| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the Department of Radiology (Drs. K.S. Lee, Yi, S.Y. Jeong, Chung, H.Y. Kim, Y.K. Kim, and K.H. Lee), Center for Imaging Science, and Biostatistics Unit (Dr. S. Kim), Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea; and the Department of Radiology (Dr. Y.J. Jeong), Pusan National University Hospital, Pusan, Korea.
Correspondence to: Kyung Soo Lee, MD, Department of Radiology, Samsung Medical Center, 50, Ilwon-Dong, Kangnam-Ku, Seoul 135–710, Korea; e-mail: kyungs.lee{at}samsung.com
Abstract
Background: Solitary pulmonary nodule (SPN) evaluation based on analyses of combined wash-in (WI) and washout (WO) values obtained by helical dynamic CT (HDCT) scanning is useful for malignant SPN characterization, because this method has higher specificity and accuracy than that based on analyses of WI values only. However, increased specificity results in reduced sensitivity and the missing of malignant SPNs. Thus, the purpose of this study was to seek the most effective method for SPN characterization during HDCT scanning.
Methods: After obtaining unenhanced CT scans, dynamic CT scanning was performed using a helical technique (images were obtained at 30, 60, 90, and 120 s, and at 5 and 15 min after the initiation of IV contrast administration) in 486 patients with a solid or partly solid SPN. Diagnostic efficacies were compared for three approaches involving considerations of WI values (in Housfield units [HU]) only, both WI and WO HU values, and WI HU values and morphologic characteristics.
Results: Considering WI values only (
25 HU), sensitivity, specificity, and accuracy for malignancy were 98% (233 of 237 nodules), 46% (114 of 249 nodules), and 71% (347 of 486 nodules), respectively. Using both a WI value of 25 HU and a WO value of 5 to 36 HU, the corresponding values were 89% (212 of 237 nodules), 79% (197 of 249 nodules), and 84% (409 of 486 nodules), respectively; for a WI value of 25 HU and a malignant morphology, the corresponding values were 92% (219 of 237 nodules), 79% (197 of 249 nodules), and 86% (416 of 486 nodules), respectively (these values were significantly different between the WI-only group and the other two groups; p = 0.001).
Conclusions: The efficacy of SPN evaluation based on analyses of WI values plus morphologic features during HDCT scanning appears to be equivalent to that based on analyses of WI plus WO values, thus obviating the need for WO scans, which saves time and reduces radiation exposure of the patient.
Key Words: CT scan • diagnosis • lung neoplasms
Morphologic evaluations of solitary pulmonary nodules (SPNs) can help to differentiate benign and malignant SPNs when they have typical benign or malignant features, but there is considerable overlap between their nodule types in terms of morphologic presentations.1 Various strategies other than morphologic evaluations have been applied to the differentiation of malignant and benign nodules.234567 Of these, helical dynamic CT (HDCT) scan characteristics and positron emission tomography findings are regarded to be the most useful for differentiating malignant and benign nodules.45
Problems encountered in HDCT scans in terms of wash-in (WI) characteristics are related to its low specificity for malignant nodules. Two studies, one from 20054 and the other from 2006,5 have suggested that SPN evaluations based on analyses of combined WI and washout (WO) values during HDCT scanning are useful for malignant nodule characterization because this method has higher specificity and accuracy than that based on analyses of WI values only. However, in these studies, increased specificity that leads to the avoidance of unnecessary surgery for benign nodules resulted in reduced sensitivity and the missing of malignant nodules. Therefore, the most efficient way of characterizing SPNs with nearly perfect sensitivity and a reasonably high specificity has yet to be determined.
While the hemodynamic attributes of WI (at > 25 Housfield units [HU]) and WO (at 5 to 31 HU) that were determined by HDCT scanning were found to be an independent factor for the differentiation of malignant nodules by multivariate analysis, morphologic characteristics (ie, nodules with a lobulated or spiculated margin) and the absence of a satellite nodule on thin-section CT scans were also found to be independent factors.4 However, nodule characterizations based on the combined interpretation of dynamic-study WI characteristics and morphologic features have not been reported. Thus, the purpose of the present study was to compare the following three different approaches in terms of their ability to characterize solid or partly solid SPNs: considerations of WI hemodynamic values only; considerations of WI and WO features; and considerations of WI values and morphologic characteristics in combination.
Materials and Methods
Patients and CT Imaging
Institutional review board approval and informed consent for this HDCT scan study were obtained. From October 2003 to March 2006, a total of 689 patients with a solid or partly solid SPN underwent dynamic chest CT scanning using an eight-row multidetector CT scanner (LightSpeed Ultra; GE Medical Systems; Milwaukee, WI) or a 16-row multidetector CT scanning (LightSpeed 16; GE Medical Systems).
Before performing dynamic CT scanning, we obtained targeted thin-section helical CT scans (helical technique; 120 kVp; 90 mA; 0.8-s gantry rotation time; beam width, 10 mm; table speed, 13.75 mm per rotation; reconstruction performed with a bone algorithm; and 1.25-mm section thickness) through the nodule concerned. Nodules were excluded if they had benign patterns of calcification (ie, diffuse, laminated, popcorn-like, or central; n = 11) or fat patterns of calcification (n = 2) on thin-section CT scan images. Nodules with predominantly ground-glass opacity and little solid portion (
5 mm in diameter; n = 8) and nodules showing total necrosis (n = 1) on thin-section CT scans were also excluded. After the above exclusions, dynamic CT studies were performed in the remaining 667 patients.
Before IV contrast administration, a series of 13 images was obtained throughout the nodule so as to cover a distance of 30 mm along the z-axis at 120 kVp, 90 mA, a 0.8-s gantry rotation time, a beam width of 10 mm, and a table speed of 13.75 mm per rotation. Thereafter, an additional six series of images were obtained at 30, 60, 90, and 120 s, and at 5 and 15 min after the initiation of the contrast medium injection (with an infusion rate of 3 mL/s, for a total of 120 mL of Iomeron 300 [Iomeprol]; Bracco; Milan, Italy) using a power injector (MCT Plus; Medrad; Pittsburgh, PA) and the same parameters as those used for the initial preenhancement series (for a total of seven series of images at time [T] 0, T30 s, T60 s, T90 s, and T120 s, and at T5 min and T15 min). Image data were reconstructed using a thickness of 2.5 mm (13 images in each cluster; total number of dynamic images, 13 x 7 = 91) and a standard algorithm. Immediately after the dynamic study performed at 120 s after the initiation of contrast medium injection, helical CT scans (120 kVp; 125 mA; beam width, 20 mm; table speed, 27.5 mm per rotation; reconstruction with a bone algorithm; and 5 mm thick) were obtained from lung apices to the level of the middle pole of both kidneys for tumor staging. All thin-section, dynamic CT scans and the staging of CT scan image data were directly interfaced to our picture archiving and communication system (MT; General Electric Medical Systems Integrated Imaging Solutions; Prospect, IL), which displayed all image data on monitors (1,536 x 2,048 image matrices, 8-bit viewable gray scale, and 60-foot-Lambert luminescence). Both mediastinal window images (window width, 400 HU; window level, 20 HU) and lung window images (window width, 1,500 HU; window level, –700 HU) were viewed on the monitors.
The technical adequacy of the CT dynamic study fulfilled the criteria detailed in a previous study.2 In six patients, inconsistent breathholds were a problem in terms of obtaining dynamic image clusters at the same level. In another 10 patients, nodules were too small ( 5 mm in diameter) to measure attenuation values (partial volume averaging precluded attenuation value measurements), and in another 4 patients it was difficult to measure the true attenuation value due to beam-hardening artifacts. With the exception of these 20 patients, the dynamic studies were technically adequate for 647 patients.
Of the 647 patients, 161 patients for whom neither follow-up CT scans nor cytologic or histologic diagnosis was available were excluded. Therefore, the final study group of 486 comprised 299 men and 187 women aged 22 to 85 years (mean [± SD] age, 56 ± 12 years). These patients underwent transthoracic needle biopsy (n = 182) or surgery (n = 170), including lobectomy (n = 150) or wedge resection (n = 20). One hundred thirty-four patients in whom a histopathologic diagnosis had not been obtained were regarded as having a benign nodule, because these nodules did not change in size (total, 41 cases; 33 cases were followed up for 18 to 24 months, and the remaining 8 cases were followed up for > 24 months), as determined by diameter measurements, or showed a decrease in size (n = 93). Nodules that did not change in size were followed up by CT scan at least once for over 18 months (follow-up CT scan schemes, 3 months after the initial study and then at 6-month intervals, thereafter; mean follow-up period, 23 ± 5.5 months; follow-up range, 18 to 33 months).
Evaluation of Enhancement Dynamics
One of two radiologists (with 3 and 15 years of experience in chest CT scan interpretation) measured the attenuation values. After viewing all 91 images of each nodule as thumbnail images on picture archiving and communication system monitors, we selected 1 image for analysis from the 13 images of the nodule obtained at a given time. The seven selected images per nodule were transverse sections with the largest diameter (scanned at the equator of the nodule). We measured nodule attenuation values in the same area on the selected image for each image cluster at each given time (from unenhanced images to 15-min images). Briefly, a circular region of interest (ROI) was placed over a nodule. This ROI covered about one half of the nodule diameter at the equator. When a calcification (n = 60; average area at the selected image, 5%, range, 2 to 20%), a cavity (n = 82; average area, 24%; range, 2 to 40%), or a necrotic area (n = 97; average area, 17%; range, 5 to 60%) was present, they were avoided, and ROIs were made as large as possible in unaffected areas. All HU measurements were performed on mediastinal-window images to ensure that the partial volume-averaging effects were minimized. All measurements were made at the time of CT scan examinations. Two measurements of nodule attenuation value were obtained on each nodule at each imaging phase. Observers recorded mean attenuation values, and then analyzed and calculated the following dynamic characteristics of tumor enhancement using WI and WO contrast medium values: peak enhancement; net enhancement (WI); and absolute loss of enhancement (WO).345
Morphologic Evaluations
One radiologist (with 7 years of experience in chest CT scan interpretation), who did not take part in the dynamic study assessment, evaluated morphologic features, which included the following: margin type (eg, smooth, lobulated, or spiculated); and the presence or absence of a satellite lesion or cavity. The long-axis diameters of nodules were also measured on lung window images.
Data and Statistical Analysis
Retrospective calculations were performed to evaluate the usefulness of morphologic features alone (ie, lobulated or spiculated margin and absence of a satellite nodule) or net enhancement (ie, WI values, calculated for each level of enhancement by varying the positive test cutoff value) during early-phase dynamic CT scans as a marker for malignant nodules; and the sensitivity, specificity, accuracy, and positive and negative predictive values were obtained. The cutoff value of WO enhancement was calculated retrospectively at different levels of cutoff to differentiate malignant nodules from benign nodules during delayed-phase dynamic CT scanning.
Diagnostic characteristics were calculated in the following three ways: considering WI hemodynamic values only; considering WI and WO features; and considering WI values and morphologic characteristics in combination, using Bonferroni correction for multiple comparisons.8 A Mann-Whitney test was used to analyze statistical differences in patients’ ages and sizes for malignant and benign nodules. A p value of < 0.05 were regarded as being significant.
Results
Of 486 nodules, 237 (49%) proved to be malignant and 249 (51%) proved to be benign (Table 1
). There were 237 patients with malignant nodules (154 men and 83 women; mean age, 59 ± 11.5 years; age range, 24 to 85 years), whereas there were 249 patients with benign nodules (145 men and 104 women; mean age, 54 ± 11.5 years; age range, 22 to 82 years; p < 0.0001 [Mann-Whitney test]). The sizes of the 486 nodules were as follows: 5.5 mm but < 10 mm in diameter, 32 patients; 10 mm but < 15 mm, 90 patients; 15 mm but < 20 mm, 117 patients; 20 mm but < 25 mm, 115 patients; and 25 mm but 30 mm, 132 patients (mean diameter, 19.6 ± 6.4 mm; median diameter, 20.0 mm; diameter range, 5.5 to 30.0 mm). Malignant nodules (mean diameter, 21.6 ± 5.9 mm; median diameter, 22.4 mm; diameter range, 6.7 to 30.0 mm) were larger than benign nodules (mean diameter, 17.8 ± 6.3 mm; median diameter, 17.0 mm; diameter range, 5.5 to 30.0 mm; p < 0.0001 [Mann-Whitney test]).
|
2 test]) [Table 2
, Fig 1
]. When considering morphologic features only (ie, a lobulated or spiculated margin and the absence of a satellite nodule), sensitivity, specificity, accuracy, and positive and negative predictive values for malignancy were 94% (222 of 237 nodules), 65% (161 of 249 nodules), 79% (383 of 486 nodules), 72% (222 of 310 nodules), and 91% (161 of 176 nodules), respectively (Table 2, Figs 1234
).
|
|
|
|
|
25 HU indicating malignancy), the following diagnostic characteristics were obtained: sensitivity, 98% (233 of 237 nodules); specificity, 45% (114 of 249 nodules); accuracy, 71% (347 of 486 nodules); positive predictive value, 63% (233 of 368 nodules); and negative predictive value, 97% (114 of 118 nodules).
|
25 HU and WO of 5 to 36 HU provided sensitivity, specificity, accuracy, and positive and negative predictive values for malignant nodules of 89% (212 of 237 nodules), 79% (197 of 249 nodules), 84% (409 of 486 nodules), 80% (212 of 264 nodules), and 89% (197 of 222 nodules), respectively (Figs 1234).
|
25 HU and WO values of between 5 and 36 HU (Fig 2). Of 249 benign nodules, 113 had a WI of < 25 HU, 48 showed persistent enhancement without WO and with a WI of 25 HU, and 36 showed a WO of > 36 HU and a WI of 25 HU (Table 5
).
|
25 HU during the dynamic CT scan, a lobulated or spiculated margin, and the absence of a satellite lesion by morphologic evaluation, the corresponding values were 92% (219 of 237 nodules), 79% (197 of 249 nodules), 86% (416 of 486 nodules), 81% (219 of 271 nodules), and 92% (197 of 215 nodules), respectively (Figs 1234). Moreover, these diagnostic characteristics of sensitivity, specificity, and accuracy were not significantly different from those yielded from criteria of WI of > 25 HU and WO of 5 to 36 HU (Table 6
).
|
The diagnostic characteristics of a combined WI of 25 HU and a malignant morphology (ie, lobulated or spiculated margin and the absence of a satellite nodule) were slightly better (sensitivity, 92% [219 of 237 nodules]; specificity, 79% [197 of 249 nodules]; accuracy, 86% [416 of 486 nodules]), although not statistically so, than those of a WI of 25 HU and a WO of 5 to 36 HU (sensitivity, 89% [212 of 237 nodules]; specificity, 79% [197 of 249 nodules]; accuracy, 84% [409 of 486 nodules], respectively). In addition, the absolute WO values have not yet been standardized. Therefore, when characterizing a malignant nodule on dynamic CT scans, WO characteristics may not be needed for nodule evaluation, and the WO characteristics can be substituted for with morphologic features of a lobulated or spiculated margin and the absence of a satellite nodule.
Because nodule characterization using WI characteristics and morphologic features was found to provide high sensitivity and accuracy and to maintain specificity, delayed scans may not be needed. This would substantially increase patient throughput because early arterial phase images only would be necessary for nodule characterization. In addition, the radiation exposure to patients would be reduced. Moreover, a previous study3 disclosed that peak nodular enhancement occurs during a dynamic CT scan 40 to 180 s after IV contrast medium injection (malignant nodules: mean, 103 ± 43 s; median, 100 s; benign nodules: mean, 119 ± 45 s; median, 120 s). Therefore, a dynamic CT study, which is composed of images obtained at T0 (unenhanced images) and at T1, T2, and T3 min after IV contrast administration initiation, would allow the calculation of peak and net enhancement attenuation values. This simplified approach for dynamic studies would contribute to a further reduction in radiation exposure to the patient.
Methods should be devised to increase sensitivity. Because in our study almost all missed malignant nodules (17 of 18 false-negative nodules; 94%) satisfied one of two criteria (ie, a WI of 25 HU and a malignant morphology; lobulated or spiculated margin and the absence of a satellite nodule) on dynamic CT scans (Table 5), we may advise patients with nodules fulfilling only one of these two malignant criteria of WI 25 HU at dynamic study and malignant morphologic features to undergo more sensitive diagnostic studies (eg, 18F positron emission tomography scan, percutaneous needle biopsy, or a surgical biopsy using video-assisted thoracoscopic surgery).91011
The two-dimensional assessment methods of nodule attenuation and enhancement patterns are simple and practical; the problem is that the attenuation is strongly affected by the slice selected or the position of the ROI in the lesion, even on helical volumetric CT scan images. A computer-aided diagnosis system using contrast-enhanced dynamic helical CT scan data may permit the internal structure of nodules to be evaluated in a three-dimensional manner serially in a given time frame.12
Our study has several limitations. First, because repeated scans were performed on a targeted nodule, a relatively large amount of radiation was delivered to the thorax. In addition, when a nodule is located in the middle lung zone and on the same transaxial plane as the breast, the breast dose would be higher. Therefore, young woman with a pulmonary nodule in the middle lung zone may not be candidates for nodule evaluation.47 Second, pathologic proof was not obtained for all benign nodules. However, follow-up CT scans helped to diagnose benign nodules by showing no growth or a decrease in nodule size. Third, our study may contain a selection bias, because we excluded 159 patients with no histologic diagnosis or follow-up images. In addition, we excluded patients with a nodule containing benign-type calcification. However, it should be noted that up to 3.3% of nodules with benign-type calcification may turn out to be malignant.13 We also excluded patients with a predominantly ground-glass opacity or an extensively necrotic nodule. Patients with these ground-glass opacities or necrotic nodules were excluded because the nodules did not have a sufficient solid component where attenuation changes were measurable during the dynamic study. Our exclusion criteria, however, did not include surgical criteria or specific features favoring benign or malignant nodules from CT scan findings. Fourth, we neither attempted to standardize the injection rate or volume of contrast material according to the cardiac output for each patient (eg, SmartPrep; GE Medical Systems14 and saline solution chaser with a double-barrel injector15) nor to standardize the volume of contrast material according to patient weight.
In conclusion, the evaluation of solid or partly solid SPNs by analyzing combined WI values and morphologic features during HDCT scans provides 92% sensitivity and 79% specificity, which are probably equivalent to or slightly better than those obtained by analyses based on WI and WO HU values. This approach of analyzing combined WI values and morphologic features obviates the need for WO scans for nodule characterization, which saves time and reduces radiation exposure to the patient.
Footnotes
Abbreviations: HDCT = helical dynamic CT; HU = Hounsfield unit; ROI = region of interest; SPN = solitary pulmonary nodule; T = time; WI = wash-in; WO = washout
This article was presented as a scientific paper at the 2006 Radiological Society of North America Scientific Assembly.
This work was supported by a Korea Research Foundation grant (No. 2004-E00132) funded by Korean Government (Ministry of Education and Human Resources Development).
The authors have reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.
Received for publication October 17, 2006. Accepted for publication December 30, 2006.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
