Neuroradiology/Head and Neck Imaging
Original Research
Optimizing the Balance Between Radiation Dose and Image Quality in Pediatric Head CT: Findings Before and After Intensive Radiologic Staff Training
Audio Available | 
Share
OBJECTIVE. The purpose of this study was to assess the radiation dose and image quality of pediatric head CT examinations before and after radiologic staff training.
MATERIALS AND METHODS. Outpatients 1 month to 14 years old underwent 215 unenhanced head CT examinations before and after intensive training of staff radiologists and technologists in optimization of CT technique. Patients were divided into three age groups (0–4, 5–9, and 10–14 years), and CT dose index, dose-length product, tube voltage, and tube current–rotation time product values before and after training were retrieved from the hospital PACS. Gray matter conspicuity and contrast-to-noise ratio before and after training were calculated, and subjective image quality in terms of artifacts, gray-white matter differentiation, noise, visualization of posterior fossa structures, and need for repeat CT examination was visually evaluated by three neuroradiologists.
RESULTS. The median CT dose index and dose-length product values were significantly lower after than before training in all age groups (27 mGy and 338 mGy ∙ cm vs 107 mGy and 1444 mGy ∙ cm in the 0- to 4-year-old group, 41 mGy and 483 mGy ∙ cm vs 68 mGy and 976 mGy ∙ cm in the 5- to 9-year-old group, and 51 mGy and 679 mGy ∙ cm vs 107 mGy and 1480 mGy ∙ cm in the 10- to 14-year-old group; p < 0.001). The tube voltage and tube current–time values after training were significantly lower than the levels before training (p < 0.001). Subjective posttraining image quality was not inferior to pretraining levels for any item except noise (p < 0.05), which, however, was never diagnostically unacceptable.
CONCLUSION. Radiologic staff training can be effective in reducing radiation dose while preserving diagnostic image quality in pediatric head CT examinations.
Keywords: MDCT, pediatric head CT, radiation dose, radiation protection, staff training
Since its introduction in the early 1970s, CT has had a dramatic evolution that has greatly improved its diagnostic performance in many clinical scenarios and vastly broadened its field of application. Even if CT represents only 11% of radiologic procedures, it accounts for as much as 70% of the total effective dose from all diagnostic radiologic studies [1–3]. However, despite clear evidence that CT can provide fundamental information for diagnosis and patient care, the risk of malignancy induced by ionizing radiation from CT examinations must be carefully considered, with specific attention to pediatric patients [4, 5]. Brenner et al. [6], using data extrapolated from atomic bomb survivor mortality rates, in 2001 showed that a child's risk of development of fatal cancer from the radiation of a single CT examination was 1 in 1000. In addition, a significant increase in cancer incidence has been found in populations exposed to ionizing radiation for medical purposes before adulthood [7–10]. Radiation-related cancer risk varies significantly with age, and among infants it is approximately 2–4 times the risk among adolescents and up to 10 times the risk among adults [11]. The attributable lifetime risk from a head CT examination of a 1-year-old child is 0.07%, and though it represents a relatively minor increase of the individual background rate of malignancy, its effect is dramatically magnified on a population basis as millions of head CT scans are annually obtained for young patients around the world [12, 13].
Children are considered at greater risk of radiation-induced cancer than adults are owing to their higher biologic sensitivity to ionizing radiation and to their longer life expectancy, resulting in more time for potential radiation-induced cancer to develop [14]. Because smaller patients attenuate x-rays less than larger ones, the radiation dose used for pediatric CT studies can usually be significantly reduced without compromising diagnostic image quality [15]. This balance between radiation dose and image quality is particularly important for potentially life-saving diagnostic procedures, such as head CT of trauma patients, which is being more and more commonly required for the assessment of children [16] and is associated with a nonnegligible and largely variable radiation dose [17]. In addition, studies performed with patients and workers exposed to ionizing radiation have shown that brain tissue is more sensitive to radiation damage than previously thought. There is a small but significant association between radiation exposure and risk of brain cancer and cerebrovascular disease [7, 8, 18–20].
For many years pediatric CT examinations have been performed with adapted adult protocols instead of natively pediatric ones. However, an arbitrary reduction of radiation dose may potentially result in substantial loss of image quality (especially if dose-saving tools compensating for higher image noise and artifacts are not applied), highlighting the need for radiologists and technologists to optimize CT protocols in an effort to balance image quality and radiation dose [21].
The aims of this study were to evaluate the radiation dose of pediatric head CT examinations performed at three different radiologic sites at the same institution and to assess the effect of intensive radiologic staff training on delivered radiation dose and image quality of examinations performed with conventional, commercially available MDCT equipment.
| Materials and Methods |   |
|---|
We collected data on the radiation dose absorbed by individual pediatric patients undergoing head CT examinations at three different radiologic centers in our institution. The examinations were performed in two time periods: January 1–June 30, 2011, and January 1–June 30, 2012. All centers involved in the study were general radiology departments of the national health system with an overall annual rate of approximately 6000 CT examinations in 2010. CT examinations were performed with one 64-MDCT scanner (center 1; LightSpeed VCT, GE Healthcare), one 40-MDCT scanner (center 2; Somatom Sensation 40, Siemens Health-care), and one 16-MDCT scanner (center 3; Aquilion 16, Toshiba Medical Systems). All images acquired at the three centers were sent to the same pool of consultant neuroradiologists via our hospital PACS for image viewing and reporting.
All CT systems were equipped with patient size– based automated tube current modulation algorithms, as provided by each manufacturer (center 1, Smart mA, GE Healthcare; center 2, CARE Dose, Siemens Healthcare; center 3, SureExposure, Toshiba Medical Systems). A total of 215 unenhanced head CT examinations of outpatient boys and girls 1 month to 14 years old were randomly selected. The pediatric population was divided into three different groups based on patient age (0–4, 5–9, and 10–14 years), and a number of no less than 30 CT studies for each group were evaluated. All selected CT examinations were free from artifacts (e.g., those due to metallic hardware or external devices) severe enough to compromise diagnostic yield. The examinations consisted of scout images plus a stack of 2.5-or 3-mm-thick axial images (depending on scanner type), which were used for image analysis. For each study, the following parameters were manually retrieved from the PACS: tube voltage, tube current, tube rotation time, acquisition mode (sequential or helical), beam pitch (for helical CT studies), detector configuration, automated tube current modulation (enabled or disabled), scan FOV, slice thickness, CT dose index (CTDI), and dose-length product (DLP).
Between July 1 and December 31, 2011, an intensive training course was organized in an attempt to verify and improve the professional performance of the radiology staff operating in the CT suite. Radiologic staff training simultaneously involved radiologists, medical physicists, and technologists and consisted of two consecutive full-day training events in which medical, biological, and technical topics related to CT (appropriateness criteria, protocol optimization, assessment of cancer risk due to ionizing radiation exposure, risk communication to patients and referring physicians, ethical and legal issues, and continuing education) were covered from both the theoretic (lectures on the principles of CT and updates from the current literature) and practical (hands-on training at the CT console) standpoints as described previously [22]. The training course was led by a group of senior radiologists and technologists with scientific and clinical experience in the fields of radiation protection, CT technology, and medical applications. The level of learning was assessed through a test administered at the end of each course, and CT protocols were progressively revised through continuous feedback among all involved staff technologists and radiologists by keeping track of individual patient characteristics, diagnostic query, radiation dose, and image quality for every CT examination. The effect of training on delivered CT radiation dose was evaluated 6 months after training with the second round of CT data collection, conducted similarly to the first.
Circular ROIs of the same size (3-mm diameter) were placed in identical locations for each CT examination, and mean and SD of CT density were measured inside each ROI at a commercially available workstation (Advantage Windows 4.4, GE Healthcare). ROIs were placed in the gray matter and white matter of selected regions at the level of the basal ganglia as shown in Figure 1. Attention was paid to avoiding inclusion of nearby structures. Gray matter conspicuity was defined as follows:
 |
where GM is gray matter and WM is white matter. Contrast-to-noise ratio (CNR) was defined in a standard manner, as follows:
 |
where SD is the SD [23].
 View larger version (318K) | Fig. 1 —9-year-old girl with head trauma. CT image shows placement of ROIs in gray (solid arrow) and white matter (dashed arrow) for measurement of gray matter conspicuity. |
Three independent observers with 7, 8, and 9 years of experience in neuroradiology visually evaluated the same set of 60 consecutive pediatric head CT studies acquired before and after the training event (i.e., 30 before and 30 after training for each reader) by reviewing axial images in 2D mode at the workstation. Readers were blinded to the technical parameters of the CT datasets, which were presented randomly. Images were ranked on the basis of the following five items: artifacts, such as beam hardening and partial volume effect; gray-white matter differentiation; noise; visualization of posterior fossa structures; and need for repeat CT examination. All readers graded gray-white matter differentiation, noise, and visualization of posterior fossa structures on a 5-point Likert scale (1, unacceptable; 2, suboptimal; 3, average; 4, better than average; and 5, excellent). A 3-point Likert scale was used to assess the presence and severity of artifacts (1, present and affecting diagnosis; 2, present but not affecting diagnosis; 3, absent) and the need for repeat CT examination (1, definitely indicated; 2 possibly indicated; 3, not indicated).
CTDI and DLP values were expressed as median, interquartile range (IQR), and highest and lowest observed values. Two-tailed independent Student t tests were performed to compare CTDI, DLP, gray matter conspicuity, and CNR values for each age group before and after radiologic staff training. Moreover, tube voltage (100, 120, or 140 kV) and tube current–rotation time product settings were compared for each age group before and after radiologic staff training by use of the Mann-Whitney U test and the independent Student t test. CTDI, DLP, tube current–time setting, gray matter conspicuity, and CNR distributions were skewed to model the log transformation to better satisfy the Student t test assumption of normally distributed outcomes.
Interrater agreement between the three readers for subjective image quality scores was assessed with the single score intraclass correlation coefficient, which was interpreted as follows: 0–0.2, poor agreement; 0.3–0.4, fair agreement; 0.5–0.6, moderate agreement; 0.7–0.8, strong agreement; > 0.8, almost perfect agreement. For each reader, intrarater agreement for image quality scores before and after training was estimated by use of the Mann-Whitney U test. A value of p < 0.05 was set as threshold for statistical significance. All statistical analysis was performed with software (Graph-Pad Prism version 5, GraphPad Software).
| Results | |
|---|
Figures 2 and 3 show observed dose values within the various age groups for head CT examinations performed before and after staff training. The training led to a statistically significant reduction in median CTDI and DLP values in all age groups: from 107 mGy and 1444 mGy ∙ cm to 27 mGy and 338 mGy ∙ cm in the 0- to 4-year-old group, from 68 mGy and 976 mGy ∙ cm to 41 mGy and 483 mGy ∙ cm in the 5- to 9-year-old group, and from 107 mGy and 1480 mGy ∙ cm to 51 mGy and 679 mGy ∙ cm in the 10- to 14-year-old group (p < 0.001). In percentage, posttraining reduction in CTDI and DLP compared with the pretraining period amounted to 64.6% and 69.1% in the 0- to 4-year-old group, to 44% and 52.8% in the 5- to 9-year-old group, and to 49% and 57.2% in the 10- to 14-year-old group. In addition, dose variability in terms of IQR of both CTDI and DLP values was lower after than before training in all age groups, especially the 0- to 4-year-old group, in which the pretraining IQR of CTDI and DLP ranged from 28 to 118 mGy and from 554 to 1705 mGy ∙ cm (with approximately tenfold variation between minimum and maximum values), whereas the posttraining IQR of CTDI and DLP ranged from 21 to 31 mGy and from 187 to 453 mGy ∙ cm.
 View larger version (35K) | Fig. 2 —Box plot shows CT dose index (CTDI) of pediatric head CT examinations performed before and after staff training. Data are median (horizontal line inside boxes), interquartile range (boxes), and highest and lowest values (ends of whiskers). p < 0.05 indicates statistical significance. 0–4, 5–9, and 10–14 indicate patient age groups in years. |
 View larger version (37K) | Fig. 3 —Box plot shows dose-length product (DLP) of pediatric head CT examinations performed before and after staff training. Data are median (horizontal line inside boxes), interquartile range (boxes), and highest and lowest values (ends of whiskers). p < 0.05 indicates statistical significance. 0–4, 5–9, and 10–14 indicate patient age groups in years. |
Table 1 shows the median tube current–rotation time product and tube voltage values before and after training. Tube current–rotation time products were significantly reduced after training, especially in younger children (i.e., from 420 to 125 mAs in the 0- to 4-year-old group, from 280 to 217 mAs in the 5- to 9-year-old group, and from 420 to 250 mAs in the 10- to 14-year-old group; p < 0.001). Automated tube current modulation was enabled in 60% of cases after training (taking into account manufacturer indications for each CT scanner) as opposed to 10% before training. Training also brought a significant reduction in tube voltage. Whereas 140 kV was the most commonly used setting (54.1% in the 0-to 4-year-old group, 50% in the 5- to 9-year-old group, and 71.4% in the 10- to 14-year-old group) and the lowest setting, 100 kV, was seldom used before training (18.9% in the 0-to 4-year-old group, 5.6% in the 5- to 9-year-old group, and 0% in the 10- to 14-year-old group), after training the 120-kV setting prevailed (64.3% in the 0- to 4-year-old group, 85.3% in the 5- to 9-year-old group, and 91.7% in the 10- to 14-year-old group) and the 100-kV setting was more frequently used in all age groups (35.7% in the 0- to 4-year-old group, 14.7% in the 5- to 9-year-old group, and 5.6% in the 10- to 14-year-old group; p < 0.001).
Gray matter conspicuity and CNR values as measured in CT examinations before and after training are tabulated in Table 2. Posttraining gray matter conspicuity (median, 0.25; IQR, 0.20–0.28) was never inferior to pretraining levels (median 0.20; IQR, 0.19–0.22) and was significantly higher in the 0- to 4-year-old group (p = 0.01). In contrast, posttraining CNR values (median, 1.05; IQR, 0.90–1.26) were significantly lower than pretraining levels (median, 1.68; IQR 1.35–1.96) in the 5- to 9-year-old group (p < 0.01) and were not significantly different in the other age groups.
Table 3 shows visual image quality scores by the three readers. After training, image quality was comparable for all items except noise (which was rated as significantly higher by all readers; p < 0.05), and one reader (reader 1) rated posttraining gray-white differentiation more frequently as average (36.7% compared with 13.3% before training) than as better than average (60% vs 66.7%) or excellent (3.3% vs 20%) (p = 0.01). However, image quality after training was never classified as suboptimal or unacceptable for any item. Interrater agreement was strong to almost perfect for all items in both the pretraining and posttraining evaluations (Table 4).
| Discussion | |
|---|
The aim of this work was to evaluate the role of radiologic staff training in reducing the radiation dose delivered to pediatric patients during head CT examinations without compromising diagnostic accuracy. In general, two strategies may be pursued to minimize children's exposure to ionizing radiation during CT. The first is limiting pediatric CT requests to only reasonable indications based on approved criteria or guidelines. The other strategy is to adjust technical parameters on CT scanners to lower radiation dose while retaining diagnostic image quality. We focused on the latter issue, and our findings indicate that staff training can be effective in improving CT protocols, at least in facilities (such as ours) without specific pediatric expertise. This was confirmed by our finding that higher CTDI and DLP values (both median and IQR) in younger children (0–4 years old) before training were significantly reduced after training, as were dose values in all other age groups. Of note, on some occasions, pretraining dose levels exceeded the maximum dose reference levels recommended by international guidelines for radiologic protection [24–26], which never occurred after training.
The aforementioned reduction in radiation dose was associated with a significant reduction in tube voltage and the tube current–rotation time settings used for CT examinations. After training the use of a higher tube voltage setting (140 kV) was significantly less frequent than before training, whereas lower tube voltage values (120 and 100 kV) were significantly more common. Lowering tube voltage settings cuts radiation dose more drastically than reducing tube current–time settings alone, because absorbed dose varies by approximately the square of tube voltage and only linearly with tube current. To put things into perspective, a reduction from 140 kV to 120 kV leads to a dose reduction of approximately 30%, and a further 30% dose reduction can be achieved by switching from 120 to 100 kV [27]. Moreover, lowering the tube voltage settings increases contrast resolution owing to the higher attenuation of lower-energy x-rays produced by a lower tube voltage, x-ray absorption occurring predominantly by photoelectric effect at lower x-ray energies. The latter effect is more pronounced with high-atomic-number objects, such as iodine from iodinated contrast material in CT angiographic studies. The increased x-ray attenuation, however, may yield overall higher tissue contrast than achieved with higher tube voltage settings, potentially improving image quality along with lower radiation dose [28–31]. In addition, the prevalence of the 140-kV setting (and of high tube current–time values) before training may reveal the trend that previously untrained staff extend adult CT protocols to pediatric patients or attempt to minimize image noise and potential artifacts by increasing tube voltage and current-time product, especially in younger children, despite the availability of consolidated evidence in the literature [32, 33] and of pediatric head CT protocols that are usually preinstalled on modern CT equipment.
On the other hand, the use of lower tube voltage and tube current–time settings has the disadvantage of greater image noise, which must be taken into account, especially in the case of unenhanced CT examinations, in which identification of subtle differences in tissue attenuation may be hampered by excessively high noise levels. Both visual and quantitative analysis of image quality revealed greater noise in target brain areas with the lower-dose head CT protocols used after training, which, however, did not result in a perceived loss of diagnostic information for any of the items evaluated. Furthermore, radiation dose levels in CT studies often exceed those required for diagnosis [34], and our finding of comparable or even higher gray matter conspicuity after training may be partially due to the higher intrinsic tissue contrast related to the more frequent use of lower tube voltage settings, which should partly offset the increased noise [25–28].
A major cause of discrepancy among CT protocols is poor awareness of radiation protection issues and lack of training on dose optimization by some technologists and radiologists (especially those who do not regularly perform pediatric imaging), as shown in several studies [35–37]. Other factors underlying this scenario are the lack of systematic auditing of dose levels and image quality according to individual patient and scanner characteristics and the lack of constructive feedback with CT application specialists, regular institutional staff education programs on dose optimization, and an effective institutional quality assurance program aimed at ensuring that diagnostic quality is achieved without exceeding dose reference levels. This may partially explain our finding of the marked dose variability (expressed as IQR) of pretraining CTDI and DLP values, especially in the youngest age group (0–4 years old), likely revealing the diversity of head CT protocols among staff members working at radiologic centers within the same institution. To this purpose, it has been found that specific training of radiologic staff may result in a significant reduction in radiation dose and overall harmonization of body CT protocols [22]. The critical importance of tailoring pediatric CT acquisition parameters to individual patients’ ages and clinical queries was emphasized by Paterson et al. [38] and Donnelly [39]. In our experience several technical improvements were introduced after training, such as individualization of tube voltage and tube current–rotation time settings depending on patient age and diagnostic query, use of automated tube current modulation, correct patient centering on the CT table, and proper choice of scout image length. These factors have proved effective in lowering radiation dose while preserving diagnostic image quality [40–42]. An International Atomic Energy Agency pediatric project [43] has revealed large variations in radiation dose levels by a factor of 55, emphasizing the need for a deeper survey of practice in pediatric CT to investigate both CT utilization in children and the level of optimization of clinical protocols. This issue is even more relevant in the case of serial CT examinations, such as in the care of trauma and cancer patients, aimed at monitoring a patient's condition [44].
Our study had limitations. First, we did not stratify pretraining and posttraining CT examinations by technologists and radiologists, which prevented us from finding a correlation between radiation dose levels and individual members of the radiologic staff. Second, we restricted our analysis to head CT examinations without IV administration of iodinated contrast material, to avoid any variation in CT image acquisition and contrast injection protocols related to different diagnostic queries, which might have resulted in differences in radiation doses delivered. It is likely that our findings could even be magnified in the case of contrast-enhanced head CT studies, owing to the higher brain tissue CNR related to the iodine content of contrast material, which can be boosted even further at lower tube voltage settings [28–31]. Third, all CT examinations were performed on scanners without iterative reconstruction algorithms, which may partly explain the higher image noise in posttraining lower-dose CT studies. However, such algorithms have been found to allow substantial reduction in radiation dose while preserving image quality compared with conventional filtered back projection [45], so our findings would likely be corroborated with the use of CT scanners with iterative reconstruction techniques.
| Conclusion | |
|---|
The radiation dose delivered to pediatric patients for unenhanced head CT examinations varies greatly within diagnostic centers at the same institution. Substantial dose reduction can be achieved through intensive hands-on training of the entire radiology staff while maintaining diagnostic image quality. Given the higher sensitivity of children to biological damage by ionizing radiation and the continuing evolution of MDCT technology and dose-saving tools, dedicated training can be crucial in reducing and harmonizing radiation dose according to the as low as reasonably achievable principle.
| Acknowledgments | |
|---|
We thank Maria Chiara Michelassi, Guido Lazzarotti, and Ilaria Pesaresi for their support in manuscript preparation.
Supported by the SUIT-Heart (Stop Useless Imaging Testing in Heart Disease) grant from ITT–Istituto Toscano Tumori of the Tuscany Region, Italy.
| References | |
|---|
Unique customized medical search engine service from ARRS! ARRS GoldMiner® is a keyword- and concept-driven search engine that provides instant access to radiologic images published in peer-reviewed journals. For more information, visit goldminer.arrs.org.
