British Journal of Radiology (2003) 76, 51-56
© 2003 British Institute of Radiology
doi: 10.1259/bjr/53215511
A method for the systematic selection of technique factors in paediatric CT
C J Kotre, PhD and
S P Willis, DCR
Regional Medical Physics Department, Newcastle General Hospital, Newcastle-upon-Tyne NE4 6BE, UK
 |
Abstract
|
|---|
A method for the systematic selection of paediatric CT technique factors is described. The approach is based on the assumption that the level of image noise acceptable for a given adult CT image is also acceptable for the equivalent paediatric examination. A simple exponential attenuation model is proposed. Effective linear attenuation coefficients were initially established from a series of phantom measurements simulating head, chest and abdomen examinations at 120 kVp, then extended for a range of tube potentials and beam qualities using a beam spectral model. Application of the method is demonstrated using phantoms representing head, chest and abdomen sections for neonate and ages 1 year, 5 years, 10 years, 15 years and adult.
|
Introduction
|
|---|
Paediatric CT examinations are performed relatively rarely in most centres, and the range of patient sizes encountered is large. It is therefore difficult to make a selection of technique factors that produces a consistently optimized balance between radiation dose and image quality. A number of recent publications have suggested that paediatric CT doses may be unnecessarily high in some cases [1, 2], information that has been reported in the UK national news [3]. Recent UK legislation requires that the dose of ionizing radiation is kept as low as reasonably practicable consistent with the diagnostic purpose, and specifically requires that special attention is paid to medical exposures of children [4]. It is also required that detailed examination protocols are put in place for all diagnostic X-ray work. In this paper a method for selecting technique factors is suggested such that the images produced for paediatric patients will be similar in terms of signal-to-noise ratio to the images produced in the equivalent adult examination, the technique factors for which will be much better established in most centres.
The approach is based on the following assumptions:
- The technique factors for the equivalent adult examination are known.
- The image signal-to-noise ratio in the equivalent adult examination is suitable for diagnosis in the paediatric case.
- The attenuation and relative proportions of body tissues are similar in adults and paediatrics.
- The tube voltage is not altered between the adult and paediatric versions of the examination.
- The reconstruction filter used in the paediatric examination is not radically different from that used in the adult examination (this point is discussed further below).
A useful approximation to patient cross-sectional size is the equivalent diameter, d. For the purposes of this paper, this is measured directly on the subject using a tape measure to determine the perimeter of the section required, then dividing by 
to give the diameter of the equivalent circular section. All equivalent diameters used in this paper are measured in this way and it should be noted that these are not directly interchangeable with those derived by different means, e.g. from patient height and weight, in other work [5–7]. By use of the equivalent diameter, the mAs per slice required to give a similar signal-to-noise ratio relative to that for the adult case in CT can be approximated using a simple monochromatic attenuation model. The relationship is 
where dA is the standard adult equivalent diameter, e.g. that of "Reference Man" [8], dP is the paediatric equivalent diameter, mAsA and SA are the tube current–time product per slice and slice width, respectively, for the adult examination, mAsP and SP are the same quantities for the paediatric examination and µeff is the effective linear attenuation coefficient for the body segment being scanned at the tube potential being used. Equation 1
includes the slice widths SA and SP specifically, as these may change between adult and paediatric examinations given the smaller size of paediatric patients. In addition, manipulation of slice width is used to modify the total number of incident X-ray photons in the experiments reported below. It should be noted, however, that reducing slice width without reducing mAs per slice does not result in a dose saving where contiguous scanning is employed, indeed the patient dose may well increase on single-slice scanners owing to the increased proportion of penumbra present at small slice widths. In all cases the slice widths refer to the reconstructed image slice thickness rather than the collimated thickness of the fan beam.
Equation 1 is based upon the assumption that the problem may be adequately modelled in terms of simple exponential attenuation. Clearly the patient exit spectrum is not monochromatic, but the heavy beam filtration normally used in CT plus the filtering effect of the patient makes the monochromatic assumption more appropriate for CT than for some other radiological imaging techniques.
|
Initial check of model validity
|
|---|
A simple initial experiment was carried out to confirm the validity of the attenuation model proposed above. Uniform Perspex CT dosimetry phantoms of 32 cm and 16 cm were axially scanned on a Siemens Somatom Plus 4 scanner (Siemens AG, Munich, Germany) at 120 kVp. The beam filtration of this scanner was estimated from measurements of half value layer (HVL) using data generated with a beam spectral simulation programme [9]. The filtration was estimated to be 12 mm Aluminium (AL), comparable with the 10 mm Al minimum quoted by the manufacturer. Substituting the phantom diameters into Equation 1 and using a value of 0.22 cm-1 obtained from the spectral simulation programme for the linear attenuation coefficient of Perspex at a depth of 16 cm, gave a ratio between the paediatric (16 cm) mAs–slice width product and the adult (32 cm) mAs–slice width product of 0.03. This large ratio was approximated in practice using 300 mAs, 10 mm slice width for the 32 cm phantom and 50 mAs, 2 mm slice width for the 16 cm phantom. Multiple noise samples were taken from each image to obtain a mean noise level in terms of pixel standard deviation (SD) and an estimate of the standard error on the mean (sem). For the 32 cm diameter phantom, the SD was 13.1 (sem=0.25) and for the 16 cm diameter phantom scanned at 0.03 of the adult mAs–slice width product, the SD was 13.6 (sem=0.16). This level of agreement was felt to be encouraging, given the very large reduction in patient dose per slice implied by this choice of technique factors.
|
Experimental derivation of effective linear attenuation coefficients
|
|---|
Measurements of perimeter were made directly on a set of male and female adult anthropomorphic phantoms (Alderson Research Labs, Stamford, CT) based on the dimensions of "Reference Man" [8] and paediatric phantoms (ATOM Ltd, Riga, Latvia) [10] constructed to the same dimensions as the software phantoms of Christy [11]. These were then converted to the diameter of a circle with the same circumference. These dimensions are taken as typical for neonate and ages 1 year, 5 years, 10 years, 15 years and adult in the experimental work that follows (Table 1).
Head, chest and abdomen examinations were performed on the set of phantoms in order to obtain µeff values for head, chest and abdomen sections. Adult male and female results were combined to produce a result equivalent to an adult hermaphrodite on the assumption that adult CT technique factors are not commonly altered for gender. Axial scans of the phantoms were carried out on the Siemens Somatom Plus 4 scanner at 120 kVp, 12 mm Al total filtration, using nominal 10 mm slice widths. Effective attenuation coefficients for other tube potentials and scanner filtrations are derived below.
To obtain values of µeff for each body segment, the level of quantum noise within a region of uniform tissue-equivalent material was measured for each of the seven phantoms using fixed technique factors. For the chest, noise measurements were confined to the near water-equivalent material of the phantom heart region. Five circular areas of interest were analysed on each image and the geometric mean value of the pixel SD calculated. Gross non-uniformities in the phantoms were avoided and efforts were made to use the same location for areas of interest when comparing noise measurements from the same section.
To measure the level of quantum noise within the phantoms accurately, any additional noise components due to inhomogeneity of the phantom material at small scales and system noise had to be eliminated. Initial measurements showed that the material of the commercial paediatric phantoms was indeed inhomogeneous at small scales, giving rise to an additional noise component. Although the adult phantoms were more uniform, they also produced a small additional noise component. All noise measurements were therefore made at two widely separated values of mAs so that the quantum noise and phantom inhomogeneity components could be separated using the relationship
where 
S is the pixel SD representing the non-quantum noise sources (including both phantom and system noise), 1 is the pixel SD measured at the high value of mAs, 2 is the pixel SD measured at the low value of mAs and A is the ratio between the high and low mAs values used. The value of this ratio in the experiments was 10.4. The pure quantum noise contribution for each measurement was then calculated by quadrature subtraction of the phantom and system noise, s established for that phantom from Equation 2, from the total noise SD. As the mean Hounsfield unit (HU) value for the paediatric phantoms was found to be significantly above the expected value for water equivalent tissue (in the range 70–90 HU), an additional correction for this anomalously high attenuation was necessary. The relative numbers of photons contributing to the image were corrected to what they would have been for zero HU and the quantum noise estimations were adjusted accordingly.
The reciprocal of the fractional change in quantum noise with equivalent diameter for each body segment was then squared to obtain the fractional change in number of X-ray photons forming the image. Finally, the values of µeff were obtained by plotting the natural logarithm of the relative number of photons forming the image against the equivalent diameter of the phantom section. The slopes of the fits to these experimental points gave the required effective attenuation coefficients.
|
Extension of µeff values using an X-ray spectral simulation programme
|
|---|
A beam spectrum simulation programme [9] was employed to extend the experimentally derived µeff values over a more useful range of tube potentials and beam filtrations. The programme was used to model the incident X-ray beam and the attenuation of the various body sections in terms of thicknesses of water and bone for the head, water for the abdomen and water and air in the case of the chest. The parameters of the model were adjusted to match the measured beam HVL and experimentally derived effective attenuation coefficient for the head, chest and abdomen sections at 120 kVp (constant potential, 10° target angle) as measured above. The kVp and total tube filtration were then varied to produce the required range of µeff values, which are given in Tables 2–4 for the head, abdomen and chest examinations, respectively. In each table the experimentally derived value is shown in bold type and the values extrapolated from it in normal type.
|
Experimental validation
|
|---|
A further series of scans of the anthropomorphic phantoms was made to verify that application of the method of paediatric factor selection proposed above results in images with a similar signal-to-noise ratio to those produced by the equivalent adult examination over a realistic range of tube potentials and beam filtrations. Two scanners were used, a Siemens Somatom Plus 4 (estimated total filtration 12 mm Al) and a Toshiba Asteion (Toshiba Corporation, Tochigi, Japan) (estimated total filtration 4 mm Al). Head, chest and abdomen sections for each of the available ages were scanned at 80 kVp, 120 kVp and 140 kVp (Somatom Plus 4) and 120 kVp (Asteion), using a 10 mm nominal slice width. Adult mA-slice width product values were based on those found in a regional CT dosimetry survey, and the values for the paediatric scans were obtained from Equation 1 using the equivalent diameters in Table 1 and the effective linear attenuation coefficients from Tables 2–4. A realistic field of view to suit the size of section being scanned was chosen in each case. As the paediatric phantoms are hermaphrodite, averaged male and female adult measurements were used to produce an equivalent hermaphrodite adult.
Where the mAs value required for the smaller phantom sections fell below the minimum mAs available for a full rotation of the scanner (37.5 mAs for the Somatom Plus 4, 30 mAs for the Asteion), the number of photons per slice was further decreased by reducing the slice width. The relationship between slice width and relative number of photons detected was previously established for each scanner in a separate series of water calibration phantom scans. Where slice width reduction was employed, a further correction was required to account for the increase in measured phantom noise component as the effect of noise averaging across the slice width is reduced. Noise measurements were again corrected to remove the effects of phantom local inhomogeneity and the elevated attenuation of the paediatric phantom material.
|
Results
|
|---|
Figure 1 shows an example of the trend in mean noise levels measured in the experimental validation for the abdominal examination. This example is for 120 kVp at 12 mm Al total tube filtration. Error bars of ±2 standard errors on the mean, calculated for each point, are shown together with the best fit line of zero gradient. The noise results are plotted against nominal phantom age, with the hermaphrodite adult point being placed at age 30 years. This is an arbitrary choice but is consistent with the age at which the relationship between body weight and age reaches a plateau [8]. Complete results for the abdomen, head and chest validations are given in Table 5.
View larger version (13K):
[in this window]
[in a new window]
|
Figure 1. Experimentally measured mean noise level plotted against phantom age for abdomen examinations at 120 kVp, 12 mm Aluminium. Error bars of ±2 standard errors on the mean calculated for each point are shown together with the best fit line of zero gradient.
|
|
View this table:
[in this window]
[in a new window]
|
Table 5. Summary of image noise results, (mean pixel standard deviation) for the experimental validation on phantoms for head, abdomen and chest sections, (standard error on the mean in parentheses)
|
|
Despite a careful analysis of propagation of errors through the experimental process, including the application of Equation 2 and the subtraction of the structure noise from the total measured noise, it was found that the zero gradient line did not pass through the error bars of ±2 standard errors on the mean for a number of the experimental points. This would seem to indicate that some sources of error have not been adequately accounted for. Possible candidates for the "missing" error include variations in the choice of noise sampling region between the scans taken to establish the phantom structure noise and the subsequent validation scans and variations in noise between individual scans. Despite this, no systematic trend in noise level with phantom age is seen across the results taken as a whole, indicating broad agreement with the expected flat noise levels.
|
Application of the method in practice
|
|---|
It is envisaged that the relationship of Equation 1 would be applied as follows. Measure the perimeter of a representative part of the section to be scanned using a tape measure on each individual patient and divide this by to obtain the equivalent diameter. Substitute the paediatric equivalent diameter into Equation 1 along with the appropriate value of µeff from Tables 2–4 and a value for dA, taken from Table 1 for the appropriate body segment. The results of this calculation could be tabulated or shown graphically as mAs against perimeter for standard examinations.
|
Discussion
|
|---|
The magnitude of the dose saving expected from the use of Equation 1 is illustrated in Figure 2, which shows the ratio of paediatric to adult mAs–slice width product for abdominal examinations as a function of phantom age, based on the average equivalent diameters given in Table 1. Curves are shown for 80 kVp, 100 kVp, 120 kVp and 140 kVp, and are based on the values of µeff from Table 2 corresponding to a total filtration of 9 mm Al. The curves shown are spline fits to calculated values at age 0 years, 1 year, 5 years, 10 years and 15 years and 100% for the nominal adult 30-year-old. At first sight the curves are surprising both for the very small proportions of adult mAs–slice width product required for younger children and for the very rapid increase from the teens to adulthood. The curve shapes do, however, seem more reasonable when the rates of change in the equivalent diameters from Table 1 are considered together with the exaggeration of those changes resulting from the application of the exponential relationship given in Equation 1. The noise results obtained from phantom body sections of realistic dimensions illustrate that the choice of mAs–slice width product value based on these relationships does produce images with similar noise content to the adult case.
View larger version (19K):
[in this window]
[in a new window]
|
Figure 2. Fraction of adult mAs–slice width product per slice required for a paediatric abdomen CT image with the same noise level plotted against phantom age, using the standard dimensions from Table 2. The curves are spline fits to points for 80 kV, 100 kV, 120 kV and 140 kV at ages neonate, 1 year, 5 years, 10 years and 15 years, and adult placed at age 30 years.
|
|
Figure 2 serves to demonstrate the relationship between mAs–slice width product and age resulting from the use of Equation 1 for the standard phantoms used. It is not recommended that this kind of relationship be used as the basis of a clinical scheme to select CT technique factors dependent on age alone, as the variation in patient size for any given age could result in non-diagnostic images for larger children.
Additional factors that could affect the noise content in the paediatric image include choice of a different image reconstruction filter, use of zoom reconstruction and use of different choices of beam hardening correction. The variations offered will change depending on the model of CT scanner used, but experimental tests on choice of adult and paediatric filters (with the same "sharpness"), zoom reconstruction and a range of fields of view showed no significant difference in image noise content. Some scanners change beam filtration for small fields of view. Since the rates of change of µeff with total filtration are small (Table 2), differences in filtration between the adult and paediatric cases within the range 3–12 mm Al will produce a maximum error on the estimated mAs–slice width product of 12%.
All experimental work described employed single slice axial scanning. The method derived can, however, be applied directly to spiral scanning protocols, as the quantum noise level in these images will still be governed by the mAs–slice width product. Changes to the pitch of the spiral run will alter the average patient dose per slice, but not the number of photons collected per tube rotation. For multislice spiral scanners the method is also applicable provided the same combination of multiple slices and pitch is maintained between the adult and paediatric cases.
The results obtained indicate that significant dose savings could be made in paediatric CT without loss of diagnostic quality by using the approach described. These findings are consistent with those of Rogella et al [12] who have reported no loss of diagnostic information for paediatric chest CT performed at 25 mAs per slice, and Huda et al [13] who found that the dose per slice for head scans of newborns could be reduced to 40% of the adult value.
Difficulty was experienced in obtaining the low mAs values required for the experimental validation on one of the CT scanners used, and changes in slice width had to be employed to simulate the effect of applying lower doses to the phantoms. For reduction in paediatric technique factors to the low levels proposed, an extended lower range of tube current settings would be of value.
|
Conclusion
|
|---|
A method for the systematic selection of paediatric CT technique factors has been described using a simple attenuation model based on measured patient equivalent diameter. Images produced for paediatric patients scanned using technique factors selected in this manner will be similar, in terms of signal-to-noise ratio, to the images produced for the equivalent adult examination, the technique factors for which will be much better established in most centres. Using this method, a significant reduction from present patient dose levels for paediatric CT would be expected.
Received for publication December 4, 2000.
Revision received August 6, 2002.
Accepted for publication August 12, 2002.
|
References
|
|---|
- Donnelly LF, Emery KH, Brody AS, Laor T, Gylys-Morin VM, Anton CG, et al. Minimising radiation dose for paediatric body applications of helical CT. AJR 2001;176:303–6.[Free Full Text]
- Paterson A, Frush DP, Donnelly LF. Helical CT of the body: are settings adjusted for paediatric patients? AJR 2001;176:297–301.[Abstract/Free Full Text]
- BBC News Online. Worry over childrens CT scans. BBC, 2001. http://news.bbc.co.uk/hi/english/health/newsid_1126000/1126776.stm
- The Ionising Radiations (Medical Exposure) Regulations 2000. London: HMSO, 2000.
- Lindskoug BA. The Reference Man in diagnostic radiology dosimetry. Br J Radiol 1992;65:431–7.[Abstract]
- Hart D, Wall BF, Shrimpton PC, Dance DR. The establishment of reference doses in paediatric radiology as a function of patient size. Radiat Prot Dosim 2000;90:235–8.[Abstract]
- Chapple C-L, Broadhead D, Faulkner K. A phantom based method for deriving typical patient doses from measurements of dose–area product on populations of patients. Br J Radiol 1995;68:1083–6.[Abstract]
- ICRP. Report of the task group on Reference Man. Publication 23. Oxford, UK: Pergamon Press, 1975.
- Institute of Physics and Engineering in Medicine. Catalogue of diagnostic x-ray spectra and other data. Institute of Physics and Engineering in Medicine Report 78. York, UK: IPEM, 1997.
- Varchenya V, Gubatova D, Sidorin V, Kalnitsky S. Children's heterogeneous phantoms and their application in roentgenology. Radiat Prot Dosim 1993;49:353–6.[Abstract]
- Christy M. Mathematical phantoms representing children of various ages for use in estimates of internal dose. NUREG/CR-1159, ORNL/NUREG/TM-367. Oak Ridge, TN: Oak Ridge National Laboratory, 1980.
- Rogella P, Stover B, Scheer I, Juran R, Gaedicke G, Hamm B. Low-dose spiral CT: applicability to paediatric chest imaging. Pediatr Radiol 1999;29:565–9.[CrossRef][Medline]
- Huda W, Chamberlin CC, Rosenbaum AE, Garrisi W. Radiation doses to infants and adults undergoing head CT examinations. Med Phys 2001;28:393–9.[CrossRef][Medline]
This article has been cited by other articles:
|
|
|
|
 
C. J. Martin
RADIATION DOSIMETRY FOR DIAGNOSTIC MEDICAL EXPOSURES
Radiat Prot Dosimetry,
March 28, 2008;
(2008)
ncm495v1.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
