Multidetector CT in children: current concepts and dose reduction strategies

Adequate prescan patient preparation should include:

This will help to reduce the anxiety of the child and positively influence their mood, increasing the success rate of the CT examination without the need for sedation or general anaesthesia. In general, children 5 years of age and older will be able to undergo a CT examination without sedation after thorough patient instruction. However, there are still situations in which sedation or general anaesthesia will be required depending on the type of investigation, age and mental ability of the child and the clinical situation and question. The way sedation and general anaesthesia is organised depends largely on local agreements and legislations, and falls beyond the scope of this article. The interested reader is referred to recent literature on this topic [31, 32].

It is worth mentioning the possibility of fixation by use of a vacuum pillow. Vacuum pillows are composed of an airtight flexible cover filled with small solid polystyrene balls. This pillow can easily be wrapped around the baby or the part of the body to be investigated. By sucking it, the vacuum one creates provides firm but soft fixation, ideal for the fixation of babies or certain parts of the body (e.g., head, shoulder, extremity) in infants and preventing the need for sedation or anaesthesia (Fig. 1).

As small children lack large amounts of intra-abdominal fatty tissue, interpretation of a CT of the abdomen will be more difficult than in adults. That is why we prefer US as the imaging modality of first choice for abdominal clinical problems in childhood. If CT of the abdomen is indicated, adequate oral contrast intake is often essential for the evaluation. Depending on the age of the child, a diluted iodinated or barium-based oral contrast agent is given by mouth or through a nasogastric tube during 24 h prior to or 1 h and 15 min before the examination (Tables 1 and 2) [33, 34]. In general, adequate opacification of the small intestine and proximal part of the colon is achieved with the latter (biphasic) approach, making it the preferred method in children. The dilution factor of these oral contrast agents (CM) is usually higher than in adults [1–2% instead of 3% of iodinated CM (300–350 mg I/ml)], which provides better contrast with low tube voltage techniques. In infants and young children a non-ionic iodinated oral contrast agent is often preferred because of the risk of aspiration. These high-density oral contrast agents are especially well suited for the differentiation of hypodense or cystic lesions from intestinal loops. However, there is increasing evidence that the use of low-density oral contrast agents may be the preferred method for opacification of intestinal loops [35, 36]. Examples of these agents are water, full-fat milk or a polyethylene glycol or mannitol (1.5%) suspension. These low-density contrast agents are especially well suited for the evaluation of mucosal lesions of intestinal loops and intraluminal or submucosal bleeding.

For the administration of IV contrast material, the use of a power injector instead of hand injection is preferred. Therefore, the largest possible IV cannula should be placed, preferably in the antecubital vein [33, 34, 37]. It is advocated to place this cannula in a comfortable environment at the in-patient or out-patient ward and after application of a local anaesthetic gel or ointment (e.g., lidocaine hydrochloride gel, Astra-Zeneca, London, UK). Central venous lines (CVL) or portacath systems can be used but are often restricted by their maximum pressure range of 30–40 pounds per square inch (psi), whereas the minimum adjustable pressure of the power injector is usually 50 psi. This means that, when using a CVL, one should inject the contrast agent manually for medico legal reasons. However, based on several recent publications and our own experience power injection of IV contrast material through a CVL is possible with a low complication rate [38‐41]. Recently, so-called PowerPort (Bard Access Systems Inc, Salt Lake City, UT, USA) implantable portacath systems have become available, combining reliable venous access with the ability for power-injected CM administration. In case of a peripherally inserted central catheter (PICC) or cannula sited on the hand, foot or skull, the contrast material can only be administered safely by manual injection.

The standard contrast material used for IV administration is a non-ionic, low-osmolar contrast agent with a concentration between 240 and 400 mg I/ml (most frequently 300 mg I/ml). As yet we are not aware of any indication for the use of high-concentration CM (370–400 mg I/ml) in children, except perhaps in cases of extreme obesity or the need for very low injection rates (e.g., in cases of fine-needle calibres). The standard CM dose is 2.0 ml/kg (concentration 300 mg I/ml), which is comparable to 600 mg I/kg. However, in case of a CT of the brain, neck or thorax usually a dose of 1.5 ml/kg (450 mg I/kg) will suffice. It is safe to increase the dose to 3.0–4.0 ml/kg (or a maximum of 150 ml) even in children, assuming that the renal function is normal. This can be of value if simultaneous opacification of the arteries and veins is needed.

In children the circulation time varies widely which makes adequate scan timing more difficult. Furthermore, the size, position and type of cannula will differ among different age groups, with as a consequence varying injection rates (Table 3). In general, an injection rate of 2.0 ml/s suffices for most paediatric indications, especially when younger than 12 years of age. When using a CVL, the injection rate should be maximized to 2.0 ml/s for safety reasons. An empirically determined fixed delay time usually suffices for most routine indications, especially in the younger age group. However, the routine use of bolus tracking techniques is strongly recommended for most body indications, especially in case of a CT angiography (CTA) or arterial-phase CT [42]. An alternative for scan timing is the test bolus technique, although this technique is not suitable in very small children as the total volume of contrast material available is often too small. Both techniques share the disadvantage of additional (monitoring) scans, increasing the radiation dose for the child. This additional dose should be weighed against the benefit of improved and individualized scan timing, and when applied the monitor scans should be obtained with a low-dose technique to limit this additional radiation dose. Table 4 summarises some guidelines for delay timing based on the indication for the MDCT scan. More details on IV CM injection issues in MDCT are given in three recent comprehensive overviews [42‐44].

Multiphase CT examinations in children should be avoided. The use of precontrast CT scans hardly ever results in clinically relevant extra information and usually should be abandoned [53]. If further characterisation of a lesion by multiple pre- and post-CM phases is indicated, MRI is currently the preferred imaging modality of choice. Several patients will receive multiple follow-up scans [15, 54]. These scans should be kept to a minimum, and a low-dose technique often will suffice [55, 56]. Furthermore, other imaging modalities such as US or MRI should be used if possible.

As children are usually smaller than adults a higher spatial resolution is needed, especially along the z-axis. This can be achieved by choosing a thin collimation, thus creating an (near-) isotropic resolution of the raw dataset consisting of voxels with (nearly) equal dimensions in all directions. However, this will increase the image noise and as small children usually lack visceral fatty tissue, less noise can be tolerated than in adolescents and adults. Furthermore, in older generation scanners thin collimation negatively influences the so-called overbeaming effect resulting in an increase in radiation dose, which will be discussed later. Therefore, the choice of collimation depends on the clinical question and size of the patient and should balance the necessary z-axis resolution, gain in three-dimensional (3-D) reformatting possibilities, noise level and low radiation dose level. The use of (near-) isotropic datasets is preferred for most indications, as the problem of the higher noise level usually can be solved by using the “scan thin—view thick” approach while interpreting the images (see also the “Image reconstruction” section).

An increase in pitch can result in a shorter scan time and (in some scanner types) in a dose reduction for the patient. However, in modern MDCT scanners this may not be the best option. This is explained by the negative dose effects of the so-called overranging which will be exaggerated by increasing the pitch. Furthermore, the spatial resolution will decrease by increasing the pitch. In those scanner types in which effective mAs is used, an increase in pitch will result in an increase in the tube current. Therefore, it is usually more dose efficient to keep the pitch as low as possible (<1) and if needed manually decrease the tube current, in order to achieve a similar tube current/pitch ratio (effective mAs) as would be the case with a higher pitch.

The scan FOV should be tailored as much as possible to the size of the body region of interest. The major advantage of a smaller scan FOV is the higher spatial resolution, as the pixel size decreases with smaller FOV. The effect of the display FOV on resolution is often different—while some increase in resolution may result, after a certain threshold pixels are only blown up.

Due to the smaller size of children it is usually possible to lower the tube voltage with maintenance or even improvement of the diagnostic image quality and resulting in a significant dose reduction. In most children a tube voltage of 80–100 kVp will suffice, especially in children with a body weight <45 kg. In adolescents, a tube voltage of 100 kVp for the thorax and 120 kVp for the abdomen is usually sufficient. However, the scanner-related parameters, such as, scanner geometry, tube filtration, detector design and efficiency, sometimes negatively influence the image quality with lower tube voltages, for instance by inducing artifacts. CT examinations with a high intrinsic contrast, such as in the chest, bones and in CTA, also justify lowering the tube voltage to 80–100 kVp. Recent studies with phantoms suggest that the optimal tube voltage in children may be even lower (approximately 60 kVp), at least for some indications [Kachelriess, unpublished data]. In this scope, it is appropriate to assume that slightly different tube voltages (for example 60 kVp compared to 120 kVp) are related to almost similar radiobiological effects. Furthermore, the use of lower tube voltages is related to a relative reduction in the production of scattered radiation.

By lowering the tube current a direct proportional decrease in the radiation dose is achieved with an increase in noise, a major drawback, but as long as the images are of diagnostic quality this increase in noise should and can be accepted [52]. Although several studies in phantoms have shown that the mA can be halved for each 3.5- to 4-cm reduction in body diameter (half value layer, HVL), in clinical practice a less stringent HVL of 4–6 cm is used in children. This is because small children usually lack large amounts of visceral fatty tissue and, therefore, less image noise is desirable. Based on the aforementioned, several body size or weight-based paediatric CT protocols have been suggested [33, 52, 56‐59]. One vendor even has adapted the Broselow-Luten system from the emergency room into colour-coded paediatric protocols based on patient weight or length, in order to reduce errors during scanning [33, 57]. However, one has to keep in mind that the body weight alone may underestimate the dose requirements in obese children, because their body diameter may be larger than that of a taller child of identical weight.

Most modern MDCT scanners apply tube rotation times of 0.3–0.5 s resulting in shorter examination times. This is advantageous in children as movement and respiration artifacts are reduced. Moreover, the need for sedation and anaesthesia is significantly reduced. However, one has to keep in mind that a shorter rotation time can result in a decreased number of profiles that will be used in image reconstruction, with the consequence that the image noise will increase. Therefore, in terms of image quality a rotation time of 0.5 s is often the best option.

Almost all modern MDCT scanners are currently equipped with some type of automatic exposure control (AEC) or automatic tube current modulation (ATCM) technique. These techniques dynamically control the tube current during scanning based on user settings and adapted to the body geometry seen on the scanogram. In this way, image quality can be improved due to a constant noise level in all slices and radiation dose can be reduced. Modulation is possible in the axial plane (angular- or XY-modulation), along the long axis of the patient (longitudinal- or Z-modulation) or by a combination of both techniques (combined- or XYZ-modulation, Fig. 2) [60].

The modulation technique and settings used differ considerably among the different vendors (Table 5) [60, 61]. Dose Right (Philips Medical Systems, Cleveland, OH, USA) asks the user to introduce reference images from prior studies into a database that will be used as reference image quality for new examinations. CARE Dose 4D (Siemens Medical Solutions, Forchheim, Germany) uses a comparable technique, with the exception that the user is asked to set a reference median effective mAs (tube current x rotation time/pitch) for the entire study based on a specific predefined image quality in a 20 kg paediatric patient (80 kg for adults). The AutomA technique (GE Healthcare, Waukesha, WI, USA) uses the so-called noise index (NI), defined as the noise level in the centre of a water phantom using a reference scan method with the selected kVp, 200 mAs, selected slice thickness and a standard reconstruction kernel. Finally, Sure Exposure (Toshiba Medical Systems, Otawara-Chi, Japan) is comparable with the former system. The primary input parameter in this system is the standard deviation (SD), reflecting the noise in all positions along the Z-axis of the patient acquired in a circular water phantom comparable to the size of the patient.

By applying ATCM techniques it is usually possible to achieve a reduction in radiation dose to the patient [51, 60, 62]. However, the amount of reduction depends largely on the user-defined settings, anatomical region scanned and, last but not least, to what extent the CT protocols already have been optimised [25, 60, 63, 64]. For instance, if ATCM is used in the chest the tube current will increase in the shoulder and liver region to maintain a constant noise level. This constant noise level is not always necessary to answer the clinical question. The German paediatric CT survey 2005–2006 by Galanski et al. [25] showed that the ATCM techniques of GE and Toshiba correct more strictly than those of Philips and Siemens, resulting in a somewhat higher patient dose compared to manual adaptation of the CT parameters. Furthermore, specific paediatric ATCM settings are not universally available on the modern scanner systems and substantial scientific literature regarding appropriate weight-, age-, body region- and indication-based reference settings is still lacking. That is why ATCM techniques should be used with care in children.

There are two dose-increasing effects that form an integral part of the image acquisition with MDCT. First, there is the overbeaming effect. In MDCT, the reconstruction algorithm requires a homogeneous illumination of all detectors. Therefore, it is necessary to widen the bundle by opening up the pre-patient collimation. This results in the formation of a penumbra outside the detector width, thereby reducing the dose efficiency (Fig. 3). The size of the penumbra depends on the focal spot size and the ratio between (a) the distance between collimator and detector, and (b) the distance between the focal spot and detector. In modern CT scanners the penumbra measures 1–1.5 mm on both sides of the collimated bundle. As the penumbra is almost constant, the dose-increasing effect of overbeaming decreases with increasing number of detectors and slice collimation width. The importance of overbeaming is minor in MDCT scanners with 32 slices and more.

The second effect is called overscanning or overranging (Fig. 4). In the helical scan mode the reconstruction algorithm requires additional raw data on both sides of the planned scan volume. These data, necessary for image reconstruction, are acquired by an additional rotation on both sides and outside the planned scan length. This will lead to radiation exposure of tissue outside the region of interest, increasing the radiation dose to the patient. The overrange length increases with increasing table movement, e.g., in case of increasing number of acquisition channels, increasing bundle collimation and/or increasing pitch. In addition, this effect is relatively more important in smaller scan lengths such as in small children. To limit the dose-increasing effect of overranging, the bundle collimation and pitch should be kept as small as possible [65].

Overbeaming and overranging always occur together but have opposite effects on the radiation exposure when changing the number of active detectors or bundle collimation. Nagel et al. [66] have calculated that for a small scan length (small children) in body CT applications, the optimal beam collimation to reduce these negative radiation dose effects will be in the order of 10–20 mm (Fig. 5). Therefore, when using a 64 slice scanner in children it will be beneficial to lower the table feed by going to a 16-slice or 32-slice mode, depending on the range that needs to be examined.

In modern scanners with 128 slices or more, overranging is the most important radiation dose increasing effect, limiting the use of these scanners in children. However, in the newest scanners (e.g., Philips iCT 256-slice, Toshiba Aquilion ONE 320-slice, Siemens Definition AS/AS+ 128-slice and Definition Flash dual-source 128-slice scanners) this effect has been (partly) eliminated by the introduction of so-called dynamic or adaptive collimators [67]. These collimators automatically move in and out at the beginning and end of the scan length, thereby hindering the bundle to reach the patient outside the planned scan volume. Furthermore, newer types of “step-and-shoot” scanning are also beneficial in children. In the volume scan mode the Toshiba Aquilion One 320-slice CT scanner has a scan range of 160 mm. In neonates and infants, this scan range is sufficiently large to make a low-dose CT of the chest or abdomen in one rotation of 0.35–0.4 s without overlap or overranging. By applying two consecutive scan volumes with a small overlap it will be possible to scan relatively dose efficiently, even in older children. However, with more than three volumes, the added radiation dose from overlap becomes equivalent to the overrange effects of helical scanning.

Dose optimisation in children usually results in quite noisy CT datasets. When interpreting the diagnostic images, the “scan thin—view thick” principle is important. By stacking thin slices in the viewing direction, the image noise is decreased considerably whereas the resolution is maintained in all directions. However, this requires a PACS environment with adequate multi-planar reconstruction (MPR) facilities or a 3-D workstation.

When acquiring CT datasets with thin slices, it is important to realise that most image reconstruction algorithms in MDCT still have problems with the reconstruction of very thin slices. A slice thickness equal to the slice collimation is associated with an additional amount of noise (up to 20%), added to the expected noise of thin slices. In order to minimise this effect, it is recommended to reconstruct slightly anisotropic voxels by increasing the slice thickness by 20–40% compared to the slice collimation.

Another important aspect of image reconstruction is the relation between resolution, noise and radiation dose. The image resolution depends largely on the scan FOV, reconstruction kernel (in the scan plane) and the effective slice thickness (along the patient axis). An increase in resolution will result in an even larger increase in noise, as the noise is proportional to the fourth power of the increase in resolution. Therefore, it is important to avoid the use of overly thin slices and “hypertrophic” voxels with a smaller dimension in the Z-axis than in the XY-plane. This can be achieved by selecting a reconstruction index that is larger than the pixel size (= FOV/matrix). With a (paediatric) FOV of 240–320 mm and a 512² matrix this will be approximately 0.5–0.6 mm.

In this scope, it is noteworthy that the image noise in reconstructed images from low-dose CT datasets can be decreased by applying a smooth reconstruction kernel. However, for the evaluation of the lung, this is usually not the optimal choice. In this case, it is better to apply a high-resolution lung or bone kernel, provided that the noise level of the dataset is not too high.