Setting up computed tomography automatic tube current modulation systems

The dose performance of CT scanners is tested through measurement of the CT dose index (CTDI) with a 100 mm ionization chamber in standard cylindrical phantoms made from polymethyl methacrylate (PMMA) 320 mm and 160 mm in diameter representing the body and head respectively (IEC 2003). Measurements are made with a CTDI chamber at the centre (CTDI_c) and four peripheral (CTDI_p) positions, for which results are averaged, in a phantom, and the weighted CTDI (CTDI_w) derived using the equation:

$$\begin{eqnarray}&&\text{CTD}{{\text{I}}_{\text{w}}}=1/3\text{CTD}{{\text{I}}_{\text{c}}}+2/3\text{CTD}{{\text{I}}_{\text{p}}}\end{eqnarray}$$

The volume averaged CTDI (CTDI_vol), is derived by dividing the CTDI_w by the pitch of the tube rotation (table movement/ beam width). It is displayed on CT scanner consoles and used in evaluation of patient doses. For dosimetry purposes, the CTDI_vol values for the two phantoms are very different, so in any study of ATCM performance it is important to ensure that it is clear which CTDI_vol is being used. The CTDI_vol was designed to provide a measure related to the average dose within a CT slice, but the actual dose will be dependent on the patient or phantom dimensions. In order to take account of changes with patient size, an adjustment can be made to the CTDI_vol to give a size specific dose estimate that approximates to the average tissue doses for patients of different size (AAPM 2014). In this study although phantoms with elliptical cross sections of different sizes are used, it will be the CTDI_vol for the standard 320 mm PMMA phantom that is displayed on the CT console which will be employed for the analyses of ATCM performance.

Image quality is assessed using other cylindrical phantoms containing contrast detail tests. It is difficult to incorporate ATCM system performance into these tests, since the phantoms have fixed dimensions. An option that provides a good test of ATCM is to scan an anatomical phantom representing a standard adult, but a set of phantoms of different sizes would be required to fully evaluate ATCM performance and data from phantoms of this type cannot be quantified readily. A geometrical phantom with sections having a range of dimensions can be used to assess the modulation for patients of different sizes, but performance quantities that might be measured will vary along the phantom. Dose levels at different positions inside a phantom will be determined by the incident air kerma and the phantom dimensions which affect the attenuation and scattering, and will vary with position both within the phantom and in the scan field. Therefore the measure of mA provided by the CT scanner is an easier quantity through which to assess the relative variation in exposure. The relationship between dose in terms of the CTDI_vol measured in a phantom and the mA will vary with the CT scanner, as it depends on the attenuation and shape of the bow-tie filter, but the mA provides a useful indicator of relative changes in dose for individual CT scanners. There are ranges of attenuations along the lengths of anatomical phantoms and these data can be used to provide an indication of the variation in mA with attenuation. Quantification of attenuation data along the lengths of such phantoms might also allow changes in mA with attenuation to be quantified. Image quality can most readily be assessed through changes in noise along the phantom. Data acquired as part of the study described in Sookpeng et al (2013) are included in the first two figures, and additional analyses of data from this study are included in later figures.

A phantom for testing ATCM systems needs to be elliptical to take account of changes in current as the x-ray tube rotates around the body. An axis ratio of 3:2 provides a realistic cross section for much of the trunk, but a wider section (e.g. ratio 2:1) would be required to mimic the shoulder region. ATCM phantoms should cover a substantial portion of the useful range of patient attenuation encountered routinely in clinical practice, with a steady progression from the smallest to the largest diameter. The majority of phantoms go from diameters less than that of the neck up to diameters larger than an average patient, typically over 400 mm in order to allow trends in dose level for larger patients to be assessed. However, the implications of size on manual handling must be considered if a phantom is to be transported between different hospitals for routine testing. There must also be a facility to suspend the phantom horizontally in the x-ray beam, without the interference of support material.

A number of phantoms have been based on cones with diameters varying continuously along the length. An elliptical conical phantom made from PMMA with axis ratio 3:2 and a largest diameter of 424 mm has been used by ImPACT (Keat et al 2005, Sookpeng et al 2013) and others have used related designs (Gutierrez et al 2007, Muramatsu et al 2007, Field 2010, McKenney et al 2014). The differences in performance between different modes of scanner operation can be seen clearly in results from scans of this design of phantom for Philips and GE CT scanners (figure 1). The mA increases with phantom diameter for both scanners over the range of cross-sectional areas from about 250 cm² to 700 cm², but starts to level off towards the upper end, with that for the GE scanner reaching the maximum mA value (figure 1(a)). The noise level in the GE scanner remains fairly constant over most of the phantom, in line with the use of the noise reference comparator, but begins to increase at the point where the mA reaches the maximum value (figure 1(b)). Whereas the noise in images from the Philips scanner, for which the comparator is a reference image, starts at a lower level in the narrower end of the phantom, but rises steadily as the diameter increases (figures 1(c) and (d)). The mA does not reach a maximum at a particular value, and the relative increases in current with size are similar for the different mAs settings. The mA for the GE scanner rises more steeply under ATCM control as the phantom diameter increases than that for the Philips one in order to maintain a constant noise level and this translates into a proportionately more rapid increase in patient dose with body size. Thus conical phantoms provide a good indication of performance in terms of the variation in mA modulation and image noise with phantom diameter. They do not provide positions where factors are constant or ones that can readily be identified for measurement of dose.

Zoom In Zoom Out Reset image size

An alternative design is a stepped phantom made up of a series of polyethylene or PMMA sections of different diameter (Sookpeng et al 2013, McKenney et al 2014). However, the sharp steps in attenuation between sections do not mimic the smooth variations encountered in the body, and there may be small air gaps between sections that can give large discrepancies in x-ray penetration. These sharp variations in attenuation affect the way in which the ATCM operates on some CT scanners, particularly those using image quality references based on noise level, as the algorithms are designed to be more responsive to small changes in attenuation, and over-compensate for step changes that do not normally occur in the human body. Figure 2 shows an example of the sharp changes in mA that may occur every quadrant with GE and Toshiba scanners. Large increases and decreases in mA occur on the next rotation following respective increases and decreases in phantom attenuation as the phantom moves through the CT gantry. Thus although this design of phantom is easier to construct it is not suitable for testing all CT scanners in use at the present time and this is discussed in more depth by Sookpeng et al (2013). The problems relating to the scanning of stepped phantoms are being addressed by manufacturers in more recent scanners. For example whereas the Toshiba SureExposure 3D version 6 based the changes in mA on the actual SPR of the phantom and gave sharp variations in mA when step changes were encountered, that in version 7 uses an internal SPR with substantially improved mA calculation at such step changes (figure 3). Phantoms with limited numbers of elliptical sections can potentially be used for dosimetry and noise measurement, but in order for them to perform satisfactorily with some CT scanners currently in service, sharp discontinuities in attenuation may need to be limited to less than 6% and air gaps between sections excluded to minimize the effects described above (Sookpeng et al 2013). The flexibility of such phantoms may be improved by incorporating annular wedges between adjacent sections to eliminate sharp discontinuities. In order to accurately characterise dose performance within sections of fixed diameter, lengths of individual sections should ideally be three times the beam width used for testing. Another option is to use a series of shorter elliptical sections with small changes in diameter between sections to create a conical phantom. This approach has been adopted in the production of the Leeds CT AEC phantom, which comprises eleven PMMA ellipses, each 25 mm thick, for which the steps in attenuation between adjacent sections are smaller (www.leedstestobjects.com/index.php/phantom/ct-aec-phantom/) (Wood et al 2015).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The phantoms described so far have been made from PMMA or polyethylene, but other materials such as nylon, which has an effective atomic number and attenuation coefficient similar to water, could be used, and properties are compared in table 1. When scanning any of these phantoms it is recommended that the SPR used to plan the scan is set within the boundaries of the phantom, as the large changes in attenuation at the start and end of the phantom can lead to unrealistic tube currents that affect the performance of some ATCM systems. In addition, it is better to carry out the tests using smaller beam widths, so that the response is averaged over a smaller range of diameters in conical phantoms and the response in the centre of a section can be quantified more reliably for sectional ones. Some phantoms also have facilities for measurement of dose in the centre with an ionization chamber either a 100 mm CTDI pencil chamber or a smaller one to give the dose at a central position. If a dedicated ATCM phantom is not available, some information on ATCM performance could be obtained using a nested PMMA CTDI phantom, that is one in which the 160 mm diameter standard head phantom fits within a larger annulus to form the 320 mm body one (Tsalafoutas et al 2013). Offsetting the head and body sections will allow three regions of differing attenuation to be created, although such a phantom will have a circular rather than an elliptical cross-section and contains sharp discontinuities.

Having a phantom with a range of diameters allows the performance of a scanner ATCM system to be assessed qualitatively, but gives little information about dose performance in clinical practice. The dose performance of CT scanners is described in terms of two measurable quantities, the CTDI_vol (IEC 2003), which relates to the dose per unit length of a patient taking account of the pitch of the x-ray tube rotation, and the dose length product (DLP), which gives a measure related to the total dose received by a patient from an examination. The dose performance of CT scanners in clinical practice is assessed through audit of dose quantities for patient examinations (IPSM 1992, Martin 2011). It is now possible to download large amounts of DLP and CTDI_vol data from Radiology Information Systems on all patients scanned within particular periods (Wood et al 2012, Martin 2016). In the past when data were only available for a limited number of patients, it was necessary to restrict the range of patient weights for each examination studied, but with larger numbers of results this is unnecessary. The median value of the distribution of dose quantities for all patients is recommended for these assessments of larger numbers of patients (Martin 2016). Comparison of median patient doses for examinations performed on CT scanners with diagnostic reference levels (DRLs) enables their performance to be assessed. If the median dose exceeds or is close to the DRL, then the scanner protocol should be reviewed and factors optimized. Valuable information on optimizing scanning protocols can be obtained by comparing results for different scanners. But judgements on the optimal dose levels must take account of the image quality and users must be assured that it is adequate for the types of examination to be performed. Figure 4 shows patient doses for chest abdomen pelvis CT examinations listed in order of increasing DLP for five Toshiba Aquilion CT scanners in which the target noise values were selected by the operators. The DLPs rise more steeply towards the end of the distribution for scanners A, B and C linked to higher dose levels for larger patients, whereas the rise is lower for scanners D and E. The median DLPs for all scanners are within 40%, but those with ATCMs using the same reference noise level for all patients gave higher doses to heavier patients. Differences between CT scanners in patient dose distributions can be identified by comparing the 3rd quartile values of the DLPs as well as the medians (figure 4) (Martin 2016). If the image quality given by scanners using a lower dose protocol is judged to be acceptable, then this can be used as a guide in setting up protocols on other scanners. The distribution of doses for groups of patients on different scanners will depend on the relative weight distributions. However, as a rough guide to whether protocols are optimized for larger patients, the user could take scanners where the third quartile is more than about 50% higher than the median of the dose distribution as the starting point for carrying out an investigation. Changes made to protocols may involve setting a higher noise reference for larger patients. In current models of CT scanner, such large increases in dose for Toshiba and GE scanners should be avoided by the automatic setting of higher reference noise levels for larger patients, but in many models in use currently, this will not be the case.

Zoom In Zoom Out Reset image size

If the dose performances of different CT scanners for a standard sized patient are adjusted so that they are similar, then the mA modulations produced by their ATCM systems along a conical or sectional phantom can be compared. The relative values for the mA for different parts of the phantom can be obtained from mA versus phantom position curves, such as those shown in figures 1(a) and (c). Results of this type are plotted as a function of the larger axis diameter for the elliptical conical phantom for CT scanners from the four manufacturers in figure 5. Tube current data have been normalized with respect to the requirement for a PMMA phantom of diameter 300 mm, which equates to an attenuation similar to that for 320 mm of water. A curve equating to the mA required to maintain the same level of transmitted air kerma is also included for comparison. The forms of the normalized curve for different image quality reference values on each individual scanner were generally similar. The scanners that apply a reference image or mAs (Philips and Siemens) use a higher mA for smaller patients, whereas those based on a noise reference (GE and Toshiba) increase the mA more for larger patients. The form of the latter group is similar to the curve representing the reciprocal of the transmitted air kerma (figure 5). There is some levelling off in mA at larger diameters for these scanners, when the mA reaches the maximum value. Two curves are included for the Toshiba scanner showing that with the 'low dose' option that starts at a lower mA, a wider mA range can be achieved, whereas with the 'standard' range the modulation is curtailed when the mA reaches its maximum value. In practice the actual mA selected for the standard option would be substantially greater than that for the low dose one, but when normalized the programmes follow similar trajectories at low and medium phantom diameters, while at large diameters the actual mA values for the two options both approach the maximum value.

Zoom In Zoom Out Reset image size

Plots comparing the variation in noise in terms of the measured SD in the image along the conical phantom for different CT scanners are shown in figure 6 for the settings regarded as the standard options. The GE and Toshiba scanners maintained a fairly constant noise level through the middle part of the phantom, whereas the noise level in the Philips scanner rose steadily, although much less than the noise when a fixed mA was used with no ATCM. The noise level for all options tended to rise towards the thicker end of the phantom as the mA approached the limiting value. Each CT scanner offered a range in the choice of image quality level from low dose and to high image quality and plots of the measured noise for the upper and lower levels along the length of the phantom for three of the scanners are shown in figure 7. The SDs of the lowest dose (LD) options were 2–3 times those for the highest image quality (HQ) in each case. There is considerable overlap between the ranges for the different scanners, with the main difference being that the noise varied less with phantom diameter for the GE and Toshiba scanners that used a noise based reference, compared to the Philips scanner. If the main factor contributing to improved contrast in larger patients is the presence of fat, then it might be argued that maintaining the same level of noise throughout the scan of an individual patient is a better approach, provided the appropriate noise reference is selected. However, if ATCM operation is based on a reference image or mAs, as practised by Philips and Siemens, it can be adapted more readily to scanning patients of differing size, as a similar image quality reference can be chosen for all patients. Thus there are arguments in favour of both approaches to the choice of image quality reference, but the important thing for the operator is to understand how the ATCM system on each CT scanner performs and adapt the settings to suit all groups of patients.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Wood et al (2015) describe how, following a dose audit on Philips CT scanners, they were able to optimize examinations based on what was considered to be the optimum protocol. They chose the third and fourth sections of a Leeds CT AEC phantom, made from short sections of PMMA with a diameter close to the standard for Philips scanners of 320 mm WET to compare performance, but another elliptical phantom with a similar diameter could be used as an alternative. They scanned the phantom using the clinical protocol with ATCM to check performance, and using the clinical protocol from the LD scanner as a guide, having checked the image quality was acceptable, they matched settings on other scanners to the protocol chosen. Finally Wood et al confirmed the performance of each scanner by rescanning the conical phantom and measuring CT number and noise along the phantom.

A similar methodology could be used for comparing performances of CT scanners from the same manufacturer, for which the image quality references are similar. For models from Philips and Siemens, scan parameters such as the maximum and minimum limits on tube currents are adjusted automatically for each patient based on the SPR and the modulations in tube current with patient size are similar in form (figure 5). For GE and Toshiba scanners that use an image quality reference based on the noise level, a similar approach based on a phantom of standard dimensions could be used for setting up the clinical protocol, but additional factors would need to be checked to ensure that the adjustments made for patient size were optimized. The various peripheral factors that influence ATCM performance are discussed in more detail in section 4, but the framework for optimization will be set out first. Establishing the CTDI_vol and noise levels required for scanning a phantom of standard size and then checking how they change with ATCM for a geometrical phantom with a range of diameters (section 3.2) provides a method for comparing performance of scan protocols. Relative values of the CTDI_vol for different diameters can then be inferred directly from the changes in tube current relative to the value for the standard phantom as depicted in figure 5. The translation of protocols between scanners from different manufacturers, particularly between ones using ATCM based on a reference image or mAs and those using a noise reference is a more difficult task, because the forms of the variations in tube current and image noise with attenuation are substantially different (figures 5–7). Nevertheless this can be tackled in a similar manner by determining values for CTDI_vol relative to those for a standard phantom from the variation in tube current with phantom diameter (figure 5). Once values of CTDI_vol for different sizes of phantom required for an optimized protocol have been determined, then values for the image quality references that will give similar CTDI_vol values for different sizes of phantom with the scanner to be optimized need to be determined. If the forms of the relationship between CTDI_vol or mA and phantom diameter are substantially different on the two CT scanners (figure 5), then scans of the geometrical phantom using different values of the image quality reference will be required to provide information on the possible settings that might be appropriate. The steps that can be followed in the optimization of clinical protocols are set out in table 2.

McKenney et al (2014) describe protocol translation between CT scanners of different type based on more detailed comparisons of displayed CTDI_vol. They employ a phantom comprising five PMMA cylinders of varying diameter each 150 mm long and determine the dependence of CTDI_vol on the image quality reference from scans of each cylinder over the full range of image quality reference settings. From these results values of the image quality references for different scanners that give similar values for the CTDI_vol can be plotted against each other and relationships between them determined. McKenney et al found that the relationships between noise index values for two GE CT scanners could be described by a linear function of the form (y  =  ax  +  b) and differences in relationships between the GE scanners were relatively small. They found the equivalent relationship between noise index for a GE scanner and the quality reference mAs for a Siemens scanner was more complex, but could be described by a power function of the form (y  =  ax^b  +  c). Such equations could then be used in translation of image quality reference values that should be selected for different protocols between CT scanners. A similar approach could be used based on achieving similar levels of image noise instead of CTDI_vol. This method of translation requires substantial numbers of experimental measurements and analysis. The simpler comparison of the principal image quality settings described in the previous section may be sufficient in many cases, but whatever approach is taken, phantoms with dimensions to represent the range of body sizes and attenuations encountered in clinical practice are required. In addition there are a number of factors that influence the performance of ATCM systems that complicate comparisons and these will be discussed individually in the following section.

Higher tube currents give proportionately lower noise levels in the images and the CT scanner operator must judge the appropriate balance. These choices are affected by factors that influence the mA selection by the ATCM. CT scanner manufacturers use the SPR to plan the ATCM to be employed in a scan. There have been a number of reports about the effect of SPR on the dose and noise levels. For example, when a single postero-anterior (PA) SPR is used instead of an AP one, dose levels 30%–60% higher have been reported for GE scanners (McNitt-Gray 2011) and 13% higher for Siemens (Söderberg 2016). These differences may be due predominantly to geometrical factors. For example the projected sizes of the vertebrae and ribs are greater in the PA projection as the spine is closer to the x-ray tube. Since the ATCM uses information about the density, size and shape of the patient tissues, the ATCM will tend to select a higher mA, because the higher density structures appear to be larger in SPRs acquired with the PA projection. Moro et al (2013) observed that for the GE scanner the overall body dimensions were also larger in PA SPRs than in AP ones, and this may be the reason for the proportionately larger increases in mA planned on PA SPRs reported for some GE scanners.

When one SPR is used for planning the ATCM, effective mAs values about 15% higher have been reported for lateral SPRs as opposed to AP ones with Philips scanners (Sookpeng et al 2015) and GE ones (McNitt-Gray 2011). However Papadakis (2007) reported the reverse, so factors may vary for individual patients and scanners and assessments should be carried out for each CT scanner. Söderberg (2016) reported a 20% decrease in CTDI_vol with a Siemens Sensation 16 scanner when performing two SPRs, the lateral and either AP or PA. For GE scanners the SPR performed last is the one used in planning the ATCM (Moro et al 2013) and performing the AP view last is recommended. For Toshiba CT scanners, mA modulation along the body is planned on the first SPR and the second one is used to derive the AP lateral current ratio. If the second SPR is not performed, a standard ratio will be used. But if a second scan is performed it must use the same exposure settings to allow a direct AP/Lateral comparison and failure to do this will change the operation of the ATCM. Using two SPRs is probably better in general for most CT scanners, if the scanner is designed to utilize both. The changes in ATCM performance have been referred to in terms of the change in dose level, but decreasing dose will lead to an increase in noise and so poorer image quality and all factors must be borne in mind when selecting the optimal settings. As iterative reconstruction and new techniques offer the potential for sub-mSv CT imaging, the SPR will become a more significant proportion of the dose for a scan, so use of a single SPR may become more appropriate.

Another factor related to the SPR that influences ATCM performance is the height of the patient couch, as this will also change the apparent patient dimension recorded on the SPR, which will be larger if the patient is closer to the x-ray tube (figure 8) (Matsubara et al 2009, Supanich 2013). Scanning with the couch off-centre will also result in misalignment with respect to the bow-tie filter and this will affect the dose distribution (Habibzadeh et al 2012, Kataria et al 2016). Therefore ensuring that the patient is centred within the scan field is important for optimizing performance. This issue has been appreciated by the manufacturers and the latest versions of Philips DoseRight and Toshiba SureExposure 3D will compensate for incorrect vertical positioning of the patient (Zhang and Ayala 2014).

Zoom In Zoom Out Reset image size

Generally scans will be performed within the bounds of the planning SPR, but if a region beyond the boundary is scanned, the behaviour of CT scanners varies. For the Philips 64 system the mA goes to the maximum, for Siemens it goes to that for the last known location used, for GE scanners the mA goes to the minimum for the protocol, and for Toshiba it goes to a default manual setting (Supanich 2013).

ATCM systems modulate mA according to the attenuation of the patient. Those using a reference image or mAs set the upper limit to that required for the most attenuating part of the patient based on the SPR, and scanners modulate the current within the chosen range, so there is no requirement to set mA limits. For scanners that employ an image noise reference, the maximum and minimum currents can be set by the operator, although values are recommended. Inappropriate choice of current limits can curtail modulation, for example setting of a minimum current of 100 mA was found to prevent any modulation for small patients on a GE Lightspeed scanner (Sookpeng et al 2014). Since contrast in images of small patients tends to be poorer, the minimum mA should be set to ensure that the image quality will always be adequate. The maximum mA limit can restrict current modulation for some patients, but the doses for larger patients have sometimes been found to be higher than necessary to achieve adequate image quality on scanners using a noise reference (Sookpeng et al 2014), so an appropriate choice of the maximum mA can provide a way in which higher dose levels can be avoided. The maximum mA limit can also be set to allow scans to be performed with a fine rather than a broad focus to achieve better resolution. The maximum achievable mA available on a CT scanner will be determined by tube performance characteristics, but if a larger mAs is required for a particular patient, then it may be necessary to increase the tube rotation time.

The image slice thickness changes the number of photons contributing to an image. The ATCM systems based on a reference mAs or reference image (Siemens and Philips) use the same mAs and will give a different noise level when the image thickness is changed (Sookpeng et al 2015). For scanners using an image noise reference (GE and Toshiba), a reduction in the image thickness used for acquisition by a factor of n will be accompanied by an increase in mAs by a factor of the order of  √n in order to maintain the same noise level (Gutierrez et al 2007, Sookpeng et al 2015). Note that this is the slice thickness that is used for the first axial reconstruction, not values chosen subsequently for the volume reconstruction images viewed by the radiologist. Use of a standard thickness recommended by the manufacturer is generally appropriate, but if the slice thickness is reduced, the reference noise would need to be increased by about  √n times if it was desired to maintain the same mAs. Modified values for the noise index have been recommended by GE for different slice thicknesses (Kanal et al 2007, McNitt-Gray 2011). The factors involved in changing the slice thickness for acquisition of scan data are large, and it is important to understand how adjustments will change the behaviour of the ATCM on each scanner. Another factor that could potentially alter mA selection by the ATCM is the pitch. Generally when pitch is lowered the mA is reduced to achieve the specified image quality reference, and the relationship between dose and image quality remains the same (Papadakis et al 2007). However, Rampado et al (2009) reported that the CTDI_vol increased as the pitch was reduced on a GE Lightspeed scanner with ATCM.

Numerical filters are applied to each pixel within digital images either to smooth the images to reduce the noisy appearance or to sharpen them to accentuate tissue boundaries. The filters are based on weighted averaging of signals from pixels surrounding each pixel in the image. The filters utilize kernels containing the weighting factors for surrounding pixels, which depend on their relative position. Smoothing filters give a positive weighting to nearby pixels to give an averaging effect, whereas sharpening filters give a high weighting to the value of the pixel itself and a negative weighting to surrounding pixels, which preserves higher spatial frequencies in order to enhance edges. This type of filter sharpens boundaries, but creates a noisier image. The type of filter selected will in most cased not affect the mAs selected by the ATCM, but alter the appearance of the image. However, in Toshiba scanners that use a target noise reference, changes in the mAs for the scan occur when different filters are selected for the first axial reconstruction (Sookpeng et al 2015).

For setting up ATCM systems, the first thing to establish is that dose and image quality performance on a standard phantom are satisfactory and decide the appropriate image quality references to be used. This should be carried out preferably using a phantom with an elliptical cross section similar to that of the standard reference phantom that might be used for the scanner, or an appropriate position in a conical or sectional phantom designed for assessment of ATCM systems. A framework for the whole process is set out in table 2, but it is important to ensure that other factors are considered. Appropriate SPRs should be used to plan scans and to check whether the order in which they are performed affects current modulation. Ensure that radiographers understand the importance of centring the patient within the CT gantry for the SPR. For scanners using an image noise reference as the basis for ATCM adjustment, first assess the optimum maximum and minimum mA settings to ensure that mA is modulated over the desired ranges, and consider whether the maximum should be set to limit the dose level for larger patients. Secondly check that the recommended settings for the slice thickness used for acquisition are selected and if these are changed then appropriate settings of the noise reference are used. Another factor for some scanners is the direction in which the scan is performed and the influence this has on the starting mA and overall dose performance should be considered (Gutierrez et al 2007, Sookpeng et al 2013). A likely outcome of optimization for scanners using a noise level reference is that a higher reference noise level is used for larger patients, if this is not selected automatically.

The measurements described in this paper are ones that should ideally be performed in the first six months after installation. The aim is to understand how different factors influence the performance of the ATCM, and feed information into setting up clinical protocols for a range of examinations. This should ideally be a collaborative project involving medical physicists, radiologists and radiographers, as recommended in COMARE (2014). Once protocols have been established, then simpler tests, which might be scanning a geometrical or anthropomorphic phantom with standard settings, should be performed periodically and results compared with those established following the early testing phase.