Introduction

Computed tomography is considered the gold standard to display structural lung disease in children with cystic fibrosis (CF) and is more sensitive than pulmonary function tests (PFTs) and chest radiography in monitoring disease progression [1,2,3,4]. With the introduction of CF transmembrane conductance regulator modulator therapies, the new standard of care in cystic fibrosis is prevention of disease progression. Imaging strategies should therefore now include the most sensitive tools to monitor lung changes [5]. CT is noninvasive, can be performed in all ages, and serves to complement functional data including forced expiratory volume in 1 s and the lung clearance index [5,6,7,8].

The need for routine CT imaging in children with CF necessitates optimizing low-dose CT protocols while preserving adequate spatial resolution and image contrast to maintain diagnostic accuracy. Much work has been done in the last decade to implement low-dose CT protocols to image children with CF, who are often imaged repeatedly from a very young age [9,10,11,12]. The relatively lower dose [13,14,15,16,17,18] and higher spatial resolution [15, 16, 18,19,20,21] of photon-counting detector CT (PCD-CT) compared to conventional energy integrating detector (EID) CT is particularly suited to imaging this patient population [21,22,23].

Photon-counting detector CT fundamentally differs from conventional EID-CT at the detector level. Conventional CT uses scintillators to convert X-rays to visible light. Reflective septae are embedded in the detectors to decrease inter-pixel scatter [21, 24] (Fig. 1). Photon-counting detectors use semiconductors to convert X-rays into electron cloud pairs, which move through a semiconductor material across a high voltage to be converted directly to signal [23, 24] (Fig. 2). The improved spatial resolution of PCD-CT compared to conventional CT is based on several resulting features: (1) Given that no septae are required between pixels in photon-counting detectors, a smaller detector pixel size can be achieved (0.18 × 0.15 mm for the NAEOTOM Alpha; Siemens Healthineers, Malvern, PA) at a higher dose efficiency [21]. The maximum spatial resolution is as high as 40 line-pairs/cm, [21] compared to up to 28 line-pairs/cm in EID-CT scanners [25]. (2) There is less optical crosstalk between detectors in PCD-CT because X-rays are not converted to visible light. (3) The direct conversion of X-rays to electronic signal in PCD-CT allows for the selective exclusion of photons below 20–25 keV, which contribute to electronic noise but not to image signal. (4) The direct conversion of X-rays to output signal in PCD-CT also allows for equal weighting of lower-energy photons, allowing for improved image contrast [21, 23, 26, 27].

Fig. 1
figure 1

Schematic of a conventional energy-integrated detector (EID) CT. A scintillator converts X-rays to visible light. The light signal is received at the photodiode, which is partitioned by reflective septae to prevent optical crosstalk between detector elements. Septae reduce the geometric dose efficiency of conventional CT

Full size image
Fig. 2
figure 2

Schematic illustrates the detector mechanism in photon-counting detector CT. X-rays are converted to an electron cloud that moves across a high voltage within the semiconductor. There are no septae in the detector, which allows for reduced pixel size. This translates to higher spatial resolution in CT images and greater radiation-dose efficiency of the detector

Full size image

Many of the detector features also reduce image noise and allow for tailored protocols that prioritize radiation dose reduction. Several studies have shown significant radiation dose reduction with PCD-CT compared to conventional CT in chest applications [28,29,30]. Symons et al. [28] showed lower image noise, better diagnostic quality and improved lung nodule contrast-to-noise ratio in PCD-CT over conventional CT of the chest. Jungblut et al. [29] showed that a 66% radiation dose reduction could be achieved without sacrificing image quality or diagnostic performance in the assessment of adult interstitial lung disease. No prior study applied PCD-CT to the imaging of pediatric patients with cystic fibrosis.

Imaging findings of CF relevant to staging the disease include elements of the Brody II scoring system: bronchiectasis, bronchial wall thickness, parenchymal opacities, air trapping and mucus plugging [5, 31]. The Brody II scoring system has been shown to correlate with PFTs, frequency of exacerbations and quality of life [6, 32, 33]. In this pilot study, we sought to determine the ability of a reduced-dose PCD chest CT protocol to reproducibly display pulmonary disease severity using the Brody II score for pediatric patients with CT scanned at radiation doses similar to those from a two-view chest radiograph.

Materials and methods

Our institutional review board approved this retrospective study. The study included pediatric patients with CF undergoing clinically indicated chest CT using our institution’s reduced-dose PCD chest CT protocol, which included inspiratory and expiratory series. Waiver of informed consent was granted for this minimal risk retrospective study. We only included children who gave general consent to use medical records in research. No children were taking intravenous antibiotics at the time of imaging.

Computed tomography technique

Contiguous inspiratory and expiratory images from the lung apices to the diaphragm were obtained in each child with CF who underwent clinically indicated non-contrast reduced-dose chest PCD-CT. Images were obtained without sedation or general anesthesia. The PCD-CT system was operated in high-resolution mode, with a detector configuration of 120 × 0.2 mm. The tube potential was 100 kV with an added tin filter, rotation time was 0.25 s and helical pitch was 1.0. Axial images were reconstructed with a Qr56 reconstruction kernel, level 3 iterative reconstruction (IR) strength, 0.8-mm slice thickness, 0.5-mm slice increment and 1,024 × 1,024 matrix. A second image series was reconstructed with a Br44 kernel, level 3 IR strength, 3-mm slice thickness, 3-mm slice increment and 512 × 512 matrix.

Image review

Three pediatric radiologists with Certificates of Added Qualification (two with 10 years, K.K.H. and P.G.T., and one with 5 years of experience, N.C.H.) independently scored each scan. They assessed diagnostic image quality using ordinal rating scales, with each reader instructed to examine the entire volumetric dataset before providing any image-quality rating. Readers were further instructed to subjectively grade the spatial resolution (distal vessel and airway conspicuity) on an ordinal scale (1: unacceptable, 2: suboptimal, 3: acceptable); noise (1: cannot evaluate, 2: readable with moderate artifact, 3: readable with mild artifact, 4: no artifact); soft tissues of the mediastinum and chest wall (1: cannot evaluate, 2: readable with moderate artifact, 3: readable with mild artifact, 4: no artifact). Image-quality metrics are reported using descriptive statistics. The readers rated presence of bronchiectasis, peribronchial thickening, parenchymal opacity, air trapping and mucus plugging in each lobe of the lung using the Brody II scoring system to categorize CF severity [1]. This was done in a similar fashion to what was performed by Loeve et al. [11], who evaluated reduced-dose non-contrast conventional chest CT images in pediatric patients with CF. In short, the Brody score rates the severity of bronchiectasis and parenchymal opacification, and the extent (central vs. peripheral lung) of bronchiectasis, mucus plugging, peribronchial thickening and hyperinflation to reflect the severity of CF within the lung [1].

Statistical analysis

Image-quality metrics of spatial resolution, noise and soft-tissue evaluation are reported with mean and standard deviation. We estimated inter-rater agreement for ordinal-scaled Brody II ratings with Krippendorff alpha using the ordinal difference function, and calculated 95% confidence intervals (CIs) using percentile bootstrap intervals with 1,000 bootstrap samples, with cluster bootstrap used to account for intra-patient correlation clustering for measures involving multiple anatomical locations read per patient. We interpreted point estimates using the recommendations of Landis and Koch [34] for kappa statistics. We estimated inter-rater agreement of the Brody II scores using the intraclass correlation coefficient (ICC) along with 95% CI. Values below 0.5 indicate poor reliability, between 0.5 and 0.75 moderate reliability, between 0.75 and 0.9 good reliability, and above 0.9 excellent reliability [35].

We calculated the mean, standard deviation and range (minimum and maximum) of radiation dose of each scan in terms of volumetric CT dose index (CTDIvol) and dose–length product (DLP). The size-specific dose estimate (SSDE) was also calculated based on the American Association of Physicists in Medicine (AAPM) task group reports 204 and 220 [36, 37]. We estimated the effective dose by multiplying the DLP by an age-specific conversion factor for the chest [38].

Results

Sixteen children were scheduled to undergo clinically indicated reduced-dose chest PCD-CT for routine clinical imaging follow-up of CF. One child could not comply with instructions during the CT examination and was excluded from the study, reducing the total number to 15 children. Four children could not breath-hold and underwent single-phase imaging. The mean age was 8.9 years (range, 2.5–17.5 years; 4 girls). The mean CTDIvol was 0.07 ± 0.03 mGy (minimum: 0.03 mGy, maximum 0.13 mGy) with the DLP being 1.8 ± 0.9 mGy∙cm (minimum: 0.64 mGy∙cm, maximum: 3.5 mGy∙cm). SSDE was 0.11 ± 0.03 mGy (minimum: 0.07 mGy, maximum: 0.18 mGy) with the mean effective dose being 0.12 ± 0.04 mSv for the inspiratory and expiratory scans combined for those with two-phase imaging, a dose range that approximates that of radiographs [39,40,41,42,43].

Spatial resolution (distal vessel and airway conspicuity) was rated on the aforementioned 3-point scale as “acceptable” by all three readers on 13/15 scans, with two readers assessing the same scan as “suboptimal” and one reader assessing a second scan as “suboptimal” (mean, 2.93 ± 0.25). None of the scans was rated “unacceptable.” All three readers assessed noise on a 4-point scale as readable with “mild” or “moderate artifact” (mean, 2.36 ± 0.56). The soft tissues were less well visualized than pulmonary structures (mean, 1.51 ± 0.78).

All three readers also scored the CT scans according to the Brody II criteria. Table 1 shows agreement among readers for each component of the Brody II score. The average Brody II score was 12.5 (range, 4–19), with moderate inter-reader reliability and ICC of 0.61 (95% CI=0.27, 0.84). Interrater reliability in terms of the Krippendorff alpha was moderate for bronchiectasis (0.52), peribronchial thickening (0.55), presence of opacity (0.62) and air trapping (0.70). Interrater reliability was poor for mucus plugging (0.09).

Table 1 Inter-rater reliability (IRR) of three readers in Brody II score assessment
Full size table

Figure 3 shows representative PCD-CT images from a child with CF to demonstrate the structural lung changes assessed with the Brody II criteria and to visually illustrate that these features are interpretable at radiation doses comparable to radiographs.

Fig. 3
figure 3

Inspiratory photon-counting detector (PCD)-CT images in a 6-year-old girl with cystic fibrosis show structural lung features assessed in Brody II scoring. a, b Axial inspiratory image (a) illustrates varicose bronchiectasis (arrow) in the right middle lobe. Coronal expiratory image (b) shows air trapping (arrow) in the left lower lobe. c, d Sagittal images show mucus plugging within a dilated bronchus (arrow in c), ground-glass opacity (arrow in d) and peribronchial cuffing (arrowhead in c, d). Images are displayed in standard lung windows (width: 1,500; level: –500). The axial image is 0.8 mm and coronal and sagittal images are 3 mm

Full size image

Figure 4 shows a side-by-side comparison of a representative PCD-CT image and an image from a conventional EID-CT obtained 16 months prior in the same girl (the same girl as shown in Fig. 3). The appearance of pulmonary structures in lung windows at 3-mm slice thickness is comparable, but the PCD-CT images were obtained at a 95% reduction in radiation dose. Figure 4 also illustrates representative axial images from the same examinations, in soft-tissue windows at similar anatomical locations at 3-mm slice thickness. There is much greater noise in the PCD-CT images. The results of our reader study attest to the limited interpretability of the soft tissues at this dose.

Fig. 4
figure 4

Photon-counting detector (PCD)-CT images in the same 6-year-old girl with cystic fibrosis as in Fig. 3. a–d Lung window images show comparable image quality, despite a 95% radiation dose reduction of the PCD-CT images (a, sagittal) from that of the conventional energy integrating detector (EID)-CT images (b, sagittal. The PCD-CT volumetric CT dose index (CTDIvol) was 0.05 mGy and the effective dose was 0.09 mSv. The EID-CT CTDIvol was 1.89 mGy and the effective dose was 2.2 mSv. Noise was measured in the paraspinous tissues on soft-tissue windows, at similar slices in the same child, and measured 40 Hounsfield units (HUs) on the PCD-CT image (c) compared to 9 HU on conventional EID-CT image (d). Comparison images were reconstructed with the same 3-mm slice thickness and similar reconstruction kernels

Full size image

Figure 5 illustrates the advantages of high temporal resolution scanning utilizing two X-ray tubes in reducing patient motion. Four of the 15 children could not follow breathing commands, including the 4-year-old girl in Fig. 5. Diagnostic images were obtained despite an imperfect breath-hold.

Fig. 5
figure 5

Photon-counting detector (PCD)-CT images in a 4-year-old girl with cystic fibrosis. a, b Axial (a) and coronal (b) images at 0.8-mm slice thickness demonstrate tree-in-bud opacities in the right lower lobe (arrows in a) as well as peribronchial thickening (arrowheads in b). Some breathing motion is seen at the diaphragms (arrows in b), and these images were obtained without sedation or general anesthesia but remained diagnostic for structural lung changes of cystic fibrosis (volumetric CT dose index of 0.03 mGy; effective dose of 0.03 mSv). Readers rated this study as having “acceptable” spatial resolution, with noise levels that were “readable with moderate artifact”

Full size image

Discussion

This pilot multi-reader study demonstrates that reduced-dose PCD-CT images obtained using high-spatial-resolution imaging in pediatric patients with CF can generate diagnostic-quality images that permit reproducible interpretation using the Brody II criteria at radiation doses that approach those of two-view chest radiographs. Using this technique, reduced-dose PCD images can be obtained in inspiration and expiration and reliably permit direct visualization and identification of bronchiectasis, peribronchial thickening, airspace opacity and air trapping. Although the absolute risks of radiation related to medical imaging in people with CF are challenging to estimate [44], the technical ability to perform CT at doses this low has significant implications for population-based imaging recommendations in people with CF, affecting up to 160,000 people worldwide [45], just under half of whom are children [46].

Imaging guidelines from the Cystic Fibrosis Foundation recommend a baseline chest radiograph within the first 3–6 months of age, once again before age 2, and every other year after age 2 [47, 48]. In pre-school-age children (ages 2–5), radiographs can be replaced with CT and be performed every 2–3 years [47, 48]. The guidelines cite sedation and radiation as the issues that combine to make the use of chest CT problematic [48]. The European Imaging Management of Cystic Fibrosis (MAESTRO) consortium cites biennial CT at radiation doses as low as reasonably achievable as the current best clinical practice [49]. PCD-CT is a new technology that is not widely available, making it difficult to be considered as the standard of care in imaging this population. However, our study does present evidence that PCD-CT can be used to stage CF with an established CT scoring system, without sedation, and at radiation doses that approach those of radiography. Subsequent to our study, technical improvements have permitted high-pitch in addition to dual-source image acquisition, which further reduces motion artifacts, particularly for pediatric patients.

All three readers assessed the spatial resolution, specifically for distal vessels and airway conspicuity, with a mean score of 2.93 ± 0.25 on a 3-point scale, with no reader assessing any scan as “unacceptable.” All three readers assessed noise levels as readable with “mild” or “moderate artifact.” The scans were acceptable for assessing airways and lung parenchyma at these radiation doses; however, our results indicate that the evaluation of soft-tissue structures is limited with this technique.

In our study, the CTDIvol was 0.07 ± 0.03 mGy (range, 0.03–0.13 mGy) for each scan and the total effective dose was 0.12 ± 0.04 mSv for the inspiratory and expiratory scans combined. The effective dose of a standard chest radiograph depends on patient size and the number of views. Dose estimates of a single-view anteroposterior (AP) or posteroanterior (PA) chest radiograph span an estimated range of 0.007–0.014 mSv from newborns to adults, with adult lateral radiographs ranging 0.031–0.038 mSv [39]. Dose estimates for two-view pediatric chest radiographs vary widely in the literature up to 0.2 mSv, depending both on technique and methods used to calculate the effective dose [39,40,41,42].

Other studies have used reduced-dose CT in people with CF, a few of which have reported effective doses below 1 mSv [50, 51]. The radiation dose estimates achieved in our study are comparable to the effective dose of 0.04 mSv (95% confidence interval [CI]: 0.03 mSv, 0.10 mSv) published by Ernst et al. [50] and 0.08 ± 0.01 mSv published by Moloney et al. [51], who used volumetric CT protocols performed at 80 kV with model-based iterative reconstruction to help minimize radiation dose. It would be useful to compare images from these studies directly to images obtained with a similar protocol on PCD CT.

Imaging strategies to reduce radiation in patients with CF have often included CT with intermittent slices. O’Connor et al. [10] compared 6-slice low-dose chest CT protocols at 1-mm and 0.5-mm slice-thickness in 14 pediatric patients and achieved effective dose ranges between 0.19 ± 0.04 mSv and 0.14 ± 0.03 mSv, respectively [10]. This is an order of magnitude lower than the dose reported in pediatric patients by Loeve et al. [11], who estimated a mean effective dose of 0.69 mSv for end-inspiratory and 0.35 mSv for end-expiratory volumetric CT, a CT protocol more comparable to our own [11].

We used the Brody II scoring system to evaluate the PCD-CT images because it represents a validated method that is applied to conventional CT images [1, 2, 31]. We report the same Brody II categories as Loeve et al. [11], with an average score of 12.5 (range, 4–19). This average score illustrates that our study consisted of children with relatively mild structural lung changes on imaging. This is not surprising given the very young age of our patient cohort, with 5 of the 15 children being age 6 years or younger. While we demonstrated moderate and acceptable reliability for the overall Brody score (ICC 0.61) and moderate to substantial agreement in all components except for mucus plugging, the interobserver agreement for mucus plugging was poor. These findings might be the result of our small population with mild and subtle disease. Other authors have examined larger cohorts with older groups of patients who had more severe manifestations of disease [52]. The purpose of our study, however, was not to validate the scoring system but to demonstrate reproducibility of the standardized scoring system using reduced-dose PCD images at a radiation dose similar to that in two-view radiographs.

Our study has multiple limitations, principally the retrospective design and the small number of children in this convenience sample. Additionally, there is no comparison to prior or concurrent imaging studies such as chest radiograph or reduced-dose EID-CT to illustrate differences using the reduced-dose PCD-CT protocol examined in this work. Further work is needed to delineate potential differences in the utility of various imaging modalities for displaying Brody II or similar imaging findings, as well as to determine whether the additional imaging findings seen on PCD-CT might change the clinical management by referring pulmonologists and improve patient outcomes.

The results of our study help illustrate how PCD-CT offers a greater degree of latitude in adjusting imaging protocols to optimize the evaluation of structural lung changes in children with CF. There are no universally agreed upon standardized CT protocols for CF [12, 49]. Intermittent slice-sampling has long been a means of radiation-dose reduction for this patient population [5, 53, 54], but structural lung changes in CF are often heterogeneous and the introduction of cystic fibrosis transmembrane conductance regulator (CFTR) modulator therapy highlights the need to track subtle changes. The ability to image whole-lung anatomy at doses comparable to those in radiographs presents new opportunities to study milder disease in younger patients. Further work is needed to determine the proper trade-offs among radiation-dose reduction, spatial and contrast resolution, and intermittent slice vs. whole lung imaging in this patient cohort.

Conclusion

This pilot study demonstrated the feasibility of performing volumetric CT in both inspiration and expiration at radiation doses that approximate radiographs to assess structural lung changes in children with cystic fibrosis. This technical advance provides the ability to image the entire lung volume at two respiratory phases, yet also balance radiation risk and image quality.