Introduction
Technological developments in computed tomography (CT) enhanced the clinical imaging possibilities in pediatric patients, sparking off a growth in the number of CT scans performed within this population [
1,
2]. Over the years, considerable efforts have been made to optimize radiation dose and image quality (IQ) [
3,
4]. Several techniques are used to optimize pediatric CT scanning protocols such as automated tube current and tube voltage adaptation, as well as the use of iterative reconstruction techniques [
5‐
8]. For an ideal working of the automatic exposure control (AEC) and to achieve ideal IQ, it is important to position the patient exactly in the center of the CT gantry [
9]. Vertical patient positioning is determined by setting the table height. Ideal positioning is defined as the table height at which the patient’s and scanner’s isocenter coincide. Patient positioning lower or higher than the scanner isocenter (i.e., table set too low or too high) affects the patient’s shape and size on a CT scan localizer radiograph, which is of direct effect on the behavior of the AEC. Positioning of pediatric patients is quite challenging, because of the wide variation in body proportions. Furthermore, when they have to be positioned in fixation aids such as a baby cradle or vacuum cushion due to lack of cooperation, it is more difficult to estimate the center of the patient. Recent studies have exhibited the benefits of using a 3D camera and a body contour detection algorithm for the accurate positioning of adult patients, resulting in smaller deviations from the ideal table height compared with manual positioning done by radiographers [
10,
11]. The camera algorithm is described in detail in our paper with regard to (adult) patient positioning [
11]. It was not yet applicable to pediatric patients due to their different body proportions compared with adults [
10,
11]. The algorithm was improved to account for the pose and body proportions of pediatric patients, too. The aim of this study was to determine the accuracy of the new improved version of the algorithm in the positioning of pediatric patients in comparison to manual positioning done by radiographers.
Discussion
We assessed the accuracy of pediatric patient positioning with the aid of a body contour detection system (3D camera) and compared it with manual positioning by radiographers. We found that positioning with the 3D camera of pediatric patients without a fixation aid allows for more accurate patient positioning than manual positioning by radiographers. This outcome is similar to the findings in adult patients [
10,
11]. Differences in positioning accuracy between the 3D camera and radiographers were not statistically significant for patients positioned in a baby cradle or a vacuum cushion. In virtually all cases of infants placed in fixation aids, like a baby cradle (Fig.
1) or a vacuum cushion (Fig.
2), it was not possible to fit a patient Avatar due to the small body size and the large occlusions. Instead, the fallback described above was applied, where the isocenter is directly estimated as geometric mean between the depth measurements and the table. This approach introduces a deviation, because these fixation aids add a considerable layer between patient and table, which in the absence of the Avatar is wrongly attributed to the patient, leading to an overestimation of patient size. Nevertheless, positioning of patients in a fixation aid seems feasible with a 3D camera. Small performance differences between camera and radiographers could not be detected due to the limited number of patients included. However, post hoc power analysis showed that the performance difference was not larger than 8.2 mm; otherwise, this would have been noted given our sample size. Thus, the fallback routine facilitates automatic positioning of a pediatric patient while keeping possible differences with a well-trained radiographer below 10 mm on average.
However, the deviation from the ideal table height could be reduced by taking the positioning devices into account. Therefore, applying an intermediate step consisting of the detection of the presence of a fixation aid like a baby cradle and a vacuum cushion (open and closed) might be of use and may be considered for further research. After detection of such aid, a correction for the thickness of such an aid can be applied to the geometric isocenter for these specific cases. The correction can be determined upfront by estimating the mean error for the vacuum cushion and by accounting for the fairly constant thickness of the baby cradle.
The two main challenges for the algorithm are the small size of the patients that are positioned with such aids and the large degree of occlusions introduced by the aids. Given a large amount of 3D camera training images, probably we could reliably fit the patient Avatar also under these circumstances. Then, as usual for cases without a baby cradle or a vacuum cushion, it would be possible to compute the center of the patient Avatar and naturally exclude additional layers such as blankets or fixation aids. The Avatar fitting was only possible in three out of nine cases when patients were positioned in a baby cradle and the fallback had to be applied in all cases when positioned in a vacuum cushion. Therefore, further work on the development with additional training data might improve the algorithm even further to reliably obtain a patient isocenter when patients are positioned in fixation aids like the baby cradle and vacuum cushion.
The 3D camera is able to assist the radiographer in positioning of pediatric patients, especially in cases without fixation aid. It should be emphasized that radiographers will continue to play an important role in patient positioning by patient guidance and verification of the table height suggested by the 3D camera, especially when fixation aids are used.
Studies demonstrated a significant impact on radiation dose and image quality when a pediatric patient is not properly positioned in the scanner isocenter [
15,
16]. In those studies, an anthropomorphic head, thorax, and/or abdomen simulating on a 5-year-old child was used. Organ doses and noise differences with several vertical table height deviations were compared with organ dose and noise levels at the scanner isocenter/center position. A noise increase of up to 45% in chest scans relative to the center position was demonstrated for table positions in the highest (+ 54 mm) and lowest (− 60 mm) vertical positions and a breast dose increase of up to 16% with 40 mm lower vertical position [
16]. Although the absolute table height deviations in our study were not always that high, maximum deviation values were high, especially with radiographers (Tables
1 and
2). Relative breast dose increase was considered to be 7% lower with 20 mm vertical lower positioning compared with the 40 mm lower position. This vertical positions are comparable to the values of the largest deviations between the 3D camera and radiographers in our study. With the tendency to position pediatric patients more often lower than the ideal table height, the noise would increase. With less extreme deviations from the ideal table height that can be obtained with the 3D camera (Tables
1 and
2), both the radiation dose and the image quality will be more constant. The same applies for organ radiation doses and image noise in head and abdominal CT [
15]. Large vertical table height deviation was of substantial influence on radiation dose and image noise, where the impact of these deviations depends on the body region and location of individual organs within the body [
15]. However, accurate and less deviations from the ideal table height are required to consolidate image quality and radiation dose. Our results were obtained in an academic facility with highly trained radiographers. It is conceivable that both the median and maximum values of deviation from the ideal positioning would be even larger when the study was obtained in a hospital without dedicated training in pediatric CT scanning.
There are limitations to this study that require considerations. For the purpose of the analysis, the algorithm used the actually scanned range to calculate the isocenter. This differs from routine operation of the camera system, whereby the algorithm uses the scan range that is defined on the planning image (=color photograph taken by the camera) prior to obtaining the localizer radiograph and scanning the patient. Consequently, the suggested ideal table height by the 3D camera based on the planned scan range may differ from the suggested table height based on the actual scan range. Nevertheless, our results demonstrate the accuracy when a 3D camera is used properly and the selected body region on the localizer radiograph and the actual scan range are the same.
In conclusion, a 3D camera for body contour detection allows for accurate pediatric patient positioning in CT. The 3D camera is able to assist the radiographer in positioning of pediatric patients, especially in cases without fixation aid. Positioning of patients in a fixation aid is feasible with a 3D camera, but evaluation of possible improvements in positioning accuracy was limited by the small sample size.
Compliance with ethical standards
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.