Usefulness of concurrent reading using thin-section and thick-section CT images in subcentimetre solitary pulmonary nodules

Aim To evaluate the differences in the characterization and recommendation for follow-up of subcentimetre solitary pulmonary nodules (SSPNs) between 5 and 1 mm section CT, and to compare the assessments generated by four radiologists Materials and methods Five hundred and twenty-nine patients who had SSPNs on chest CT reconstructed using both 5 and 1 mm sections were enrolled. Two image subsets of 5 and 1 mm CT images of each nodule were interpreted independently by four radiologists. Nodule size, consistency (solid, partly solid, non-solid), the presence of calcification, and recommendations for follow-up were evaluated. If a non-calcified solid nodule was confirmed using CT, recommendation for follow-up was based on Fleischner Society guidelines. Data assessed by each radiologist were compared, and interobserver agreements were determined using the intraclass correlation coefficients and kappa value. Results Using 1 mm CT images, the nodule sizes were significantly larger than on 5 mm CT images (paired t -test, p < 0.01). The presence of calcification and nodule consistency were significantly different between 5 and 1 mm CT images (McNemar test for the presence of calcification, p < 0.01; Wilcoxon signed test for nodule consistency, p < 0.01). On 1 mm CT images there was significantly higher agreement regarding nodule consistency than on 5 mm CT (kappa = 0.78 and 0.67, respectively). Conclusions Concurrent use of thin-section and thick-section CT can provide more accurate nodule assessment and higher interobserver agreement in SSPN. Introduction With widespread clinical use of computed tomography (CT), subcentimetre solitary pulmonary nodules (SSPN) have became a common finding during CT examinations. 1–3 The majority of these lesions are benign, 4 even in patients at high-risk of developing lung cancer or metastatic disease; therefore, histological proof may not be necessary in every case. Unnecessary follow-up CT examinations can be reduced if radiological features more accurately predict benignity with satisfactory certainty. The thin-section CT obtained through the nodule has been shown to provide more accurate information regarding nodule size, consistency, edge characteristics, and the presence of calcification, cavitation, or fat. 5 However, reading the whole volume data obtained with thin-section chest CT requires long interpretation times. If CT is obtained using submillimetre collimation, both thin- and thick-section CT images could be reconstructed from the same raw data without additional radiation exposure. 6 With the concurrent use of thick- and thin-section CT, the nodule can be detected using thick-section CT, and then can be characterized using thin-section CT by reviewing the selective images that cover the nodule. Therefore, the purpose of the present study was to evaluate the differences in the characterization and recommendation for follow-up of SSPNs between 5 and 1 mm section CT examinations, and to compare the assessments generated by four different radiologists, in order to evaluate the usefulness of concurrent reading with 5 and 1 mm section CT in SSPNs. Materials and methods This study was approved by the Institutional Review Board and patients' informed consent was waived in this retrospective study. Between January 2006 and January 2007, a consecutive review of the CT findings was retrospectively performed and patients who met the following CT criteria were included: 5 and 1 mm sections concurrently obtained; a single pulmonary nodule detected; and a lesion that had a maximum diameter ≤10 mm. The decision for nodule presence was established by consensus reading between two radiologists with 5 and 25 years experience, respectively. A nodule was defined as a round or oval lesion of lung parenchyma with intact surrounding lung, pleura, and mediastinum. 6 The enrolled study population consisted of 529 participants (350 men and 179 women) ranging in age from 35–83 years (mean 52 ± 9 years). Chest CT was performed using unenhanced spiral CT (Sensation 16; Siemens Medical Solutions, Forchheim, Germany) at end-inspiration in one breath-hold. Examination parameters were 120 kVp, 40 mAs, 16 × 0.75 mm collimation, and 18 mm/rotation table feed. Two image sets were reconstructed concurrently from the same raw data into section thicknesses of 5 and 1 mm by means of a high special frequency algorithm. Therefore, no additional radiation exposure was necessary to obtain the two datasets. The two radiologists, who ascertained the presence of the nodules, collected the images containing the lung nodules and annotated the nodules. Four radiologists (with 20, 13, 11, and 9 years experience, respectively) independently interpreted the 5 mm CT image subsets first, and subsequently, the 1 mm CT image subsets, using the picture archiving and communication system (PACS), with an interval of at least 2 weeks between the two readings. Only the age, sex, and smoking history of the participants were provided to the reviewers. The characteristics of the nodules that were recorded included: size, consistency, and the presence of calcification separately for 5 mm and 1 mm image subsets. Nodule size was measured manually by averaging the length and width in the transverse plane with an electronic calliper. Consistency was divided into three categories: solid, partly solid, and non-solid. Reviewers were requested to give a recommendation for follow-up. If the CT image indicated a non-calcified, solid nodule, recommendation for follow-up was given based on Fleischner Society guidelines. 7 If a partly solid or non-solid nodule was detected, recommendation for follow-up was recorded subjectively due to the absence of widely accepted guidelines. SPSS version 11 software for Windows (SPSS, Chicago, IL, USA) was used for statistical analyses. Nodule sizes measured by each radiologist from the 5 and 1 mm sections were compared using the paired t -test. For the presence of calcification and recommendation for follow-up, comparisons were assessed using the McNemar test. The Wilcoxon signed-rank test was used to compare nodule consistency and follow-up periods. Interobserver agreements among four reviewers were determined through intraclass correlation coefficients 8 for continuous variables and kappa value for categorical variables. Results Comparison between CT Interpretations of the 5 and 1 mm sections In terms of nodule size, the diameter of nodules measured on 5 mm section CT were usually smaller than that on 1 mm section for all reviewers ( Table 1 ). Smaller nodules tended to show higher variation in calculated diameter compared with larger nodules. The difference in the diameters between 5 and 1 mm section CT was statistically significant ( p < 0.001). On 1 mm section CT, 28–48 more calcifications (5–9% of the nodules) were detected within the nodules than on 5 mm section CT. There were statistically significant differences between the 5 and 1 mm section CT images for the presence of calcification ( p < 0.001; Table 2 , Fig. 1 ). Nodule consistency was also significantly different between the 5 and 1 mm CT images ( p < 0.001; Table 3 , Fig. 2 ). Twenty-four to 64 partly solid or non-solid nodules (4.5–12.1% of the total nodules) on 5 mm CT were interpreted to be solid nodules on 1 mm CT (47, reviewer 1; 64, reviewer 2; 53, reviewer 3; 24, reviewer 4). Nodules <4 mm contributed greatly (52–68% of the given nodules) to this difference in the assessment of nodule consistency (32, reviewer 1; 33, reviewer 2; 34, reviewer 3; 13, reviewer 4). In terms of the recommendation for follow-up based on Fleischner Society guidelines for solid nodules, follow-up was not recommended on 1 mm CT even when it was recommended on 5 mm CT images, in 24 to 34 nodules ( Table 4 ). Interobserver agreement Table 5 lists the interobserver agreement of each finding. For the size, the reviewers' agreement using 1 mm CT images was higher than that using 5 mm CT images. Regarding the consistency of the nodules, using the 1 mm CT images there was significantly higher agreement compared with the 5 mm CT images. In the recommendation for follow-up ( Table 6 ), the inter-reviewer agreement using 1 mm CT was also higher than that using 5 mm CT. In the analysis of all nodules including partly solid or non-solid nodules, recommendation for follow-up had significantly lower agreement than that of solid nodules alone, using the Fleischner Society guidelines. Discussion The results of the present study clearly showed significant differences in the characterization and recommendation for follow-up of SSPN between 5 and 1 mm section CT. The reduced section thickness improved the accuracy of the measurement of nodule size and increased the detection rate of calcification. 9,10 More accurate detection of benign calcified nodules could help to reduce unnecessary follow-up, especially in patients with low cancer risk. This is why only patients with SSPN at initial CT were selected. Furthermore, in terms of the nodule consistency, the volume-averaging effect in thick sections caused solid nodules to mimic partly solid or non-solid nodules. 11 Because persistent, partly solid nodules are most likely to be malignant, 12–14 misinterpreted partly solid or non-solid nodules may cause problems at repeat CT or at follow-up examinations. The tasks involved in the interpretation of SSPN, such as discrimination of non-calcified nodules from benign lesions, measurement of nodule size, and assessment of the risk of malignancy, inevitably have some degree of subjectivity and variability among readers. 15 In the present study, 1 mm CT images showed higher interobserver agreement, except for the presence of calcification. Thin-section CT is a promising method of reducing observer variability at routine chest CT interpretation and screening CT interpretation. In a previously published study, 15 authors had reported greater disagreement in the measurement of nodule size of small nodules compared with large nodules. In the present study, the effect of lesion size on agreement was also confirmed. If the presence of calcification is suspected within a SSPN, it is hard to determine the pattern of calcification suggesting a benign nature 16 and to measure CT number due to its small size. If the nodule is <5 mm in diameter, the presence of calcification can be determined by comparing the attenuation of the nodule with that of the ribs on the images obtained with bone and lung window. 17 This suggestion is very subjective and, consequently, the presence of calcification depends on the radiologists' subjective decision. Relatively lower interobserver agreement was found on the presence of calcification using 1 mm CT images compared with 5 mm CT images. This might have been caused during the reconstruction of images with a sharp algorithm. Edge enhancement in the nodule boundary, especially using 1 mm section CT images, may result in the false-positive interpretation of a non-calcified nodule. Although a sharp reconstruction algorithm helps characterizing nodule shape, internal consistency, and margins, the results of the present study suggest that the use of a standard reconstruction algorithm using 1 mm CT images may improve interobserver agreement regarding the presence of calcification. Regarding partly solid or non-solid nodules, only the inter-reviewer's agreements of follow-up recommendation were evaluated, unlike the solid nodules. The reason for analysing the recommendation for follow-up between the 5 and 1 mm CT image subset in solid nodules was to validate the potential of thin-section images to reduce unnecessary follow-up and to prolong the follow-up intervals by better confirming the benignity of SSPN; therefore, the potential radiation hazard of frequent CT examination can be avoided. 18 However, for the partly solid or non-solid nodules, the concept of the recommendation for follow-up is different from the solid nodules. The report of Hasegawa et al. 19 supports the use of extended follow-up intervals for small partly solid or non-solid nodules, even in high-risk patients. Fleischner Society guidelines for incidentally detected partly solid or non-solid nodules suggest longer interval follow-up 7 considering the slower growth rate of lung cancer appearing as partly solid or non-solid nodules. However, follow-up of partly solid or non-solid nodules would be needed only if the nodule persists on the short-term follow-up CT within 3 months, thus excluding the possibility of a transient inflammatory nodule. For similar reasons, it is not known how often and how long partly solid or non-solid nodules should be followed up using CT. In this study, the assessment using a subjective standard rating from each radiologist in the view of recommendation for follow-up of partly solid or non-solid nodules varied considerably even though they were experienced chest radiologists. Recommendations for follow-up of solid nodules had a “very good” degree of interobserver agreement under the guidelines, but those of partly solid or non-solid nodules showed relatively lower interobserver agreement. This is mainly because the latter is less standardized. The observations of the present study suggest that a standardized guideline for the management of partly solid or non-solid nodules is needed. The present study had several limitations. There was no histological proof for the detected nodules. An edge-enhancing kernel was used, which may falsely interpret a non-calcified nodule as a calcified nodule. Nodule characterization was the focus of the study, not on the nodule detection. The interpretation time for SSPN was not assessed, even though interpretation time is clinically a major concern in evaluating the subcentimetre nodules using thin-section CT images. The main drawback of CT using thin sections or small reconstruction intervals is the large size of the image datasets and the considerable amount of interpretation time. 20 To reduce interpretation time, nodule search may be performed using thick-section CT, and then nodule characterization can be performed using thin-section CT by selectively reviewing the images that cover the nodule. Despite several limitations, concurrent use of thin-section and thick-section CT may help to increase the agreement on routine chest CT and screening CT interpretation of SSPN. In conclusion, concurrent use of thin-section and thick-section CT can provide more accurate nodule assessment and higher interobserver agreement in SSPN. Acknowledgement This study was supported by Seoul R&BD program (10888). References 1 C.I. Henschke D.I. McCauley D.F. Yankelevitz Early Lung Cancer Action Project: overall design and findings from baseline screening Lancet 354 1999 99 105 2 S. Diederich D. Wormanns M. Semik Screening for early lung cancer with low-dose spiral CT: prevalence in 817 asymptomatic smokers Radiology 222 2002 773 781 3 H. Saito Y. Minamiya H. Kawai Usefulness of circumference difference for estimating the likelihood of malignancy in small solitary pulmonary nodules on CT Lung Cancer 58 2007 348 354 4 D. Wormanns S. Diederich Characterization of small pulmonary nodules by CT Eur Radiol 14 2004 1380 1391 5 C.V. Zwirewich S. Vedal R.R. Miller Solitary pulmonary nodule: high-resolution CT and radiologic–pathologic correlation Radiology 179 1991 469 476 6 D.R. Aberle G. Gamsu C.I. Henschke A consensus statement of the Society of Thoracic Radiology: screening for lung cancer with helical computed tomography J Thorac Imaging 16 2001 65 68 7 H. MacMahon J.H. Austin G. Gamsu Guidelines for management of small pulmonary nodules detected on CT scans: a statement from the Fleischner Society Radiology 237 2005 395 400 8 P.E. Shrout J.L. Fleiss Intraclass correlations: uses in assessing rater reliability Psychol Bull 86 1979 420 428 9 H.T. Winer-Muram S.G. Jennings C.A. Meyer Effect of varying CT section width on volumetric measurement of lung tumors and application of compensatory equations Radiology 229 2003 184 194 10 J.M. Goo T. Tongdee R. Tongdee Volumetric measurement of synthetic lung nodules with multi-detector row CT: effect of various image reconstruction parameters and segmentation thresholds on measurement accuracy Radiology 235 2005 850 856 11 C.M. Park J.M. Goo H.J. Lee Nodular ground-glass opacity at thin-section CT: histologic correlation and evaluation of change at follow-up RadioGraphics 27 2007 391 408 12 C.I. Henschke D.F. Yankelevitz R. Mirtcheva CT screening for lung cancer: frequency and significance of part-solid and nonsolid nodules AJR Am J Roentgenol 178 2002 1053 1057 13 M. Nakata H. Saeki I. Takata Focal ground-glass opacity detected by low dose helical CT Chest 121 2002 1464 1467 14 F. Li S. Sone H. Abe Malignant versus benign nodules at CT screening for lung cancer: comparison of thin-section CT findings Radiology 233 2004 793 798 15 D.S. Gierada T.K. Pilgram M. Ford Lung cancer: interobserver agreement on interpretation of pulmonary findings at low-dose CT screening Radiology 246 2008 265 272 16 E.A. Zerhouni F.P. Stitik S.S. Siegelman CT of the pulmonary nodule: a cooperative study Radiology 160 1986 319 327 17 New York Early Lung Cancer Action Project Investigators CT screening for lung cancer: diagnoses resulting from the New York Early Lung Cancer Action Project Radiology 243 2007 239 249 18 D.J. Brenner Radiation risks potentially associated with low-dose CT screening of adult smokers for lung cancer Radiology 231 2004 440 445 19 M. Hasegawa S. Sone S. Takashima Growth rate of small lung cancers detected on mass CT screening Br J Radiol 73 2000 1252 1259 20 J.S. Kim J.H. Kim G. Cho Automated detection of pulmonary nodules on CT images: effect of section thickness and reconstruction interval—initial results Radiology 236 2005 295 299