Effect of Slab Thickness on the CT Detection of Pulmonary Nodules: Use of Sliding Thin-Slab Maximum Intensity Projection and Volume Rendering
Abstract
OBJECTIVE. The objective of this study was to evaluate the detection rates of pulmonary nodules on CT as a function of slab thickness using sliding thin-slab maximum intensity projection (MIP) and volume rendering (VR).
SUBJECTS AND METHODS. Eighty-eight oncology patients (33 women, 55 men; mean age, 59 years; age range, 18–81 years) who routinely underwent chest CT examinations were prospectively included. Two radiologists independently evaluated each CT examination for the presence of pulmonary nodules using MIP and VR, with each image reconstructed using three different slab thicknesses (5, 8, 11 mm). The standard of reference was the maximum number of detected nodules, which were classified by localization and size, judged to be true-positives by a consensus panel. Interreader agreement was assessed by kappa value on a nodule-by-nodule basis. Sensitivities for both reconstruction techniques and for the three slab thicknesses were calculated using the proportion procedure for survey data with the patient as the primary sample unit and were compared using the Wilcoxon's signed rank test with Bonferroni correction for both readers separately.
RESULTS. One thousand fifty-eight true-positive nodules were detected. Interreader agreement was fair to moderate. Sensitivity for pulmonary nodules was superior for 8-mm MIP (reader 1, 84%; reader 2, 81%) and was significantly better than the sensitivities of all other tested techniques for both readers (p < 0.001 each) independent of nodule localization and size (except for one reader's analysis of 8-mm MIP versus 11-mm MIP for nodules > 8 mm). A higher sensitivity was achieved using MIP than VR.
CONCLUSION. MIP with a slab thickness of 8 mm is superior in the detection of pulmonary nodules to all other tested techniques.
Introduction
The introduction of MDCT scanners and the ability to reduce slice thickness to 1 mm or less led to an increase in the number of images per CT examination [1]. Consequently, CT interpretation became more time consuming, and the demand for alternative image reading techniques emerged to reliably assess the increasing amount of data in a reasonable amount of time.
In this context, the detection of pulmonary nodules is one example of a repetitive process performed multiple times during the daily routine of a radiologist, not only in the dedicated clinical context of detecting primary tumors or metastases, but also for every chest CT study performed for other purposes, such as pulmonary embolism and so on.
Viewing image stacks of CT examinations at a workstation with a continuous sequence (cine viewing) instead of reading single images on hard copies was an early major technologic advance [2]. However, application of postprocessing techniques, such as maximum intensity projection (MIP) or volume rendering (VR), has been shown to improve detection rates of pulmonary nodules in comparison with cine viewing of nonpostprocessed axial images [3–6].
In the MIP algorithm, only the highest-attenuation voxels along lines projected through the volume data set are selected to create the final image [7].
VR is a 3D reconstruction method in which the final image contains data from each voxel. Each voxel, in turn, represents proportions of the enclosed different types of tissue (percentage classification) [7].
One important parameter the radiologist is able to adjust in both viewing techniques is the thickness of the sliding thin slabs.
The purpose of this study was to compare the detection rates for pulmonary nodules of MIP and VR depending on the thickness of the sliding thin slabs for three different slab thicknesses: 5, 8, and 11 mm.
Subjects and Methods
Patient Population
One hundred eighteen consecutive patients with a known malignancy undergoing a chest CT examination for clinical reasons between July 9 and August 10, 2007, were included in this prospective study, which was approved by the ethics committee. Exclusion criteria were age of < 18 years, > 60 nodules in both lungs (n = 5), coexisting lung disease (e.g., consolidations, interstitial lung disease) (n = 19), and refusal to attend (n = 6). Eighty-eight patients (33 women, 55 men; mean age ± SD, 59 ± 14 years; age range, 18–81 years) met the inclusion criteria.
CT Parameters
CT studies were performed on a 16-MDCT scanner (Aquilion 16, Toshiba Medical). Tube current was adjusted automatically using an automatic dose adjustment technique (RealEC, Toshiba Medical) to achieve an SD of 7.5 HU. The remaining acquisition parameters were identical in all patients: tube potential, 120 kV; collimation, 16 × 1 mm; section thickness, 1 mm; pitch, 1.4; gantry rotation, 0.5 second; reconstruction kernel, FC51 (high resolution); and reconstruction increment, 0.6 mm.
IV contrast agent (120 mL of iohexol [Accupaque 300, Amersham Health]; iodine concentration, 300 mg/mL) was administered at a flow rate of 3 mL/s using a bolus-tracking technique (SureStart, Toshiba Medical). Scanning was started when the CT attenuation in the ascending aorta measured 120 HU.
Image Interpretation
Independent evaluations of 5-, 8-, and 11-mm MIP and VR images were performed on a workstation (Vitrea 2, version 3.9.0.1, Vital Images) by one resident with 3 years of experience (reader 1) and one board-certified chest radiologist with 12 years of experience (reader 2). The MIP and VR images were reconstructed interactively (i.e., sliding thin-slab technique). Several window settings for MIP and VR were tested in advance to get the most suitable settings for the purpose of lung nodule detection. On the basis of these preliminary studies, MIP images were displayed with a window center of –300 HU and a window width of 1,600 HU and VR images with a center of –500 HU and a width of 1,500 HU. The opacity settings for VR were obtained from a study by Peloschek et al. [4]: linear ramp with 100% opacity at 200 HU and 50% opacity at –400 HU.
Every CT examination was read six times using the following techniques: MIP with a slab thickness of 5, 8, and 11 mm and VR with a slab thickness of 5, 8, and 11 mm. The slab thicknesses of 5, 8, and 11 mm were chosen on the basis of recently published articles [3–5, 8]. The decision was also influenced by our experience obtained during daily work. Slabs larger than 11 mm are confusing because of the overlay of a multitude of pulmonary vessels. Slabs smaller than 5 mm seem to be too small to differentiate longitudinal structures, such as vessels and pulmonary nodules, well enough. Image readings for the same patient with different techniques were performed with a time delay of at least 2 weeks between reading sessions to avoid memory effects. Assessment consisted of detection of pulmonary nodules as defined by the Fleischner Society's recommendations [9] supplemented by newer classifications including focal nodular areas of ground-glass attenuation [10–12].
To standardize the reading sessions, we instructed radiologists that they were not allowed to change any parameter such as window settings or slice thickness and that they should read the CT examinations as quickly as they would in clinical routine and with a uniform reading pattern. For this purpose, both lungs were virtually divided into a ventral part and a dorsal part separated by an imaginary line proceeding in the coronal direction through the center of both lungs. Reading was performed by starting with the ventral part of the right lung followed by the dorsal part of the right lung and then moving to the ventral part followed by the dorsal part of the left lung, while adhering to the caveat that each part must be viewed from the lung apex to the lung base. During the preparatory stage of the study, the readers jointly assessed several CT chest examinations, which were not part of the study to train in the use of the reading pattern and to standardize the definition of a pulmonary nodule.
During independent image evaluation, every pulmonary nodule was marked by an arrow and documented on a printout and by archiving the nodule image in PACS.
A consensus panel consisting of both readers chaired by a third radiologist with 28 years of experience analyzed the results of both readers and classified each detected nodule as true- or false-positive using a combination of MIP, VR, the original 1-mm axial images, and multiplanar reconstructions in two additional imaging planes (sagittal and coronal).
The maximum diameter of each nodule was measured and classified according to the Fleischner Society's guidelines [13] into the following categories: ≤ 4 mm, > 4–8 mm, and > 8 mm. The categories > 4–6 mm and > 6–8 mm were merged into a > 4–8 mm category. Depending on their localizations, nodules were grouped as hilar, located within 2 cm around the hilum; pleural, with pleural contact; and central, between hilar and pleural.
Statistical Analysis
Because of the lack of a true reference standard, the maximum number of intrapulmonary nodules detected by both readers and classified as true-positive by the consensus panel was considered the standard of reference. Statistical analysis was performed using SPSS software (version 13.0, SPSS) and Stata software (version 10.0, StataCorp) for Macintosh (Apple Computer).
Interreader Agreement
To assess interreader agreement, kappa values were determined for all nodules on a nodule-by-nodule basis. Agreement was classified using a grading system for agreement measures for categoric data established by Landis and Koch [14]: poor agreement, κ < 0.00; slight, κ = 0.00–0.20; fair, κ = 0.21–0.40; moderate, κ = 0.41–0.60; good, κ = 0.61–0.80; and excellent, κ = 0.81–1.00.
The following statistical analyses were performed separately for both readers.
Sensitivity
Sensitivities including 95% CIs for each postprocessing technique using the three different slab thicknesses were assessed using the proportion procedure for survey data used by the Stata software with the patient as the primary sample unit.
Statistical Significance
The sensitivities of the six methods were compared for each reader on a patient basis by analyzing the numbers of true findings per patient using the Wilcoxon's signed rank test. A p value of < 0.0033 was considered significant after Bonferroni correction for multiple tests.
The same analysis was performed after grouping nodules in the mentioned size and location categories.
Results
A total of 1,132 nodules were detected, of which 74 (6.5%) were scored by the consensus panel as false-positives. The remaining 1,058 true-positive nodules were grouped according to their localizations and sizes: 205 nodules (19.4%) had pleural contact, 84 (7.9%) were at the hilum, and 769 (72.7%) were centrally located. Seven hundred thirty-five nodules (69.5%) were ≤ 4 mm; 225 (21.3%), > 4–8 mm; and 98 (9.3%), > 8 mm. The mean number of true-positive nodules per patient (± SD) was 12 ± 10.4 (range, 0–57 nodules). With each technique, the maximum number of false-positive nodules per patient for each reader was two or fewer with no significant difference.
Interreader Agreement
The interreader agreement was fair for 5-mm MIP (κ = 0.32) and 8-mm MIP (κ = 0.22) and moderate for 11-mm MIP (κ = 0.44), 5-mm VR (κ = 0.47), 8-mm VR (κ = 0.53), and 11-mm VR (κ = 0.57). The interreader agreement within the size and location categories was similar.
Nodule Detection Rate
An analysis of the sensitivities using the proportion procedure for survey data of the Stata software with the patient as the primary sample unit resulted in the performance data shown in Table 1. Nodule detection sensitivity was superior with 8-mm MIP in comparison with all the other techniques for both readers. Sensitivities for 5-mm MIP and 11-mm MIP were similar for both readers (reader 1, 61% and 58%; reader 2, 55% and 55%, respectively). MIP had a higher sensitivity than VR at all slab thicknesses. The results of both readers for 5-mm VR, 8-mm VR, and 11-mm VR did not show major differences.
| Sensitivity in % (No. of Nodules) [95% CI] | ||
|---|---|---|
| Technique | Reader 1 | Reader 2 |
| MIP | ||
| 5 mm | 61 (641) [56–65] | 55 (578) [50–59] |
| 8 mm | 84 (893) [81–88] | 81 (857) [77–85] |
| 11 mm | 58 (612) [52–63] | 55 (581) [50–59] |
| VR | ||
| 5 mm | 41 (435) [36–47] | 41 (432) [37–45] |
| 8 mm | 40 (427) [34–46] | 42 (441) [37–46] |
| 11 mm | 40 (422) [34–46] | 41 (433) [37–45] |
Note—Total number of true-positive nodules = 1,058
In Tables 2 and 3, the sensitivities after grouping nodules according to their maximum diameters and localizations are shown. The highest detection rate for both readers was seen for 8-mm MIP independent of nodule size or localization.
| Sensitivity in % (No. of Nodules) [95% CI] | ||||||
|---|---|---|---|---|---|---|
| Reader 1 | Reader 2 | |||||
| Technique | Pleural | Hilar | Central | Pleural | Hilar | Central |
| MIP | ||||||
| 5 mm | 70 (143) [63–77] | 64 (54) [53–75] | 58 (444) [53–63] | 62 (127) [53–71] | 51 (43) [37–65] | 53 (408) [48–58] |
| 8 mm | 88 (180) [83–93] | 90 (76) [84–97] | 83 (637) [79–87] | 82 (169) [75–90] | 80 (67) [69–90] | 81 (621) [76–85] |
| 11 mm | 64 (132) [55–74] | 62 (52) [49–75] | 56 (428) [50–62] | 63 (129) [54–72] | 51 (43) [41–62] | 53 (409) [49–58] |
| VR | ||||||
| 5 mm | 57 (117) [46–68] | 40 (34) [28–53] | 37 (284) [31–43] | 48 (99) [40–56] | 49 (41) [39–59] | 38 (292) [33–43] |
| 8 mm | 45 (93) [35–55] | 40 (34) [27–54] | 39 (300) [33–45] | 44 (91) [35–53] | 54 (45) [42–65] | 40 (305) [35–45] |
| 11 mm | 45 (93) [37–54] | 51 (43) [38–64] | 37 (286) [30–44] | 43 (89) [34–53] | 57 (48) [46–68] | 38 (296) [34–43] |
Note—Total number of true-positive nodules = 1,058: pleural nodules, n = 205; hilar nodules, n = 84; and central nodules, n = 769
| Sensitivity in % (No. of Nodules) [95% CI] | ||||||
|---|---|---|---|---|---|---|
| Reader 1 | Reader 2 | |||||
| Technique | ≤ 4 mm | > 4 to 8 mm | > 8 mm | ≤ 4 mm | > 4 to 8 mm | > 8 mm |
| MIP | ||||||
| 5 mm | 54 (397) [49–59] | 72 (163) [66–79] | 83 (81) [75–91] | 47 (347) [42–52] | 68 (152) [61–74] | 81 (79) [72–89] |
| 8 mm | 79 (581) [75–83] | 95 (214) [92–99] | 100 (98) | 78 (574) [74–82] | 86 (193) [80–91] | 92 (90) [84–100] |
| 11 mm | 47 (344) [41–52] | 80 (180) [75–85] | 90 (88) [84–95] | 47 (346) [43–51] | 71 (160) [64–79] | 77 (75) [67–87] |
| VR | ||||||
| 5 mm | 28 (205) [23–33] | 66 (148) [58–73] | 84 (82) [76–92] | 29 (216) [26–33] | 63 (141) [55–70] | 77 (75) [68–85] |
| 8 mm | 29 (210) [24–33] | 62 (140) [51–73] | 79 (77) [68–89] | 30 (224) [27–34] | 62 (140) [54–70] | 79 (77) [71–87] |
| 11 mm | 25 (183) [21–29] | 69 (155) [60–78] | 86 (84) [78–93] | 29 (216) [26–33] | 63 (141) [53–73] | 78 (76) [69–86] |
Note—Total number of true-positive nodules = 1,058: nodules ≤ 4 mm, n = 735; nodules > 4–8 mm, n = 225; and nodules > 8 mm, n = 98
Statistical Significance of Differences in Detection Rate
A comparison of the sensitivities among the different techniques showed that the detection rate of 8-mm MIP was significantly better for both readers than all other postprocessing techniques (Table 4). There was no significant difference between 5-mm MIP and 11-mm MIP for both readers. The detection rate between 5-mm VR, 8-mm VR, and 11-mm VR showed no significant differences for both readers.
| MIP | VR | |||||
|---|---|---|---|---|---|---|
| Technique | 5 mm | 8 mm | 11 mm | 5 mm | 8 mm | 11 mm |
| MIP | ||||||
| 5 mm | < 0.001a | 0.2407 | < 0.001a | < 0.001a | < 0.001a | |
| 8 mm | < 0.001 a | < 0.001a | < 0.001a | < 0.001a | < 0.001a | |
| 11 mm | 0.5066 | < 0.001 a | < 0.001a | < 0.001a | < 0.001a | |
| VR | ||||||
| 5 mm | < 0.001 a | < 0.001 a | < 0.001 a | 0.398 | 0.904 | |
| 8 mm | < 0.001 a | < 0.001 a | < 0.001 a | 0.421 | 0.382 | |
| 11 mm | < 0.001a | < 0.001a | < 0.001a | 0.881 | 0.607 | |
Note—The p values for reader 1 are shown in lightface and the p values for reader 2 are shown in boldface. MIP = maximum intensity projection, VR = volume rendering
a
Statistically significant in the following order: 8-mm MIP better than (5-mm MIP, 11-mm MIP) better than (5-mm VR, 8-mm VR, 11-mm VR); sensitivity for each technique is shown in Table 1
For both readers, nodule localization did not affect the detection rate when using 8-mm MIP because this method proved to be significantly superior to all other techniques for pleural, hilar, and central nodule detection (p ≤ 0.002). For both readers, 5-mm MIP and 11-mm MIP had significantly better detection rates than VR at all slab thicknesses for nodules with central localizations (p < 0.001). For reader 2, 11-mm MIP was significantly superior to VR at all slab thicknesses for nodules in pleural and central localizations (p ≤ 0.001).
An analysis of the influence of nodule size on the detection rate showed that 8-mm MIP was significantly superior to all other techniques for both readers (p ≤ 0.003) except for the comparison of 8-mm MIP and 11-mm MIP for nodules > 8 mm for reader 1 (p = 0.004). Detection of nodules ≤ 4 mm was significantly better with 5-mm MIP and 11-mm MIP than 5-mm VR, 8-mm VR, and 11-mm VR for both readers (p < 0.001).
Discussion
In several previous studies, investigators have shown that the detection rates for lung nodules using postprocessing techniques such as MIP and VR are superior to those achieved through conventional transverse section viewing [3–5, 15]. Diederich et al. [16] compared, among others, 15-mm MIP and 30-mm MIP and found 15-mm MIP to be slightly superior. However, MIP images were reconstructed from CT data sets with 5- and 10-mm collimation. Today, a narrower collimation is usually used, which may influence detection rates. Peloschek et al. [4] compared VR and MIP for only one fixed slab thickness of 7 mm, and Yoneda et al. [6] evaluated VR and MIP with a slab thickness of 15 mm. Gruden et al. [3] compared 10-mm MIP with 3.75-mm axial images because the results of their prior investigations [3] gave rise to the assumption that 10-mm MIP is superior to 30-mm and 5-mm MIP.
To the best of our knowledge, a comprehensive analysis of the influence of slab thickness on the effectiveness of pulmonary nodule detection using VR and MIP has not been conducted.
In several studies, investigators are even one step ahead by comparing the sensitivities for lung nodules of computer-assisted detection software with that of the radiologist without an optimized detection technique [5, 17–23]. Most of the studies mentioned used conventional axial images with a slice thickness of between 0.75 and 7.5 mm as a standard of reference for nodule detection by the radiologist [17, 19–21, 23]. In one study, MIP postprocessing was used, but the slab thickness was not mentioned [24], whereas in other studies, MIP postprocessing with only one slab thickness of 5 mm [8] or 6 mm [5], respectively, was applied without having optimized the technique in advance.
In the current study, similar source data (collimation, 16 × 1 mm; section thickness, 1 mm) were used for reconstruction of the MIP and VR images in comparison with recent studies in which source data had a collimation ranging from 16 × 0.75 mm to 16 × 1.5 mm and a section thickness of 1 or 2 mm, respectively [4, 5, 8].
The aim of our study was to optimize manual postprocessing-aided detection of pulmonary nodules using MIP and VR as a function of slab thickness.
Our results show that the nodule detection rate using MIP is superior to that using VR independent of slab thickness. Furthermore, the detection rate using 8-mm MIP is significantly higher than all other tested techniques and slab thicknesses. Statistical results suggest a grading system with 8-mm MIP as the highest-ranking technique; 5-mm MIP and 11-mm MIP as the second; and 5-mm VR, 8-mm VR, and 11-mm VR together as third-rank techniques.
Yoneda et al. [6] and Das et al. [8] found the effectiveness of MIP images for pulmonary nodule detection, when assessed using receiver operating characteristic curves, to be significantly superior to detection using VR images. In their study, Peloschek et al. [4] drew a different conclusion: They found VR to be superior to MIP for a single slab thickness of 7 mm. There are several possible reasons for that result. First, VR images can be modified by the radiologist not only by changing the slab thickness but also by modifying other parameters such as color, opacity, and window settings [7]. We used the same opacity curves as Peloschek et al. in their study. Our experience suggests that the window settings may have considerable influence on detection rate. During the preparatory stage of our study, we tested several window settings and chose settings for MIP and VR that seemed to be most suitable for pulmonary nodule detection. In other studies MIP images were displayed using window levels between –700 and –530 HU and with ranges of window widths from 1,200 to 1,600 HU [3, 4, 15, 16, 25, 26]. In the only two other studies investigating MIP and VR, window settings were not indicated [4, 6], and in one study [4], the software did not allow the adjustment of window settings when using VR (e-mail communication with Peloschek P, May 2007).
Second, each vendor has its own VR algorithm that leads to differences in image presentation that may affect nodule contrast and detection rate [7].
In comparison with VR, MIP is less vulnerable because of a lower number of adjustable postprocessing parameters in case the direction of the projecting rays is parallel to the original data axes. In the current study, only axial postprocessed images were assessed. Difficulties arise only when the rays are cast at angles that are not parallel to the original data axes because nonparallel rays will not pass through the centers of all the pixels. As a result, the image must be resampled at the points of interest along the ray [27], and such resampling algorithms are vendor-specific. In case of axial MIP, slab thickness and window settings are the only adjustable parameters [7].
By contrast, VR displays 3D relationships between different structures such as nodules and vessels better than MIP. However, given our results, this capability does not seem to play an important role in pulmonary nodule detection.
Pulmonary nodule localization did not affect the detection rate of 8-mm MIP, which was superior to all other tested techniques for the detection of hilar, pleural, and central nodules. Peloschek et al. [4] found VR to be superior to MIP especially in the detection of perihilar nodules.
Nodule size did not affect detection rate of 8-mm MIP in the current study. Other studies comparing 5-mm MIP [15] and 10-mm MIP [25] versus conventional axial imaging have shown MIP to perform better only for the detection of nodules < 5 mm. In the previously mentioned study comparing MIP and VR [4], investigators found VR to be significantly better than MIP in the detection of nodules < 10 mm.
The fair to moderate interreader agreement may be interpreted as one of the limitations of the current study. Other studies have shown a fair to moderate interreader agreement as well, at least between some of the readers [5, 8, 22]. Strict definition of the window settings, reading pattern, and a training phase at the beginning of the reading sessions were aimed to eliminate factors impairing interreader agreement. One possible reason for the fair to moderate agreement might be the instruction to assess the CT images at the same speed used in clinical routine. Gruden et al. [3] limited the reading time to a maximum of 2 minutes, but the maximum number of nodules per patient was nine in their study, whereas it was 57 in our study. Especially in case of a large number of nodules, a limited time frame may diminish the nodule detection rate and cause a decrease in interreader agreement. However, although achieving almost the same sensitivity (Table 1), the two readers overlooked a certain number of nodules. Overlooking different nodules caused a decrease in interreader agreement. The readers, therefore, worked equally well with respect to nodule detection, but they detected different nodules. This is, however, not a fault of poor study design or poor reader performance. The moderate interreader agreement is caused by a combination of limited sensitivity of the detection techniques and of the random nature of overlooking different nodules.
The fact that one of the readers was a resident could be a limiting factor of the study. However, in the current study, the resident and the chest radiologist gained similar sensitivities (Table 1) in nodule detection, showing a robust behavior of the technique with regard to professional experience. In daily routine, this is an important and favorable characteristic of an image assessment technique especially when used in teaching hospitals.
Another limiting factor is the lack of a true standard of reference shared with all other clinical studies involved with pulmonary nodule detection. The reference standard for this study consisted of the maximum number of intrapulmonary nodules detected by both readers and classified as true-positive by the consensus panel. Only experimental studies with artificial creation of nodules in laboratory animals are able to provide a true standard of reference [28].
In conclusion, MIP with a thickness of the sliding thin slab of 8 mm has a significantly higher sensitivity for nodule detection in chest CT than all other tested techniques: 5-mm MIP, 11-mm MIP, 5-mm VR, 8-mm VR, and 11-mm VR.
Footnote
Address correspondence to N. Kawel ([email protected]).
References
1.
Rubin GD. Data explosion: the challenge of multidetector-row CT. Eur J Radiol 2000; 36:74-80
2.
Seltzer SE, Judy PF, Adams DF, et al. Spiral CT of the chest: comparison of cine and film-based viewing. Radiology 1995; 197:73-78
3.
Gruden JF, Ouanounou S, Tigges S, Norris SD, Klausner TS. Incremental benefit of maximum-intensity-projection images on observer detection of small pulmonary nodules revealed by multidetector CT. AJR 2002; 179:149-157
4.
Peloschek P, Sailer J, Weber M, Herold CJ, Prokop M, Schaefer-Prokop C. Pulmonary nodules: sensitivity of maximum intensity projection versus that of volume rendering of 3D multidetector CT data. Radiology 2007; 243:561-569
5.
Jankowski A, Martinelli T, Timsit JF, et al. Pulmonary nodule detection on MDCT images: evaluation of diagnostic performance using thin axial images, maximum intensity projections, and computer-assisted detection. Eur Radiol 2007; 17:3148-3156
6.
Yoneda K, Ueno J, Nishihara S, et al. Postprocessing technique with MDCT data improves the accuracy of the detection of lung nodules. Radiat Med 2007; 25:511-515
7.
Fishman EK, Ney DR, Heath DG, Corl FM, Horton KM, Johnson PT. Volume rendering versus maximum intensity projection in CT angiography: what works best, when, and why. RadioGraphics 2006; 26:905-922
8.
Das M, Muhlenbruch G, Mahnken AH, et al. Small pulmonary nodules: effect of two computer-aided detection systems on radiologist performance. Radiology 2006; 241:564-571
9.
Austin JH, Müller NL, Friedman PJ, et al. Glossary of terms for CT of the lungs: recommendations of the Nomenclature Committee of the Fleischner Society. Radiology 1996; 200:327-331
10.
Beigelman-Aubry C, Hill C, Grenier PA. Management of an incidentally discovered pulmonary nodule. Eur Radiol 2007; 17:449-466
11.
Munden RF, Hess KR. “Ditzels” on chest CT: survey of members of the Society of Thoracic Radiology. AJR 2001; 176:1363-1369
12.
Yang ZG, Sone S, Takashima S, et al. High-resolution CT analysis of small peripheral lung adenocarcinomas revealed on screening helical CT. AJR 2001; 176:1399-1407
13.
MacMahon H, Austin JH, Gamsu G, et al. Guidelines for management of small pulmonary nodules detected on CT scans: a statement from the Fleischner Society. Radiology 2005; 237:395-400
14.
Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33:159-174
15.
Eibel R, Turk TR, Kulinna C, Herrmann K, Reiser MF. Multidetector-row CT of the lungs: multiplanar reconstructions and maximum intensity projections for the detection of pulmonary nodules [in German]. Rofo 2001; 173:815-821
16.
Diederich S, Lentschig MG, Overbeck TR, Wormanns D, Heindel W. Detection of pulmonary nodules at spiral CT: comparison of maximum intensity projection sliding slabs and single-image reporting. Eur Radiol 2001; 11:1345-1350
17.
Awai K, Murao K, Ozawa A, et al. Pulmonary nodules at chest CT: effect of computer-aided diagnosis on radiologists' detection performance. Radiology 2004; 230:347-352
18.
Brown MS, Goldin JG, Rogers S, et al. Computer-aided lung nodule detection in CT: results of large-scale observer test. Acad Radiol 2005; 12:681-686
19.
Brown MS, Goldin JG, Suh RD, McNitt-Gray MF, Sayre JW, Aberle DR. Lung micronodules: automated method for detection at thin-section CT— initial experience. Radiology 2003; 226:256-262
20.
Lee JW, Goo JM, Lee HJ, Kim JH, Kim S, Kim YT. The potential contribution of a computer-aided detection system for lung nodule detection in multidetector row computed tomography. Invest Radiol 2004; 39:649-655
21.
Marten K, Seyfarth T, Auer F, et al. Computer-assisted detection of pulmonary nodules: performance evaluation of an expert knowledge-based detection system in consensus reading with experienced and inexperienced chest radiologists. Eur Radiol 2004; 14:1930-1938
22.
Rubin GD, Lyo JK, Paik DS, et al. Pulmonary nodules on multi-detector row CT scans: performance comparison of radiologists and computer-aided detection. Radiology 2005; 234:274-283
23.
Yuan R, Vos PM, Cooperberg PL. Computer-aided detection in screening CT for pulmonary nodules. AJR 2006; 186:1280-1287
24.
Wormanns D, Beyer F, Diederich S, Ludwig K, Heindel W. Diagnostic performance of a commercially available computer-aided diagnosis system for automatic detection of pulmonary nodules: comparison with single and double reading. Rofo 2004; 176:953-958
25.
Valencia R, Denecke T, Lehmkuhl L, Fischbach F, Felix R, Knollmann F. Value of axial and coronal maximum intensity projection (MIP) images in the detection of pulmonary nodules by multislice spiral CT: comparison with axial 1-mm and 5-mm slices. Eur Radiol 2006; 16:325-332
26.
Remy-Jardin M, Remy J, Artaud D, Deschildre F, Duhamel A. Diffuse infiltrative lung disease: clinical value of sliding-thin-slab maximum intensity projection CT scans in the detection of mild micronodular patterns. Radiology 1996; 200:333-339
27.
Schreiner S, Paschal CB, Galloway RL. Comparison of projection algorithms used for the construction of maximum intensity projection images. J Comput Assist Tomogr 1996; 20:56-67
28.
Coakley FV, Cohen MD, Johnson MS, Gonin R, Hanna MP. Maximum intensity projection images in the detection of simulated pulmonary nodules by spiral CT. Br J Radiol 1998; 71:135-140
Information & Authors
Information
Published In
Copyright
© American Roentgen Ray Society.
History
Submitted: August 16, 2008
Accepted: October 9, 2008
Keywords
Authors
Metrics & Citations
Metrics
Citations
Export Citations
To download the citation to this article, select your reference manager software.