Systematic Error in Lung Nodule Volumetry: Effect of Iterative Reconstruction Versus Filtered Back Projection at Different CT Parameters
Abstract
OBJECTIVE. Iterative reconstruction potentially can reduce radiation dose compared with filtered back projection (FBP) for chest CT. This is especially important for repeated CT scanning, as is the case in patients with indeterminate lung nodules. It is currently unknown whether absolute nodule volumes measured with iterative reconstruction are comparable to those measured with FBP. We compared nodule volumes measured with iterative reconstruction and FBP at different CT parameters.
MATERIALS AND METHODS. An anthropomorphic chest phantom was scanned using a 256-MDCT scanner at various tube voltages and tube current–time products. Raw data were reconstructed using FBP or a commercially available iterative reconstruction algorithm. Five inserted nodules with 100 HU radiodensity and different sizes (3, 5, 8, 10, and 12 mm) were measured by two observers using semiautomatic software. Volumetric nodule measurements were performed using thin-slice reconstructions.
RESULTS. For very small nodules (volume, 14.1 mm3; diameter, 3 mm), FBP and iterative reconstruction measurements exhibited large errors and overestimated the nodule size by up to 160%. For larger nodules (volume, ≥ 65.4 mm3; diameter, ≥ 5 mm), CT underestimated the actual size, but errors were small (within 25%) and remained small when the tube voltage and tube current–time product were reduced, even without iterative reconstruction.
CONCLUSION. In a phantom model, no clinically relevant differences beyond reported interscan variation levels between lung nodule volumes were measured in nodules 5 mm or larger at reduced tube voltage and tube current–time product, with radiation dose reductions up to 90.6% for both FBP and iterative reconstruction, suggesting that it is safe to convert FBP protocols to iterative reconstruction and reduce tube voltage and tube current–time product for lung nodule follow-up. CT appears to slightly underestimate actual nodule volume.
Lung nodules are a common incidental finding on chest CT [1], but differentiation between benign and malignant nodules can be difficult. Although most small solitary lung nodules have a benign cause [2], typically benign characteristics (e.g., smooth edges, presence of fat, calcification, small size, and perifissural location) are not always present on CT [1, 3–5]. Because nodule size and growth are strong predictors for malignancy, the accurate assessment of size at baseline and growth on follow-up CT is important in the diagnostic workup of nodules that reach a certain size [6]. According to the Fleischner Society’s recommendations [1], a nodule of larger than 4 mm average diameter can be regarded as benign if no growth is observed after 24 months on low-dose unenhanced CT. Nodule growth can be quantified using either diameter or volume. Recent studies point to an increasingly important role for volumetric measurements, because malignant nodules may grow asymmetrically, and, therefore, their growth may remain unnoticed with manual diameter measurements [2]. Furthermore, manual 2D measurements of small nodules have modest repeatability [7].
Currently, CT scans for the measurement of nodule size and growth, at baseline and on follow-up, use the conventional filtered back projection (FBP) algorithm for image reconstruction [8]. Recent advances in CT technology and increased computational power have permitted the use of iterative reconstruction algorithms for CT image reconstruction [8]. Iterative reconstruction can decrease radiation dose by 50% or more without an increase in image noise and has the potential for significant dose reduction without a loss of diagnostic value [9]. Radiation dose saving is especially important in patients with lung nodules because of the frequent follow-up examinations. Despite the wide range of research in CT measurements of pulmonary nodules [8, 10–14], it is unknown whether nodule volume measurements using FBP are different from volume measurements using iterative reconstruction at different tube voltages and tube current–time products.
Therefore, the aim of our study was to systematically compare volumes of lung nodules measured with iterative reconstruction and FBP at different CT parameters in an anthropomorphic chest phantom.
Materials and Methods
Phantom
A commercially available anthropomorphic chest phantom (PH-1, Kyoto Kagaku) with five manually inserted spherical lung nodules (volume, 14.1, 65.4, 268.1, 523.6, and 904.8 mm3, respectively; diameter, 3, 5, 8, 10, and 12 mm, respectively) was used. The nodules were supplied by the manufacturer of the phantom, and they all had a radiodensity of 100 HU (Table 1). The phantom was an accurate life-sized anatomic human torso model with dimensions of 43 × 40 × 48 cm. The inner components consisted of mediastinum, pulmonary vasculature, spine, ribs, and abdomen block (Fig. 1).
CT Protocol
Scanning was performed on a 256-MDCT scanner (Brilliance iCT, Philips Healthcare). The following parameters were used: detector collimation, 128 × 0.625 mm; slice thickness, 0.9 mm; gantry rotation time, 0.33 s; pitch, 0.76; and matrix size, 512 × 512 pixels. Tube voltages of 80, 100, and 120 kVp were used, whereas tube current–time products varied among 25, 50, and 100 mAs. The different acquisition protocols are listed in Table 2. The phantom was scanned once with each acquisition protocol. All CT scans were performed sequentially on a single day without changing the positions of either the phantom itself or the nodules within it. Volume CT dose index was recorded for each scan. Dose reduction was calculated according to normal dose parameters of 120 kVp and 100 mAs.
Image Reconstruction
Raw data were reconstructed using standard FBP or an iterative reconstruction algorithm (iDose, Philips Healthcare). iDose is a recently developed reconstruction algorithm consisting of two denoising components [15, 16]. First, a Poisson noise distribution–based maximum likelihood denoising algorithm is applied. Second, reconstructed images are iteratively adjusted to decrease uncorrelated noise. Thus, iDose provides reduction of noise. The level of noise reduction is adjustable by selecting one of seven levels (higher levels have more iterations and noise reduction). In this study iDose levels 2, 4, and 6 were used. Theoretically, levels 2, 4, and 6 result in 16%, 29%, and 45% noise reduction, respectively. In addition, spatial domain sharpness filter C was used [17], which is used commonly in our hospital for lung imaging.
Fig. 1 —Axial CT images of phantom nodule 2 (diameter, 5 mm; volume, 65.4 mm3) in lung setting.
A–D, Images of nodule (arrow) were reconstructed using filtered back projection (A), iDose (Philips Healthcare) level 2 (B), iDose level 4 (C), and iDose level 6 (D).
Volumetry Protocol
The volumes of the five lung nodules were measured on a CT workstation using commercially available semiautomatic software (IntelliSpace, Philips Healthcare). All volumetric nodule measurements were performed using 0.9-mm slice thickness reconstructions. By placing a cursor on the nodule of interest and clicking with the mouse, the user instructed the software to automatically delineate the nodule and quantified its volume. Because the automatic measurement function of the program was adequate in all measurements, as judged by visual inspection of the resulting segmentation, no manual adjustments had to be made. All measurements were performed by two independent observers who both measured all nodules.
Data Analysis
de Hoop et al. [11] found that the minimum change needed to detect growth of solid lung nodules is 18.5–25.6% for nodules smaller than 8 mm in diameter and 12.9–17.1% for nodules greater than 8 mm in diameter. Therefore, a 25% or more difference in volume was used to define a clinically relevant nodule volume difference in this study. Interobserver variability was assessed visually, and because only one nodule had a different volume between observers, no statistical test was applied. Comparison of relative differences in nodule volumes between the CT protocols was done graphically.
Results
Dose reduction percentages and volume CT dose index values are listed in Table 2.
Nodule Volumes at Various Levels of Tube Current–Time Product and Tube Voltage
Differences between measured and actual lung nodule volumes with FBP reconstruction and various tube voltages and tube current–time products are displayed in Figure 2 and listed in Table 3. Measurements of the smallest nodule, nodule 1 (volume, 14.1 mm3; diameter, 3 mm), showed volume overestimations, ranging from 26.2% to 159.6%. Except for one case of overestimation of 13.0% in nodule 2 (100 mAs, 80 kVp, and 1.95 mGy), the other nodules were underestimated, ranging from –0.9% to –23.9%.
Fig. 2 —Differences between actual and measured nodule volumes using filtered back projection reconstruction.
A, Tube voltage is 120 kVp, and tube current–time product is variable.
B, Tube current–time product is 100 mAs, and tube voltage is variable.
For nodule 1, greater tube current–time product values resulted in more accurate volume measurements, but this did not apply for the larger nodules. Compared with measurements with tube voltage values of 80 and 120 kVp, volume differences with 100 kVp (3.95 mGy) were greatest for all but nodule 4 (volume, 523.6 mm3; diameter, 10 mm). In general, all errors in volume estimation of nodules 5 mm or larger (volume, ≥ 65.4 mm3) remained within 25%. We did not observe larger errors at lower dose with dose reductions up to 90.6% for any but the smallest nodule.
Nodule Volumes at FBP and Various iDose Levels
Differences between measured and actual lung nodule volumes at 80 and 120 kVp tube voltage, tube current–time product of 25 mAs, and varying reconstruction techniques (FBP and iDose levels 2, 4, and 6) are displayed in Figure 3 and listed in Table 3. For nodule 1 (volume, 14.1 mm3; diameter, 3 mm), all measurements again showed volume overestimations ranging from 1.4% to 159.6%. Larger nodules (≥ 5 mm) were underestimated, ranging from –1.1% to –17.0%, except for one overestimation of 5.8% in the lowest dose measurement (80 kVp, 25 mAs, and 0.49 mGy) of nodule 4 (volume, 523.6 mm3; diameter, 10 mm) using iDose level 6. Figure 3 and Table 3 show that iDose level 4 shows worse results compared with iDose levels 2 and 6 for nodules 1–4, but not for nodule 5. We cannot provide an explanation for this finding (with errors well below reported interscan variation [11]) other than that it is most likely a random effect. We did not observe clinically relevant differences between nodule volumes with iterative reconstruction and FBP, both at 67.9% (1.68 mGy) and 90.6% (0.49 mGy) dose reduction, especially not for the more clinically relevant larger nodules with diameter 5 mm or larger and volume 65 mm3 or larger (nodules 2–5). Figure 1 shows the same nodule using different reconstruction techniques.
Fig. 3 —Difference between actual nodule volume and measured nodule volume using filtered back projection reconstruction (FBP) and iterative reconstruction algorithm (iDose, Philips Healthcare).
A, Tube voltage is 120 kVp, and tube current–time product is 25 mAs.
B, Tube voltage is 80 kVp, and tube current–time product is 25 mAs.
Differences Between Observers in Nodule Volume Measurement
Measured lung nodule volumes of the two observers were identical for all nodules except for one measurement (the smallest nodule was imaged at 120 kVp, 25 mAs, 1.68 mGy, and iDose level 6). One observer measured a volume of 9.5 mm3, whereas the other measured 19.1 mm3 (14.1 mm3 actual volume). For further analysis, the mean volume of the two observers (14.3 mm3) was used.
Discussion
Recent developments in CT technology have improved the quality of chest CT images [8]. Because of increased use of CT, radiologists are routinely confronted with indeterminate lung nodules in clinical practice and screening trials [18]. Because iterative reconstruction has the potential to reduce radiation dose in chest CT [19–21], this technique is of great interest for follow-up CT in patients with indeterminate lung nodules.
We showed that, for clinically relevant nodules (≥ 5 mm), volumes are comparable between FBP and iterative reconstruction and that significant decreases in tube voltage and tube current–time product with corresponding dose reductions up to 90.6% are possible without significant changes in nodule volumetry, even when using FBP. In addition, we found that, for nodules with diameter 5 mm or larger, reconstruction of CT data with both FBP and iterative reconstruction in general slightly underestimated actual nodule volume in our phantom model, whereas the nodule volume of smaller nodules was overestimated and was associated with large measurement errors.
The finding that volumetry is reliable at reduced tube voltage and tube current–time product when using routine FBP as well as iterative reconstruction, even at our extreme protocol of 80 kVp, 25 mAs, and iDose level 6, is relevant. The Fleischner Society recommends the use of low-dose CT for nodule follow-up [1]. The present study shows that dose reduction does not hamper volume measurements. Thus, our results suggest that, for nodule volumetry, it is safe to reduce tube voltage and tube current–time product and convert FBP protocols to iterative reconstruction, even for patients who are already in a follow-up scheme. Nevertheless, at our institution we maintain FBP nodule volumetry for patients who are already in follow-up and apply iterative reconstruction only for newly detected and followed nodules because of its superior image quality for chest CT, as described elsewhere [19, 22, 23], but we did not measure this in the current study.
The finding that CT underestimates the size of larger nodules is clinically relevant because nodule management guidelines are based on absolute nodule size thresholds. We showed that the combination of our CT scanner and software underestimated the nodule size, which may result in certain patients’ undergoing a less-intensive management strategy. However, we think that, for evaluation of growth rate itself, this is of lesser importance given the systematic nature of this error. It would be interesting to know whether other CT scanner and software combinations also underestimate absolute nodule volume.
This study has several limitations. First, an anthropomorphic chest phantom was used, and all measured lung nodules were spherical with sharp margins and had the same radiodensity of 100 HU. Yi et al. [24] showed that median densities of malignant and benign lung nodules are 92 and 86 HU, respectively, using contrast-enhanced chest CT and 50 and 45 HU, respectively, using unenhanced CT. On the basis of the present study, we cannot prove that our observations apply to unenhanced CT, but we do not expect a difference, given the high contrast difference between lung tissue and nodules. No nodules with lobulated, irregular, or speculated borders and no ground-glass nodules were used. Moreover, the phantom was not able to simulate breathing or cardiac motion and contained no pathologic lung structure; thus, further analyses using in vivo data are recommended. Despite the identical measurements of both observers, one measurement of the smallest nodule differed. This discrepancy is probably caused by different mouse cursor positioning within the nodule, resulting from the spherical gradually expanding volume algorithm of the semiautomatic software package, as described by Ashraf et al. [10]. Furthermore, only one semiautomated software package was used, and the results might have been different if software packages from other manufacturers were used. Finally, one CT scanner of a single vendor was used, and it is unknown whether our results apply to other CT scanners and iterative reconstruction technology of other vendors.
In conclusion, our study showed no clinically relevant differences exceeding the interscan variation in volume of lung nodules 5 mm or larger for FBP and iterative reconstruction. Volumetry remained reliable at reduced tube voltage and tube current–time product, with radiation dose reductions of up to 90% for both FBP and iterative reconstruction. Therefore, our study suggests that it is safe to reduce tube voltage and tube current–time product and to convert FBP protocols to iterative reconstruction for lung nodule follow-up CT. Finally, CT appears to slightly underestimate actual nodule volume for nodules with clinically relevant size.
Acknowledgments
We thank Robert Valkenburg for technical assistance.
References
1.
MacMahon H, Austin JH, Gamsu G, et al. Guidelines for management of small pulmonary nodules detected on CT scans: a statement from the Fleisher Society. Radiology 2005; 237:395–400
2.
Yankelevitz DF, Reeves AP, Kostis WJ, Zhao B, Henschke CI. Small pulmonary nodules: volumetrically determined growth rates based on CT evaluation. Radiology 2000; 217:251–256
3.
Ost D, Fein AM, Feinsilver SH. Clinical practice: the solitary pulmonary nodule. N Engl J Med 2003; 348:2535–2542
4.
Ahn MI, Gleeson TG, Chan IH, et al. Perifissural nodules seen at CT screening for lung cancer. Radiology 2010; 254:949–956
5.
Henschke CI, McCauley DI, Yankelevitz DF, et al. Early Lung Cancer Action Project: overall design and findings from baseline screening. Lancet 1999; 354:99–105
6.
Xu DM, van der Zaag-Loonen HJ, Oudkerk M, et al. Smooth or attached solid indeterminate nodules detected at baseline CT screening in the NELSON study: cancer risk during 1 year of follow-up. Radiology 2009; 250:264–272
7.
Revel MP, Lefort C, Bissery A, et al. Pulmonary nodules: preliminary experience with three-dimensional evaluation. Radiology 2004; 231:459–466
8.
Yanagawa M, Honda O, Kikuyama A, et al. Pulmonary nodules: effect of adaptive statistical iterative reconstruction (ASIR) technique on performance of a computer-aided detection (CAD) system—comparison of performance between different-dose CT scans. Eur J Radiol 2012; 81:2877–2886
9.
Sagara Y, Hara AK, Pavlicek W, Silva AC, Paden RG, Wu Q. Abdominal CT: comparison of low-dose CT with adaptive statistical iterative reconstruction and routine-dose CT with filtered back projection in 53 patients. AJR 2010; 195:713–719
10.
Ashraf H, de Hoop B, Shaker SB, et al. Lung nodule volumetry: segmentation algorithms within the same software package cannot be used interchangeably. Eur Radiol 2010; 20:1878–1885
11.
de Hoop B, Gietema H, van Ginneken B, Zanen P, Groenewegen G, Prokop M. A comparison of six software packages for evaluation of solid lung nodules using semi-automated volumetry: what is the minimum increase in size to detect growth in repeated CT examinations. Eur Radiol 2009; 19:800–808
12.
Hein PA, Romano VC, Rogalla P, et al. Variability of semiautomated lung nodule volumetry on ultralow-dose CT: comparison with nodule volumetry on standard-dose CT. J Digit Imaging 2010; 23:8–17
13.
Wang Y, van Klaveren RJ, van der Zaag-Loonen HJ, et al. Effect of nodule characteristics on variability of semiautomated volume measurements in pulmonary nodules detected in a lung cancer screening program. Radiology 2008; 248:625–631
14.
Wormanns D, Kohl G, Klotz E, et al. Volumetric measurements of pulmonary nodules at multi-row detector CT: in vivo reproducibility. Eur Radiol 2004; 14:86–92
15.
Funama Y, Taguchi K, Utsunomiya D, et al. Combination of a low-tube-voltage technique with hybrid iterative reconstruction (iDose) algorithm at coronary computed tomographic angiography. J Comput Assist Tomogr 2011; 35:480–485
16.
Habets J, Symersky P, de Mol BA, Mali WP, Leiner T, Budde RP. A novel iterative reconstruction algorithm allows reduced dose multidetector-row CT imaging of mechanical prosthetic heart valves. Int J Cardiovasc Imaging 2011 [Epub ahead of print]
17.
Schaller S, Wildberger JE, Raupach R, Niethammer M, Klingenbeck-Regn K, Flohr T. Spatial domain filtering for fast modification of the tradeoff between image sharpness and pixel noise in computed tomography. IEEE Trans Med Imaging 2003; 22:846–853
18.
Kim JS, Kim JH, Cho G, Bae KT. Automated detection of pulmonary nodules on CT images: effect of section thickness and reconstruction interval—initial results. Radiology 2005; 236:295–299
19.
Hu XH, Ding XF, Wu RZ, Zhang MM. Radiation dose of non-enhanced chest CT can be reduced 40% by using iterative reconstruction in image space. Clin Radiol 2011; 66:1023–1029
20.
Pontana F, Duhamel A, Pagniez J, et al. Chest computed tomography using iterative reconstruction vs filtered back projection. Part 2. Image quality of low-dose CT examinations in 80 patients. Eur Radiol 2011; 21:636–643
21.
Singh S, Kalra MK, Gilman MD, et al. Adaptive statistical iterative reconstruction technique for radiation dose reduction in chest CT: a pilot study. Radiology 2011; 259:565–573
22.
Prakash P, Kalra MK, Digumarthy SR, et al. Radiation dose reduction with chest computed tomography using adaptive statistical iterative reconstruction technique: initial experience. J Comput Assist Tomogr 2010; 34:40–45
23.
Sato J, Akahane M, Inano S, et al. Effect of radiation dose and adaptive statistical iterative reconstruction on image quality of pulmonary computed tomography. Jpn J Radiol 2012; 30:146–153
24.
Yi CA, Lee KS, Kim BT, et al. Tissue characterization of solitary pulmonary nodule: comparative study between helical dynamic CT and integrated PET/CT. J Nucl Med 2006; 47:443–450
Information & Authors
Information
Published In
Copyright
© American Roentgen Ray Society.
History
Submitted: February 3, 2012
Accepted: April 20, 2012
First published: January 14, 2013
Keywords
Authors
Metrics & Citations
Metrics
Citations
Export Citations
To download the citation to this article, select your reference manager software.