Main points

• Cross-sectional study shows that ultra-high-resolution (UHR) photon-counting detector computed tomography (PCD-CT) imaging provides more detailed information regarding interstitial lung disease (ILD) in patients with rheumatoid arthritis (RA) than HR PCD-CT.

• PCD-CT might be a promising tool in the early diagnosis of subclinical ILD and bronchiectasis in patients with RA.

• Low-dose, HR PCD-CT could serve as a means of preselecting candidates for more detailed, UHR measurements in the assessment of RA-related ILD.

Contrary to conventionally utilized energy-integrating detectors (EIDs), photon-counting detectors (PCDs) are able to directly convert X-ray photons into electric pulses.¹ This process leads to better spatial resolution, decreased beam hardening, reduced noise and radiation dose, and to register the energy of photons.^2,3,4 PCD-computed tomography (PCD-CT) has gained increasing interest in pulmonary imaging due to its high spatial resolution.⁵ To date, preclinical studies have mostly investigated lung PCD-CT imaging by analyzing optimal reconstruction parameters, such as matrix size, optimal slice thickness, and iterative reconstruction algorithms.^6,7

Rheumatoid arthritis (RA) is a systemic autoimmune disease associated with several pathologies, including parenchymal and pleural involvement, bronchiolitis, rheumatoid nodules, and vascular abnormalities. Interstitial lung abnormalities (ILA) occur in 1.8%–67% of patients with RA, among whom established interstitial lung disease (ILD) accounts for a significant proportion of mortality and morbidity.^8,9 High-resolution (HR) CT remains the main imaging modality for the evaluation of lung involvement in RA.¹⁰ The spectrum of RA-related interstitial abnormalities in parenchymal involvement includes ground-glass opacity (GGO), fibroreticular changes, bronchiectasis, and honeycombing. The most frequent phenotype of RA-ILD is usual interstitial pneumonitis (UIP).¹¹ It has been demonstrated that the early identification of subclinical ILAs (which show a tendency to progress to ILD) by CT promotes early intervention that stabilizes further interstitial changes and, thus, significantly improves prognosis.¹²

The primary goal of this study is to evaluate the performance of PCD-CT in the detection of early and subclinical parenchymal lung involvement in patients with RA.

Methods

Patients

From February 2022 to November 2022, 334 patients were enrolled in this study (Figure 1). Inclusion criteria were as follows: (i) patients diagnosed with RA (according to 2010 American College of Rheumatology/European League Against Rheumatism classification criteria); and (ii) patients >40 years of age. Exclusion criteria were as follows: (i) patients with pulmonary infection, lung malignancies, or previously diagnosed ILD. The demographic data, clinical manifestations, and routine laboratory test results of study participants were recorded. All patients were checked regularly and received medical treatment as recommended by the attending rheumatologist, following standards of care. First second of forced expiration, forced vital capacity, and diffusion capacity for carbon monoxide (DLCO) were measured by pulmonologists via pulmonary function tests. Patients underwent yearly chest X-rays following their RA diagnosis; however, based on earlier radiographs, none had an HR CT-based indication to rule out ILD. Written informed consent was obtained from all study participants. Three hundred twenty-six patients underwent HR CT imaging, and, following the detection of abnormalities by the attending radiologist, 61 patients underwent a subsequent ultra-high resolution (UHR) CT on the same day for further evaluation of lung parenchyma (Figure 1).

Research ethics

This trial was registered on the clinicaltrials.gov website (IV-2683-1/2022/EKU) and approved by the local ethical review board (2021, National Scientific and Research Ethics Committee, Hungary). This work was carried out in accordance with the Helsinki Declaration (JAMA 2000; 284:3043–3049).¹³

Patient PCD-CT measurements

HR (slice thickness: 0.4 mm) and UHR (slice thickness: 0.2 mm) CT scans were carried out with a PCD-CT scanner (Naeotom Alpha^®, Siemens Healthineers, Erlangen, Germany). Both imaging techniques were performed with a large field of view (FOV) [median (quartile 1, quartile 3): 35 (32, 38) cm] and 1024 × 1024 matrix. Additionally, quantitative iterative reconstruction algorithms were utilized to enhance image quality (Table 2). To exclude GGO from dependent atelectasis, prone inspiratory HR CT measurements were performed.

Phantom studies

To compare the image quality of PCD- and EID-CT methods, prone chest region measurements with matched parameters (similar slice thicknesses and equivalent rotation times, voltage and current, pitch values, and scan lengths) were taken of a phantom (CT Whole Body Phantom, PBU-60) with a 1-year-old PCD-CT scanner (Naeotom Alpha^®, Siemens Healthineers, Erlangen, Germany), a 2.5-year-old 128-slice EID-CT scanner (Philips Incisive^®, Philips, Amsterdam, The Netherlands), and a 2-year-old 128-slice EID-CT scanner (GE Revolution EVO^®, GE Healthcare, Chicago, Illinois, USA). Subjective image quality was rated independently by four radiologists on a five-point scale (5 being best, 1 being worst). Contrast-to-noise ratio (CNR) was calculated based on the formula: [average pixel values in signal region of interest (ROI) (bronchial wall) – average pixel values of background ROI (air)] / standard deviation (SD) of background ROI (air). Signal-to-noise ratio (SNR) was calculated based on the formula: average pixel values in signal ROI (bronchial wall)/SD of background ROI (air).

Dose values

Volume CT dose index (CTDIvol) and total dose length product (TDLP) values were extracted from patient protocol data (syngo.via software, Siemens Healthineers). Approximate effective dose values were calculated from TDLP values as follows: Effective dose = TDLP × k-factor (0.014 mSv/mGy*cm).^14,15

Evaluation of parenchymal abnormalities

All PCD-CT images were reviewed, and specific ILD patterns were determined by consensus, by two radiologists with 6 and 13 years of experience. A third thoracic radiologist with 13 years of experience then reviewed the images and spoke with the ILD board to reach an agreement with the pulmonologists. Interstitial abnormalities were classified into four categories: GGO (parenchymal opacity with perceptible underlying bronchovascular structure without architectural distortion), reticulation (thickening of interlobular septae or intralobular septae and traction), bronchiectasis (dilatation of bronchial tree), and honeycombing (clustered, subpleural, multilayered, cystic air-spaces).

To test whether the UHR protocol provided additional information about interstitial pathologies, a semiquantitative scoring system was utilized. The extent of pulmonary parenchymal abnormalities for each lobe was scored using a Likert-type scale (0 = absent; 1 = 1%–25%; 2 = 26%–50%; 3 = 51%–75%; 4 = 76%–100%). Total GGO, reticulation, bronchiectasis, and honeycombing scores were calculated by adding up the scores of all five lung lobes, with final values ranging from 0–20. All scores were combined to produce a total ILD HR CT score ranging from 0–80 (Figure 2).^16,17

As ILD is a heterogeneous group of parenchymal lung disorders, observed abnormalities were classified into patterns defined in Table 3.

Statistical analysis

Distribution was defined by the Kolmogorov–Smirnov test. Descriptive statistics (median and quartiles) and mean ± SD were used to represent abnormally and normally distributed variables, respectively. A paired t-test was used to compare normally distributed data, and the Wilcoxon test was used to compare non-parametric data. In the case of parenchymal changes with a definitive ILD pattern, Cohen’s kappa (κ) was used to test agreement between readers (0–0.20 = poor agreement; 0.21–0.40 = fair agreement; 0.41–0.60 = moderate agreement; 0.61–0.80 = substantial agreement; and 0.81–1.00 = almost perfect agreement).¹⁸ The Pearson correlation coefficient was used to find correlations between total lung HR CT scores and pulmonary function tests. Differences between the image quality parameters of different CTs were evaluated using a One-Way ANOVA test with a subsequent Tukey post-hoc analysis. Significance was established at P values <0.05 were considered statistically significant. Categorical variables were reported as frequencies and percentages. Statistical analyses were performed using GraphPad Prism v. 6.0 software.

Results

Demographic and clinical characteristics

Of the 334 patients initially enrolled, eight were excluded due to the presence of an acute lung infection, and 265 were excluded due to the absence of significant interstitial changes in the initial HR CT scan (total ILD score <2) (detailed descriptions in the section entitled: assessment of parenchymal abnormalities). Sixty-one patients underwent both HR and UHR scans, and 51 patients underwent pulmonary function tests (Table 1). The mean age of study participants was 68.6 ± 9.73 years, and 40 (65.57%) were female. Average time since disease onset was 15.75 ± 12.85 years. Forty-two (68.85%) patients were seropositive, and 33 (53.22%) had previously been smokers. The detailed characteristics of the study population are summarized in Table 1.

Assessment of parenchymal abnormalities

Parenchymal abnormalities were visually recognizable on both HR and UHR scans (Figure 3). UHR CTs yielded higher total ILD scores than HR CTs [6 (4, 9) vs. 4 (2.5, 8), P < 0.001). Additionally, bronchiectasis and reticulation scores were significantly higher in the UHR protocol compared with the HR protocol [2 (0, 2) vs. 0 (0, 2), P < 0.001 and 2 (0, 3.5) vs. 0 (0, 3) P < 0.001, respectively]; however, GGO and honeycombing scores did not differ [2 (2, 4) vs. 2 (2, 4), P = 0.944 and 0 (0, 0) vs. 0 (0, 0), P = 0.641, respectively] (Figure 4). Visually identified patterns did not differ significantly between UHR and HR PCD-CT protocols (Figure 5). UIP patterns, non-specific interstitial pneumonia, desquamative interstitial pneumonia, respiratory bronchiolitis-ILD, pleuroparenchymal fibroelastosis, and post-coronavirus disease-2019 parenchymal changes were registered in the examined population on both HR and UHR protocols (Figure 5). No patterns of lymphocytic interstitial pneumonitis or organizing pneumonia were identified on the scans. Inter-reader reliability pattern scores varied between moderate and perfect agreement (Figure 5). Most cases were small-extent, otherwise non-specified parenchymal abnormalities. Total ILD scores of both HR and UHR protocols showed a mild but significant negative correlation with DLCO values (HR: r = –0.297, P = 0.034; UHR: r = –0.294; P = 0.036) (Figure 6).

Dose considerations and phantom studies

Reduced-dose, 0.4 mm scans had significantly lower CTDIvol values [median (quartile 1, quartile 3): 0.67 (0.69, 1.06) mGy] compared with non-reduced, 0.2 mm scans [8.18 (6.80, 9.23) mGy, P < 0.001]. The 0.4 mm slice thickness HR acquisitions had approximately 8.6 × lower TDLP [29.0 (24.48, 33.20) mGy*cm] compared with 0.2 mm slice thickness, non-reduced-dose UHR scans [250 (218, 305) mGy*cm, P < 0.001]. Median effective radiation doses were ~0.4 mSv for  low-dose (LD) HR CT scans and 3 mSv for UHR CT scans (Table 2). Dose-matched phantom studies confirmed that, compared with EID-CT scans, PCD-CT measurements had improved subjective and objective image quality values (Figure 7).

Discussion

A relatively small number of studies on the clinical application of PCD-CT in lung diseases have been published. This current work extends previous observations. This study demonstrates that LD PCD-CT chest scans could be used to evaluate the quality and extent of ILA in a majority of patients with RA, and higher-accuracy UHR imaging can add further information about lung parenchymal involvement. Thus, HR PCD-CT with a low effective radiation dose may serve as a valuable screening tool in the selection of RA-ILD patients for a more detailed, higher-dose UHR PCD-CT screening.

RA is a systemic autoimmune disease, and lung involvement may be its most frequent extra-articular manifestation and highest contributor to a worsening prognosis.¹⁹ The prevalence of interstitial lung involvement is reported in a wide spectrum of patients with RA, and ILD can be a predictor of the development of articular manifestations.^9,20,21 Some forms of ILD are progressive, and, in addition to their patterns, the ILD board considers the extent of lung involvement an important parameter in its multidisciplinary discussion.^22,23 Screening for ILD may be advisable in select cases of RA, as early detection of parenchymal changes could help direct antirheumatic treatment.²⁴

The identification of interstitial lung involvement requires high spatial resolution scans, as the subtlety of parenchymal changes (e.g., intralobular reticulations, bronchiectasis, and honeycombing) are frequently indefinite.²⁵ To date, there is no worldwide consensus on screening recommendations because the benefits of lung parenchymal involvement screening have had to be balanced with the inherent risks of ionizing radiation. Large FOV chest PCD-CT scans with 1024 × 1024 matrix sizes conferred better overall image quality and SNR than standard EID-CT scans.^6,26 According to previous studies, LD image acquisition with PCD-CT showed better SNR and attenuation accuracy compared with conventional CT, especially at lower doses, where attenuation decreased significantly with EID-CT.⁷ Better image quality was also observed, especially in areas with known beam hardening (e.g., paravertebral spaces). Prior investigations have shown that PCD-CT images have 15.2%–16.8% less noise at two different dose levels. Furthermore, studies have proven that HR parameters could be preserved while applying LD protocols in lung evaluation.²⁶ In this examination, fast gantry rotation times (0.25–0.5 s) were used to reduce scanning time and motion artifacts.²⁷ We used a 1024 × 1024 matrix, large FOV (35 ± 3 cm, depending on the size of the patient), and 0.2 or 0.4 mm slice thickness parameter protocols and found a satisfactory detection of parenchymal pathologies, including GGOs, fibrotic reticulations, bronchiectasis, and subpleural cysts (Table 2, Figure 4). According to the literature, lower tube currents are optimal for pulmonary nodule detection (approximately 25 mA).²⁸ However, for subtle parenchymal anomalies, higher currents are inevitable to reach better resolutions. In the case of this study, 100 kV (for HR) and 120 kV (for UHR) voltages, as well as automated mA parameters, were utilized to obtain a HR image with reduced dose values (Table 2).

Optimized dose efficiency, combined with iterative reconstruction algorithms, can decrease noise levels and allow for large matrix reconstructions that lead to ultra-LD protocols.^29,30 Due to increased data complexity and spectral information, a novel algorithm, quantum iterative reconstruction, with four strength levels (QIR-1–4) has been developed for PCD-CT.²⁹ According to a preceding article, QIR-3 dispensed the highest spatial resolution and noise texture; thus, we applied QIR-3 for our protocols.²⁹ Additionally, it has been described that significant dose reduction and conservation of HR parameters for lung parenchyma assessment is possible with PCD-CT, either with or without iterative reconstruction.

In a pilot study, Inoue et al.³¹ demonstrated that PCD-CT produced better image quality and enhanced diagnostic confidence for lung parenchymal abnormalities at reduced radiation doses. Jungblut et al.³² further confirmed that PCD-CT provides good image quality with lower radiation doses, compared with EID-CT. Our phantom studies confirmed that, compared with EID-CT measurements, PCD-CT protocols produce improved dose-matched CNR and SNR values (Figure 7). Previous LD EID-CT protocols (i.e., <1 mSv) are not recommended for diagnostic use, as their impaired image quality could lead to the misclassification of ILD.³³ Chest HR EID-CT has an effective dose of approximately 6–9 mSv, according to the literature, while in our study, the effective dose of PCD-CT was 0.4 mSv for HR scans and 3.5 mSv for UHR scans. Previously reported average LD EID-CT protocols had an effective dose of 2.1–2.4 mSv, significantly higher than our HR PCD-CT dose value.^13,26,32,34 LD CT has been increasingly used in the assessment of pulmonary cancer; however, this is not the only pulmonary disorder in which the risk–benefit ratio could be positive. For instance, the follow-up of ILD at low doses of radiation could be of interest.^{33,35,36,37,38} Our data suggest that PCD-CT is a promising tool in radiation dose optimization, which is crucial in optimizing the risk–benefit ratio of CT lung screening.³⁹

UHR scans proved to be more sensitive in the detection of bronchiectasis and reticulation; hence, their total score values were slightly higher. However, the identification of GGO and honeycombing values was the same between protocols (Figure 4). The UHR protocol was slightly more sensitive to interstitial pathologies; however, the magnitude of differences was not protruding. Moreover, the same ILD patterns were identified with both protocols. Taking into consideration the dose values that were notably lower (~7.4×) in LD HR scans, the LD measurements were able to assess interstitial changes with good proximity.

While this study has limitations (e.g., the relatively low number of patients enrolled), it is comparable to other international investigations. Furthermore, it would be interesting to set against our results from other PCD-CT protocols with different parameters. However, the benefits of better image quality need to be balanced with the risks of higher radiation exposure in these patients. Phantom studies to compare different detector-type CT protocols can be conducted to avoid increased radiation doses for patients. Additionally, ILD multidisciplinary team discussions are needed to gauge the difference between these two CT modalities in clinical settings to improve team diagnosis, especially of early cases. Longitudinal radiological data on natural behavior and disease-specific treatment of early RA-ILD are also needed.

In conclusion, wide-scale clinical experience with UHR CT imaging to assess lung involvement in patients with RA does not exist. In this proof-of-concept study, we found that a UHR PCD-CT protocol provided more detailed images compared with an HR PCD-CT protocol. The HR PCD-CT protocol provided detailed information regarding interstitial lung involvement; however, in the case of an extended or complex pathology, additional UHR imaging may prove beneficial. Further studies are needed to determine if an HR PCD-CT protocol, with its reduced radiation doses, could serve as an initial screening tool before selecting patients for further UHR imaging. From a clinical perspective, the higher effective radiation dose of UHR PCD-CT is balanced by its better characterization of pulmonary involvement, which provides the potential for earlier anti-fibrotic treatment, a very important intervention in RA patients with ILD.

Acknowledgements

We would like to thank for the help and support of the colleagues mentioned afterwords:

Hanna Balogh MD¹, Leila Szeibel MD¹, Tamas Purczel MD¹, Tamas Munkacsi MD¹, Nora Kerkovits MD¹, Klaudia Borbely MD¹, Laszlo Szakacs¹, Aniko Kubovje¹, Kinga Karlinger MD, PhD¹, Dora Sarvari MD², Timea Petri MD²

¹Department of Radiology, Medical Imaging Centre, Semmelweis University, Budapest, Hungary

²Buda Hospital of the Hospitaller Order of Saint John of God, Budapest, Hungary

Conflict of interest disclosure

The authors declared no conflicts of interest.

Funding

Nikolett Marton MD, PhD (first author) received a Bolyai Research Scholarship from the Hungarian Academy of Sciences.