Article Text
PDF
Abstract
Objectives The objective of this study is to assess diagnostic accuracy for the detection of interstitial lung disease (ILD) in image series with high increment and reduced number of slices in patients with systemic sclerosis (SSc).
Methods 45 patients with SSc underwent high-resolution CT (HRCT). Three series of secondary captures were reconstructed as follows: series 1, series with 10 mm increment and 1 mm slices; series 2, seven axial images with baso-apical gradient; series 3, three axial images were obtained at the apical, at the level of the carina and basal. The presence and extent of ILD, and the degree of diagnostic confidence were recorded. The effective dose for each image series was estimated. Standard HRCT was the standard of reference.
Results The prevalence of ILD was 55% (25/45). Diagnostic sensitivity and accuracy of series 1, series 2 and series 3 were 100% and 94.4%, 94% and 97.8%, 92% and 97.8%, respectively. The extent of ILD was underestimated in series 3 (p<0.05) and was comparable to the standard HRCT in series 1 and 2 (p>0.05). Estimated dose reduction was more than 90% in all image series.
Conclusions HRCT image series with low sampling rate allow an accurate detection of ILD with very-low-radiation dose, making this approach potentially valuable for screening in patients with SSc.
Statistics from Altmetric.com
Introduction
Interstitial lung disease (ILD) is frequent in systemic sclerosis (SSc) and represents the leading cause of disease-related death.1 Thus, evaluation of lung involvement at the time of SSc diagnosis is essential.2 High-resolution CT (HRCT), a non-invasive and sensitive technique, represents the gold standard for ILD diagnosis because it can detect lung involvement prior to appearance of symptoms and provides prognostic information.2,–,4
On the downside, CT uses ionising radiation3 that has been linked to an increased risk of cancer.5 Various adjustments of acquisition settings have been assessed to reduce radiation, and lowering the tube current is the most widely employed adjustment setting in chest CT.6 Alternatively, a reduced CT scan protocol with a limited number of slices can also be used.7 8 However, this technique carries the risk that with the decrease in image numbers, abnormal findings remain undetected. Therefore, images should be obtained in predefined lung regions, representative of the distribution of parenchymal abnormalities.
In this study, we aimed at developing CT scan protocols with reduced radiation to make CT scanning feasible for screening purposes. In an image series with a different number of slices and regional distribution, we assessed radiation dose and diagnostic accuracy for the detection of ILD compared to standard HRCT in patients with SSc.
Methods
Patient population
Seventy-five patients who were a part of the SSc registry at the Department of Rheumatology, University Hospital Zurich, underwent HRCT between 2003 and 2010. Exclusion criteria consisted of the following: (1) lung pathologies not associated with SSc (ie, lung emphysema, lung cancer, pneumonia; n=10), (2) extensive pleural effusion (n=2), (3) extensive motion artefacts (n=2) and (4) external CT scans when secondary reconstruction could not be performed (n=16). The remaining 45 patients made up the study population.
Recruitment into the SSc cohort and assessment of clinical parameters was done prospectively, and image analysis, retrospectively.
All patients signed informed consent. The study was approved as part of the SSc registry by the institutional review board.
Image acquisition
HRCT scans were obtained with 16-slice and 64-slice CT scanners ((Somatom Definition AS/Definition Flash/Sensation 64/Sensation 16; Siemens Healthcare, Forchheim, Germany) (Light speed VCT; GE Healthcare, Munich, Germany) (Brilliance 16P; Philips Healthcare, Eindhoven, the Neatherlands)).Patients were examined in the supine position and at end inspiration with generally accepted acquisition parameters.9 The original HRCT scans were the standard of reference.
Using the data of the original HRCT scan, three image series were reconstructed as following: series 1, series with 10 mm increment and 1 mm slices; series 2, seven axial images with baso-apical gradient (at the level of the sternoclavicular joint; at the carina; five images basal with 10 mm increment), according to the regional distribution of ILD;2 series 3, three axial images (at the level of the sternoclavicular joint; at the carina; at the top of the diaphragm) (figure 1).
The three topograms displaying levels of CT scans for secondary reconstructed image series: Series 1 consisted of images with 10 mm increment and 1 mm slices (a); series 2 consisted of seven images with baso-apical gradient (at the level of the sternoclavicular joint, at the level of the carina and five images basal with 10 mm increment) (b); series 3 consisted of three images (at the level of the sternoclavicular joint, at the level of the carina and at the top of the diaphragm) (c).
Image assessment
Images were anonymised and displayed at lung parenchyma setting (B60f, W1200:L-600).
On the original HRCT scans, the presence of ILD (1, ILD present; 2, no ILD) was assessed by two blinded readers (AW, 2 years of experience; TF, 10 years of experience) in consensus. The extent of ILD was graded as no lung involvement, lung involvement less than 20%, lung involvement more than 20% (ie, extensive disease) or indeterminate (ie, extent not readily classifiable).3 In cases of discrepancies concerning the presence of ILD, the written report, clinical findings and follow-up CT scans were taken into account. Furthermore, predominant patterns of parenchymal abnormalities were assessed as previously done.10
Reduced image series were assessed by two blinded radiologists (NB, 2 years or experience; TF, 10 years of experience). Again, the presence and extent of ILD were assessed. Furthermore, the degree of diagnostic confidence was recorded (score 1 to 4, 1=certain that ILD is present; 2=quite sure that ILD is present; 3=quite sure that no ILD is present; 4=certain that no ILD is present).
Radiation dose estimation
One patient was scanned according to the three image series to estimate radiation dose reduction compared to standard HRCT (Somatom Definition Flash; Siemens Healthcare, Germany).
Statistical analysis
Statistical analysis was performed using SPSS V.18.0. Continuous variables were reported as mean±SD, and categorical variables, as frequencies or percentages. Cohen's Kappa statistics were calculated for interobserver agreements. The presence of ILD on the scans was compared using Cochran's Q test and McNemar test. The extent of parenchymal abnormalities was compared using the Wilcoxon signed ranks test.
Sensitivity, specificity and accuracy were calculated with regard to the detection of ILD. CI was 95%. The average values among the two readers were estimated to reflect average community practice.11
Statistical significance was inferred at a p-value that was below 0.05.
Results
Patient demographics
Of the 45 patients included (39 women; age, 54.0±13.0 years), 38 fulfilled the criteria for SSc as proposed by LeRoy et al,19 five were diagnosed as having early SSc without skin fibrosis according to the criteria by LeRoy and Medgser.20 Two patients had an overlap syndrome of SSc with polymyositis and rheumatoid arthritis, respectively (table 1).
Patients' characteristics
Image assessment
In the original HRCT scans, ILD was found in 55% (n=25) of the patients (predominant patterns as found in standard HRCT are listed in table S1 in the online supplement).
In the secondary reconstructed series (k, 0.87–1), the detection of ILD did not significantly differ (p>0.05 for both readers). Good sensitivity, as well as high certainty for either presence or absence of ILD (ie, degree of confidence score 1 or 4), was obtained in all image series (table 2).
Estimated accuracy and diagnostic certainty in detecting ILD on secondary reconstructed HRCT image series
The extent of ILD (k, 0.65–0.82) was underestimated in series 3 (p<0.05 vs standard HRCT), and was comparable to standard HRCT in series 1 (reader 1, p=1; reader 2, p=0.617) and series 2 (reader 1, p=0.48; reader 2, p=1) (online supplementary figure S1).
Radiation dose estimation
Estimated effective radiation dose was 0.154 mSv (equivalent to a dose reduction of 90.52%, standard of reference 1.6 mSv) for series 1, 0.028 mSv (a dose reduction of 98.27%) for series 2 and 0.014 mSv (a dose reduction of 99.14%) for series 3.14
Discussion
ILD is the leading cause of death in patients with SSc.1 HRCT represents the gold standard for ILD diagnosis and provides essential information for its treatment and prognosis.2 3 However, ionising radiation is associated with an increase in the risk of cancer.5 15 Patients with SSc are often women and are younger at disease onset and, thus, more sensitive to radiation. Furthermore, cumulative effects due to serial testing play an important role.16 The International Commission on Radiological Protection states that the benefits of an imaging procedure must exceed the risks and that attempts should be made to reduce the potential risks as low as reasonably achievable.15 Thus, various adjustments of acquisition settings have been explored to reduce radiation dose of chest CT.6 Furthermore, other imaging modalities with no or less ionising radiation, including chest sonography and tomosynthesis (a three-dimensional imaging technique), have been recently described. However, both techniques allow only limited assessment of the lung.17 18 In this study, we employed another strategy: we decreased the sampling density of CT scans, thereby reducing the effective scan length and, consecutively, the radiation dose, as previously done by Bankier et al and de Jong et al for other indications.7 8 We could herein demonstrate an estimated dose reduction of more than 90% compared to standard HRCT while maintaining high sensitivity for ILD detection.
However, the extent of parenchymal abnormalities may be underestimated in image series with very-low-sampling rates (ie, series 3) and in image series with a higher sampling rate at the lower zone of the lung (ie, series2), which, according to the classical distribution of ILD,2 should be preferred. Similarly, Bankier et al proposed that due to the heterogeneity of distribution (ie, air trapping in patients with bronchiolitis), the extent of abnormalities found depends on the examination protocol and that different protocols (with reduced number of slices) result in a different extent.7 Thus, concerning ILD, the choice should be towards protocols with higher sampling rate at the lower zone.
Limitations of the study are the retrospective assessment of CT images and the relatively small number of patients, with SSc being an orphan disease. Second, we only assessed supine sequences at end inspiration, and not at expiratory or prone sequences, which may be part of a full HRCT protocol. Furthermore, our image series were secondary reconstructed series obtained from a single initial CT scan, and not real examination protocols. Thus, a true comparison of radiation dose in all patients was not possible. Other lung pathologies were not assessed; it might be possible that with a limited protocol, critical abnormalities including lung cancer might be missed.
In conclusion, we assessed the diagnostic accuracy of image series with high increment and reduced the number of slices in the detection of ILD associated with SSc. Herein, we could demonstrate an accurate detection of ILD with an estimated radiation dose reduction of more than 90%. These results might have influence on the clinical care of patients, because CT protocols with low radiation exposure might allow the use of annual screening programs in patients with SSc leading to an earlier detection of SSc-ILD.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Web Only Data - This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Web Only Data - This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
-
Competing interests None.
-
Patient consent All patients signed informed consent.
-
Ethics approval Ethical approval was obtained from the institutional review board.
-
Provenance and peer review Not commissioned; externally peer reviewed.