is the aim of our study. Since the head sizes of children and infants vary substantially, and children are at greater risk of suffering detrimental effects from exposure to ionizing radiation than adults [4], the authors chose paediatric brain CT scans for this study.
|
Methods and materials
|
|---|
Although head size increases with age, individual variations can be great. Thus age is not a good indicator of head size. It is well known that attenuation of X-rays increases with the size of the object being scanned, thus mAs should be increased with an increase in object size to maintain an appropriate image signal-to-noise ratio. A good index for mAs determination should be directly measurable from a pilot or scout view. The maximum anteroposterior diameter (MAPD) islogically a good choice for this purpose, as shown in Figure 1
. MAPD is defined as the largest anteroposterior distance drawn parallel to the orbital meatal (OM) line.
In this study, two groups of paediatric patients undergoing brain CT (28 infants, <1 year; 172 children, 1–12 years) were randomly selected. At our institute, 12 years of age is considered the upper limit of paediatrics, and patients above 12 years are scanned using an adult protocol.
Accounting for the attenuation from skull bone to the MAPD, the axial scan with the largest anteroposterior diameter was selected. Skull bone thickness was measured at the widest lateral diameter, as shown in Figure 2. The skull bone thickness was then converted into water equivalent thickness based on the following equations:
where HU is the measured CT number of the skull bone, and µw and µb are the linear attenuation coefficients for water and bone, respectively.
View larger version (111K):
[in this window]
[in a new window]
|
Figure 2. Axial image at maximum anteroposterior diameter, where total skull bone thickness xb=distance 1 + distance 2 was measured.
|
|
For the same attenuation, µbxb=µwxw, where xw and xb are the water equivalent thickness and skull bone thickness, respectively. By rearranging the terms:
Thus, the water equivalent MAPD, MAPDw (Table 1), with respect to the measured MAPD, MAPDm, can be calculated as:
To access the image noise level for different MAPD and mAs values, cylindrical water phantoms of different diameters (100 mm, 120 mm, 145 mm and 160 mm) were used in the study. Scans with gantry tilting of 0–30° (at 5° increments) were performed to simulate the actual elliptical shape of the brain and the skull base alignment. A single CT machine (PQ6000; Picker International, Highland Heights, OH) was used in this study for all scanning. Selection of imaging parameters, except for mAs, was based on the manufacturer's recommended values (Table 2). For each size of cylindrical phantom, scans with four mAs values (150 mAs, 225 mAs, 300 mAs and 375 mAs) were performed. A region of interest (ROI) approximately half the size of the phantom was located in the centre of the phantom to measure the image noise by means of the standard deviation (SD) of the mean CT number.
For another two groups of randomly selected paediatric brain CT studies (60 infants and 180 children), mAs values were retrospectively calculated using the new method and compared with the original mAs subjectively determined by radiographers. In each study, the average CT number and the associated SD were measured for four ROIs defined in the axial image with the MAPD. As shown in Figure 3, the ROIs were: (A) basal ganglion; (B) thalamus; (C) white matter; and (D) grey matter. Their sizes were 60 mm2. For each ROI, an image noise level was determined as the ratio of the measured SD to the mean CT number (i.e. noise level=SD/mean CT number).
View larger version (117K):
[in this window]
[in a new window]
|
Figure 3. Axial image showing where CT number and standard deviation were measured for four defined regions of interest: A, basal ganglion; B, thalamus; C,white matter; D grey matter.
|
|
|
Results
|
|---|
In the phantom studies, the image noise (i.e. SD of the mean CT number) was found to increase linearly with the MAPD using linear regression analysis. The results are illustrated in Figures 4a,b and the correlation coefficients are listed in Table 3. It is assumed that the manufacturer's recommended mAs values (i.e. 200 mAs for infant (100 kVp) and 250 mAs for child (120 kVp)) produce images with "acceptable" noise for the average MAPD of infants (137±14 mm) and children (160±16 mm). The "acceptable" image noise level could be interpolated from the figures (infant SD=2.32; child SD=2.15). With knowledge of the "acceptable" noise levels, the relationship between mAs and MAPD was found to be linear, with correlation coefficients of 1.00 for infants and 0.97 for children. These findings were then formulated into a first order polynomial equation. Therefore, the "appropriate" mAs setting can be objectively estimated using Table 4 or the following equations:
View larger version (16K):
[in this window]
[in a new window]
|
Figure 4. Variations of image noise with maximum anteroposterior diameter (MAPD) for (a) infant protocol (100 kVp) and (b) child protocol (120 kVp).
|
|
View this table:
[in this window]
[in a new window]
|
Table 3. Correlation coefficients of the image noise and maximum anteroposterior diameter by linear regression both for infant and child protocols
|
|
View this table:
[in this window]
[in a new window]
|
Table 4. Recommended mAs setting for infants and children based on the measured anteroposterior (AP) diameter (mm)
|
|
In the retrospective comparison study, the distribution of occasions on which the calculated mAs was larger than the applied value, or vice versa, was about 50:50. The distribution of the mAs difference (i.e. calculated mAs-applied mAs) and the percentage difference between the calculatedand applied mAs (i.e. [(calculated mAs-applied mAs)/calculated mAs]x100) for the infant and child studies are shown in Figures 5a,b. The average calculated mAs values (infant 206±69 mAs; child 270±36 mAs) were comparable with the applied values (infant 212±41 mAs; child 274±52 mAs). However, individual calculated mAs values deviated markedly from the original values. For infant studies, the deviation varied from -213 mAs to +123 mAs or from -246% to +38%. Similarly, a deviation from -156 mAs to +169 mAs or from -81% to +46% was observed for child studies. Plots of the variation of image noise level with the percentage difference between calculated mAs and applied mAs are illustrated in Figures 6a,b.
|
Dicussion
|
|---|
Image quality in CT depends on many things, such as the inherent contrast, contrast degradation due to scatter, as well as statistical noise in the image. Noise represents statistical variation of measured density values expressed in Hounsfield units when scanning a homogeneous object such as water. Other things being equal, decreasing noise requires increasing radiation dose [5]. In addition, because noise is regarded as the limiting factor for low contrast resolution [6], anything that increases noise in the image may require an increase in patient radiation dose to maintain the same degree of "detectability".
Relationships between dose at a point, image noise, reconstruction algorithm, section thickness and patient thickness have been derived [7–9]. However, this involves some assumptions and approximations that may not hold true in the practical situation. For example, the noise from the system, other than that caused by photon statistics, has been ignored. In practice, two main factors must be considered in the selection of appropriate mAs values. First, different makes and models of CT scanners with different detector efficiencies and architectures can deliver different radiation dose for a typical CT head examination [10]. Thus, no standard mAs can be applied equally well for different scanners. Therefore, there is a need to calibrate individual CT scanners to establish the relationship between image noise and mAs settings. The calibration methodology would be similar to the first part of our study. Making use of the manufacturer recommended mAs value and the water cylinders, the acceptable noise level could be determined. The acceptable noise level is used as a reference point for setting up the technique used in different protocols.
A second factor is the wide range of mAs that should be used for different patients. Patient thickness is one of the crucial factors of determining image noise for the "appropriate" mAs setting [5, 11]. We found that MAPD increases linearly with image noise for a constant mAs. From the technique chart derived by phantom studies, the required mAs for obtaining images with the same noise level is found to correlate linearly with MAPD. Using these results, appropriate mAs values for different thicknesses could be estimated objectively by a simple measure of the MAPD from a pilot view.
From the results of the retrospective study it was found that the required mAs values varied widely (69–323 mAs for infant and 194–382 mAs for child). Without any objective method of measurement, it is very difficult to set the mAs value confidently and correctly to such an extent. In this study, although the average calculated mAs value was found to be comparable with the applied value for both infant and child studies, individual calculated mAs values could deviate markedly from the original values (infant: deviation from -213 mAs to +123 mAs; child: deviation from -156 mAs to +169 mAs). The negative figures mean that if we use the proposed change in procedures to determine the applied mAs, some of the paediatric patients will receive a radiation dose reduced by as much as 246%. Moreover, as illustrated in Figures 6a,b
, the rate of increase in image noise level is markedly higher in the positive percentage difference region (i.e. applied mAs > calculated value) than the negative region (i.e. applied mAs < calculated value). Clearly this demonstrates that the applied mAs was really too small to maintain steadily low image noise level for those studies belonging to the positive region. In contrast, a decrease of image noise by an increase of applied mAs was inefficient and barely observed when the applied mAs was larger than the calculated value. In other words, the increase in patient dose may not be justified in those cases since the image quality did not improve.
There were more cases where the applied mAs was larger than the calculated value, since larger mAs produced better looking images. However, it was found that the chance of one being larger than the other was equally probable. Thus, there is no overall tendency for any bias in choosing the applied mAs in these studies.
|
Conclusion
|
|---|
The relationship between MAPD and the appropriate mAs value can be easily determined for each CT machine through a phantom study. For paediatric brain CT, the appropriate mAs value can be objectively determined by a simple measure of MAPD in a pilot or scout view.
Received for publication February 7, 2000.
Accepted for publication May 14, 2001.
|
References
|
|---|
- Shrimpton PC, Jones DG, Hillier MC, Wall BF, LeHeron JC, Faulkner K. Survey of CT practice in the UK. Part 2: dosimetric aspects, NRPB-R249. London: HMSO, 1991:1–121.
- Shrimpton PC, Wall BF. The increasing importance of X-ray computed tomography as a source of medical exposure. Radiat Prot Dosim 1995;57:413–5.[Abstract]
- McCrohan JL, Patterson JF, Gagne RM, Goldstein HA. Average radiation doses in a standard head examination for 250 CT systems. Radiology 1987;163:263–8.[Abstract/Free Full Text]
- International Commission on Radiological Protection. 1990 Recommendations of the International Commission on Radiological Protection, Publication 60. Oxford: Pergamon Press, 1991.
- Rothenberg LN, Pentlow KS. AAPM tutorial: radiation dose in CT. Radiographics 1992;12:1225–43.[Abstract]
- Sprawls P. AAPM tutorial: CT image detail and noise. Radiographics 1992;12:1041–6.[Abstract]
- Brooks RA, DiChiro G. Statistical limitations in X-ray reconstructive tomography. Med Phys 1976;3:237–40.[Medline]
- Chesler DA, Riederer SJ, Pelc N. Noise due to photon counting statistics in computed X-ray tomography. J Comput Assist Tomogr 1977;1:64–74.[Medline]
- Pentlow KS, Beattie JW, Laughlin JS. Parameters and design considerations for tomographic transmission scanners. In: Ter-Pogossian MM, Phelps ME, Brownell GL, Cox JR Jr, Davis DO, Evens RG, editors. Reconstructive tomography in diagnostic radiology and nuclear medicine. Baltimore, MD: 1977:267–79.
- Conway BJ, McCrohan JL, Antonsen RG, et al. Average radiation dose in standard CT examinations of the head: results of the 1990 NEXT survey. Radiology 1992;184:135–40.[Abstract/Free Full Text]
- Polacin A, Kalender WA, Marchal G. Evaluation of section sensitivity profiles and image noise in spiral CT. Radiology 1992;185:29–35.[Abstract/Free Full Text]
This article has been cited by other articles:
|
|
|
|
 
J Pages, N Buls, and M Osteaux
CT doses in children: a multicentre study
Br. J. Radiol.,
November 1, 2003;
76(911):
803 - 811.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
, infant protocol; 
, child protocol.
), the thalamus (
), white matter (
) and grey matter (*).