Original Research
Cardiopulmonary Imaging
March 20, 2015

Lung Nodule Detection by Microdose CT Versus Chest Radiography (Standard and Dual-Energy Subtracted)

Abstract

OBJECTIVE. The purpose of this study was to investigate the feasibility of microdose CT using a comparable dose as for conventional chest radiographs in two planes including dual-energy subtraction for lung nodule assessment.
MATERIALS AND METHODS. We investigated 65 chest phantoms with 141 lung nodules, using an anthropomorphic chest phantom with artificial lung nodules. Microdose CT parameters were 80 kV and 6 mAs, with pitch of 2.2. Iterative reconstruction algorithms and an integrated circuit detector system (Stellar, Siemens Healthcare) were applied for maximum dose reduction. Maximum intensity projections (MIPs) were reconstructed. Chest radiographs were acquired in two projections with bone suppression. Four blinded radiologists interpreted the images in random order.
RESULTS. A soft-tissue CT kernel (I30f) delivered better sensitivities in a pilot study than a hard kernel (I70f), with respective mean (SD) sensitivities of 91.1% ± 2.2% versus 85.6% ± 5.6% (p = 0.041). Nodule size was measured accurately for all kernels. Mean clustered nodule sensitivity with chest radiography was 45.7% ± 8.1% (with bone suppression, 46.1% ± 8%; p = 0.94); for microdose CT, nodule sensitivity was 83.6% ± 9% without MIP (with additional MIP, 92.5% ± 6%; p < 10–3). Individual sensitivities of microdose CT for readers 1, 2, 3, and 4 were 84.3%, 90.7%, 68.6%, and 45.0%, respectively. Sensitivities with chest radiography for readers 1, 2, 3, and 4 were 42.9%, 58.6%, 36.4%, and 90.7%, respectively. In the per-phantom analysis, respective sensitivities of microdose CT versus chest radiography were 96.2% and 75% (p < 10–6). The effective dose for chest radiography including dual-energy subtraction was 0.242 mSv; for microdose CT, the applied dose was 0.1323 mSv.
CONCLUSION. Microdose CT is better than the combination of chest radiography and dual-energy subtraction for the detection of solid nodules between 5 and 12 mm at a lower dose level of 0.13 mSv. Soft-tissue kernels allow better sensitivities. These preliminary results indicate that microdose CT has the potential to replace conventional chest radiography for lung nodule detection.
Approximately 90% of lung cancer lesions in men and 60% of lung cancer lesions in women are smoking related, for both active and former smokers [1]. Worldwide, lung cancer represents the leading cause of cancer death [2]. Most patients present at advanced stages of the disease—40% in stage IV and 30% in stage III, resulting in a poor 5-year average survival rate of 16% [3, 4]. Therefore, early detection is preferable. The combination of early detection in a high-risk group with the recent advantages offered by low-dose CT has led to the implementation of lung cancer screening programs in many countries [5, 6]. With the advantages of more efficient detector assemblies, low tube voltage imaging, and low tube current–exposure time products, it is now feasible to minimize the radiation dose that a patient receives [7].
According to recent investigations, the benefits of lung cancer screening outweigh the disadvantages of radiation exposure, as well as the costs [8]. Depending on an individual's personal risk profile, a screening routine for each patient can be established [9]. Furthermore, many institutions continue to perform conventional chest radiography for tumor screening and follow-up, as well as for exclusion of lung metastases [10].
It is necessary to reconstruct maximum intensity projections (MIPs) of the axial CT dataset to increase the sensitivity of CT for lung nodule detection [11]. Recent studies have also shown computer-assisted detection software to be beneficial in screening scenarios [12].
The literature has documented the decreased sensitivity of conventional chest radiography with decreasing lesion diameter, as well as its insufficiency to detect sub-solid (ground-glass) nodules [1315]. Nevertheless, clinicians often request chest radiographs as a first-line modality for nodule detection, primarily because of the concerns associated with radiation dose from CT.
On the basis of established guidelines, screening is performed at low dose levels, between 1 and 2 mSv [5]. With the latest advances in CT hardware and software, a microdose CT examination is feasible that imparts an effective dose similar to that from radiographs (between 0.05 and 0.2 mSv), depending on the size of the patient and the technique. In particular, a recently introduced integrated circuit detector (Stellar, Siemens Healthcare) yields effective dose reduction. At microdose level, the implementation of this new detector has the potential to reduce the radiation dose by up to 70% while maintaining diagnostic image quality [7]. Recent studies on low-dose (submillisievert) imaging of the chest using various scanners have shown that other CT systems are also capable of ultralow dose imaging, especially of the chest. Khawaja et al. [16] were able to achieve comparable low dose levels for chest CT scans within the submillisievert range, and they achieved a mean dose reduction of 69% by utilizing a new iterative reconstruction algorithm (Iterative Model Reconstruction [IMR], Philips Health-care). To our knowledge, no prior studies have focused on the comparison of different scanners for microdose lung nodule screening using an integrated circuit detector. Thus, the low dose of chest radiography can be combined with the better capability of CT for lung lesion detection.
The aim of our study is to investigate the feasibility of ultralow-dose (i.e., microdose) CT, using a dose comparable to that of a conventional chest radiograph (in two planes), in terms of the nodule detection rate with radiography versus microdose CT. Furthermore, the image quality of low-dose CT studies using different reconstruction kernels is compared. We hypothesize that microdose CT, with a dose similar to that of chest radio-graphs, provides better lesion detection.

Materials and Methods

For the current study, we used an anthropomorphic chest phantom (Chest Phantom N1, Kyoto Kagaku) with artificial lung nodules (Fig. 1). The phantom is an accurate life-size anatomic model of a human male torso with a synthetic heart, trachea, pulmonary vessels (right and left), and abdomen (diaphragm) block. The thickness of the chest wall is based on clinical data. The soft-tissue substitute material (polyurethane [specific gravity, 1.06]) and synthetic bones (epoxy resin) have x-ray absorption rates that are very close to those of human tissues. The arm-abducted position of the torso is useful for CT. The pulmonary vessels are spatially traceable. The phantom size is 43 × 40 × 48 cm, with a chest girth of 94 cm and a weight of 18 kg. Furthermore, artificial lung nodules with determined attenuations representing solid lung lesions were randomly placed. The spherical nodules had diameters of 5, 8, 10, and 12 mm. The CT attenuation was 100 HU (Fig. 1). We used only solid nodules because subsolid (ground-glass) nodules are usually not detectable on chest radiographs and are instead diagnosed on CT [17]. We used a segment-based randomized distribution of lung nodules in the lung parenchyma. Randomization for the axial distribution (central and peripheral) was performed with Excel 2011 (version 14.0, Microsoft). The total number of lung nodules placed per phantom (one to four) was also randomized using Excel software. In total, 60 phantoms with nodules and five without nodules were scanned. A total of 141 solid lung nodules were placed in the artificial lung parenchyma. The five lung phantoms without lung nodules represented the control group.
Fig. 1A —Artificial lung nodules in phantom.
A, CT images show spherical nodules with increasing cross-sectional diameters of 5 (A), 8 (B), 10 (C), and 12 (D) mm.
Fig. 1B —Artificial lung nodules in phantom.
B, CT images show spherical nodules with increasing cross-sectional diameters of 5 (A), 8 (B), 10 (C), and 12 (D) mm.
Fig. 1C —Artificial lung nodules in phantom.
C, CT images show spherical nodules with increasing cross-sectional diameters of 5 (A), 8 (B), 10 (C), and 12 (D) mm.
Fig. 1D —Artificial lung nodules in phantom.
D, CT images show spherical nodules with increasing cross-sectional diameters of 5 (A), 8 (B), 10 (C), and 12 (D) mm.
To localize the lung nodules, we divided both lungs into anatomic segments. Then, each lung segment was further divided into central and peripheral zones. The border was drawn radially in the middle between the hilum and pleura. A random generator attributed the nodule location (side, segment); position of the artificial nodules in each phantom was documented as a standard of reference (slice position). After placing the lung nodules in the phantom, we first performed microdose CT using a helical scan and the following imaging parameters: kilovoltage, 80 kVp; tube current–exposure time product, 6 mAs; reconstruction slice thickness, 1 mm; FOV, 32 cm. Care-kV and reference tube current–time product were disabled for the lowest radiation dose. A FLASH spiral with a pitch of 2.2 guaranteed the lowest exposure.
Scans were performed on a 128-MDCT scanner (Somatom Definition Flash, Siemens Health-care). The CT scanner was equipped with the integrated Stellar detector system. For maximum dose reduction, iterative reconstruction algorithms (iterative reconstruction in imaging space [IRIS], Siemens Healthcare) were used, and three levels of iterations were chosen for the image reconstruction process. For image reconstruction, I30f, I50f, and I70f kernels with lung window settings (level, −600 HU; width, 1200 HU) and axial orientation were used. Additionally, MIPs were reconstructed (slice thickness, 8 mm; increment, 2 mm). After CT, the phantom, including the unchanged nodules, was moved to the chest radiograph unit. Radiographs in two planes, including dual-energy subtraction of the bones, were acquired with the phantom in prone position using a flat panel detector (Fuji XU-D1, Fuji-film). Bone subtraction was performed with single-shot dual-energy radiography in the posteroanterior direction. The detector consists of a 0.8-mm copper filter mounted between two phosphor plates. The chest radiograph unit produces one image before and another after the copper filter, resulting in a high-energy projection. For bone suppression, the software calculates a subtraction image from the high-energy picture on the second plate and the image on the first phosphor plate. The chest radiograph unit settings for the phantom were identical to those used in the clinical setting. Thus, the standard tube voltage was set to 125 kVp; the tube-detector distance was held constant at a distance of 2 m; the reference dose was 3.5 μGy; and automated tube current modulation was implemented. The spatial resolution was 5 pixels/mm, equaling 2140 × 1760 pixels/image, with a format of 43 × 35 cm.
The study design consisted of two major parts. There was an initial, pilot study using only the 20 initial phantoms containing 45 randomly distributed nodules to determine the best reconstruction kernel, and we assessed the sensitivities of four readers with and without the aid of MIP for lung nodules with three reconstruction kernels (I30f, I50f, and I70f). After conducting the pilot study, we performed the main evaluation. In addition to the 20 phantom scans used in the pilot study, another 40 phantoms containing 96 additional artificial lung nodules were included in the main evaluation, giving a total of 141 nodules distributed in 60 phantoms. On the basis of the results of the pilot study, only the soft-tissue kernel was used in the main evaluation.

Image Evaluation

All analyses were performed on a PACS (PACS R11.4.1, Philips Healthcare; or Sectra PACS IDS7, Sectra). The readout was done by one radiologist specializing in thoracic imaging with 12 years' experience, a fellow in thoracic imaging with 5 years' experience, and two residents each with 4 years' experience in general radiology (including 2 years of training in chest imaging). The radiologists were blinded to the true-positive and MIP results. On average, the readers examined a stack of 329 continuous, axial 1-mm slices and 162 MIP slices. In the initial, pilot study, the readout order of the three kernels and the phantoms was random for the four readers, with a consistent break of 3 weeks between the three readouts of the 45 nodules. After 3 months (second part of the study), the radiologists reread the phantoms from the pilot study plus the additional phantoms (all 141 nodules), each with only the best kernel. The first 45 nodules were used to evaluate intraobserver agreement. For the assessment of interobserver agreement, all 141 nodules were used (including the 45 nodules from the pilot study). The radiologists started with the microdose CT readout so that any potential recall bias would be in favor of the hypothesized inferior examination (i.e., chest radiography). The readout of the chest radiographs and the microdose CT scans was performed in random order, with a break of 3 weeks between the two sessions.
The radiologists rated the microdose CT images according to their subjective image quality. We used a scale ranging from 1 to 5, as follows: 1, inferior image quality, nondiagnostic; 2, poor image quality, with diagnostic confidence significantly reduced; 3, moderate reduction in image quality, but still sufficient for diagnosis; 4, good image quality; 5, excellent image quality. The nodules detected by the radiologists were documented in an Excel table, with the corresponding table position at the maximum lesion diameter.
Then, the readers measured the signal, contrast, and noise for the pilot study. For the objective assessment of image quality, we calculated the signal-to-noise ratio and the contrast-to-noise ratio. Therefore, the readers measured the attenuation (in HU) and its SD (i.e., noise) in an ROI with a diameter of 2 cm (area, 3.14 cm2). The ROIs were placed in air outside the phantom, anterior to the sternum, in the bone (middle of a vertebral body, excluding the vertebral endplates), and in soft tissue (heart). The measurements were standardized at the level just above the diaphragm (Fig. 2). Measurements were recorded for air, soft tissue, and bone in each of the lung kernels (I30f, I50f, and I70f). The image quality, represented by the signal-to-noise ratio, was calculated for soft tissue by dividing the signal intensity (SI) of the soft tissue by its SD (i.e., background noise), as follows:
where SIst is the signal intensity (in HU) of soft tissue and SD is the soft-tissue background noise (in HU). For air, only noise was recorded because the signal in air should be negligible. Contrast-to-noise ratio was defined as the difference between the signal intensities of bone and soft tissue (in HU), divided by the soft-tissue background noise, as follows:
where SIb and SIst are the signal intensities (in HU) of bone and soft tissue, respectively, and SD is the soft-tissue background noise (in HU). Next, the axial image reconstructions were systematically examined for lung lesions. Additional lung nodules identified with the aid of the MIP were recorded separately. For the readout of the conventional chest radiographs, the radiologists evaluated the radiographs in both planes and subsequently interpreted the posteroanterior projection with bone subtraction. The four readers, who were blinded to the nodule distribution, searched the radiographs for lung nodules. Lesions were then recorded on a standardized readout sheet by marking their location on a chest schematic. A 3-week break was instituted between sessions to prevent reader bias.
Fig. 2A —Signal-to-noise and contrast-to-noise ratio measurements.
A, CT images show ROIs depicting mean (SD) signal intensity in soft-tissue kernel (I30f) (A), intermediate kernel (I50f) (B), and hard reconstruction kernel (I70f) (C). ROIs are drawn in air outside phantom (1), in mediastinum (2), and at vertebral body (3). Diameter is held constant at 2 cm.
Fig. 2B —Signal-to-noise and contrast-to-noise ratio measurements.
B, CT images show ROIs depicting mean (SD) signal intensity in soft-tissue kernel (I30f) (A), intermediate kernel (I50f) (B), and hard reconstruction kernel (I70f) (C). ROIs are drawn in air outside phantom (1), in mediastinum (2), and at vertebral body (3). Diameter is held constant at 2 cm.
Fig. 2C —Signal-to-noise and contrast-to-noise ratio measurements.
C, CT images show ROIs depicting mean (SD) signal intensity in soft-tissue kernel (I30f) (A), intermediate kernel (I50f) (B), and hard reconstruction kernel (I70f) (C). ROIs are drawn in air outside phantom (1), in mediastinum (2), and at vertebral body (3). Diameter is held constant at 2 cm.

Dose Calculation

The dosage was represented by the dose-length product (DLP) (in mGy · cm), which the scanner automatically calculated for a phantom diameter of 32 cm for each scan, for a constant scan length of 35 cm. For the conventional chest radiographs, the dose was calculated with a preinstalled airkerma measurement tool (KermaX plus DDP, Scanditronix Wellhofer).

Effective Dose Calculation for Conventional Radiographs

Based on model simulations, the effective dose of radiography is given by the product of the dose-area product and the organ-specific conversion factor, as follows:
where E is the effective dose (in mSv), DAPCR is the air-kerma dose-area product (in mGy · cm2), and kDAP is a conversion factor for thorax of 0.21 mSv/ (Gy · cm2) for the posteroanterior projection and 0.11 mSv/(Gy · cm2) for the lateral projection (approximating the effective dose as calculated from Monte Carlo simulations of phantom measurements) [18].

Effective Dose Calculation for Microdose CT

The effective radiation dose for microdose CT is given by the product of the dose-length product and the organ-specific conversion factor. According to the recommendations of the International Commission on Radiological Protection (ICRP) [19], the conversion factor for chest CT at 80 kVp for an adult is 0.0147. Therefore,
where E is the effective dose, DLP is the dose-length product (9 mGy · cm), and CF is the conversion factor (0.0147) [20]. We calculated the adapted effective doses, including the organ-specific correction factors (mSv), for both modalities.

Statistics

The objective and subjective imaging qualities of the microdose CT scans were compared using the Wilcoxon test. The doses for the chest microdose CT scans and for the conventional radiographs were also analyzed using the Wilcox-on test. The diagnostic per-nodule sensitivity and the false-positive rate, as well as the per-phantom accuracy for the lung nodules for both modalities (microdose CT and chest radiographs), were determined and compared using the McNemar test. The inter- and intraobserver variabilities were assessed with the Fleiss kappa test [21].

Results

Ethical approval was waived owing to the phantom nature of the study. In total, 141 solid lung nodules were imaged in 65 chest phantoms. Five phantoms were empty—that is, with no nodules placed. All included examinations were performed using diagnostic standards. Each reader rated the subjective image quality as sufficient for diagnostic purposes.

Pilot Study

The soft-tissue reconstruction kernel (I30f) delivered significantly better sensitivities for all four radiologists than the hard kernel (I70f), with mean sensitivities of 91.1% ± 2.2% versus 85.6 ± 5.6% (p = 0.041), respectively. The mean sensitivity of the medium reconstruction kernel (I50f) was 88.9% ± 4.4% and was not significantly different from those of the other kernels. The imaging quality features (Fig. 3) were best for the soft-tissue reconstruction kernel. The mean image noise values in soft tissue for I30f, I50f, and I70f were 60.3 ± 10.3, 153.9 ± 11.5, and 292.6 ± 22.1 HU, respectively. The mean signal-to-noise ratios in soft tissue for I30f, I50f, and I70f were 0.28 ± 0.07, 0.11 ± 0.03, and 0.12 ± 0.01, respectively. The mean contrast-to-noise ratios for I30f, I50f, and I70f were 8.1 ± 0.4, 3.3 ± 0.3, and 1.6 ± 0.1, respectively. Additionally, the subjective image quality favored the soft-tissue reconstruction kernel; for all readers, the mean subjective image quality for I30f, I50f, and I70f was rated as 4.3 ± 0.5, 4.0 ± 0.8, and 2.7 ± 0.5, respectively.
Fig. 3 —Graph compares imaging quality features among three reconstruction kernels.

Main Study

Because the objective and subjective results of the pilot study achieved the best results with the I30 kernel, only this kernel was used for the main study.
Per-nodule analysis for chest radio-graphs and microdose CT—The clustered analysis for lung nodule assessment in the regular chest radiograph studies exhibited a mean sensitivity for the four radiologists of 45.7% ± 8.1%. When additional bone suppression was used, there was only a slight increase in the sensitivity to 46.1% ± 8% (p = 0.94). The mean sensitivity for nodule detection with microdose CT was 83.6% ± 9% without MIP; when additional MIP was used, the sensitivity was increased to 92.5% ± 6% (p < 10–3). In comparison between micro-dose CT (with and without MIP) and chest radiographs (with and without bone subtraction), microdose CT always showed significantly better capability for lung lesion detection (p < 10–6) (Fig. 4). The individual sensitivities of the readers in the per-nodule analysis are provided in Table 1.
Fig. 4 —Comparison of lung nodule detection between chest radiographs and microdose CT.
A, Posteroanterior chest radiograph.
B, Coned-down excerpt from A shows lung nodule in right lower lobe.
C, Microdose CT image clearly facilitates lesion detection.
TABLE 1: Per-Nodule Sensitivity Analysis
ModalityReader 1Reader 2Reader 3Reader 4
Microdose CTa84.3 (95.7)90.7 (96.4)68.6 (82.1)90.7 (95.7)
Chest radiographsb42.9 (44.3)58.636.445.0

Note—Data shown are individual sensitivity values (%) by reader for 140 nodules.

a
Values in parentheses are with additional maximum intensity projection (MIP), which are shown only where they differed from values without additional MIP.
b
Values in parentheses are with bone subtraction (BS), which are shown only where they differed from values without BS.
Size-related sensitivity analysis for chest radiographs and microdose CT—The lung nodule detection rate is size dependent for both modalities and increases with lesion diameter (Table 2). Microdose CT showed a mean sensitivity for the smallest nodules (5 mm) of 45.6% ± 21.1%. For the largest nodules (12 mm), microdose CT reached a sensitivity of 100% ± 0%. In contrast, radiography showed mean sensitivities for the smallest (5 mm) and largest (12 mm) nodules of 4.4% ± 7.6% and 80% ± 2%, respectively. For every lung nodule size, microdose CT was better than chest radiographs (Table 2). The additional use of MIP improved the sensitivity of microdose CT in lung lesions with sizes ranging from 5 to 10 mm. Bone suppression was not able to improve the sensitivity of chest radiographs significantly for any nodule size.
TABLE 2: Nodule-Dependent Sensitivity Analysis
Diameter (mm)No. of NodulesMicrodose CTaChest Radiographsbpc
53445.6 ± 21.1 (75.0 ± 17.8)4.4 ± 7.6< 10−6
83590.0 ± 11.2 (95.0 ± 5.5)40.0 ± 11.2< 10−6
103697.2 ± 4.8 (99.3 ± 1.2)59.7 ± 12.5 (60.4 ± 12.2)< 10−6
1236100.0 ± 0.080.0 ± 2.0 (80.7 ± 2.4)< 10−6

Note—Except where otherwise indicated, data shown are mean ± SD sensitivity values (%) among four readers.

a
Values in parentheses are with additional maximum intensity projection (MIP), which are shown only where they differed from values without additional MIP.
b
Values in parentheses are with bone subtraction (BS), which are shown only where they differed from values without BS.
c
For both without and with maximum intensity projection and bone subtraction.
Per-phantom analysis for microdose CT—Microdose chest CT exhibited a mean sensitivity of 96.2% ± 1.3%, a mean specificity of 100% ± 0%, and an overall mean accuracy of 96.9% ± 1.1%. When MIP was used in conjunction, the sensitivity, specificity, and accuracy were identical. The individual sensitivities, specificities, and accuracies for each reader are provided in Table 3.
TABLE 3: Per-Phantom Analysis of Sensitivity, Specificity, and Accuracy
ModalityReader 1Reader 2Reader 3Reader 4
Microdose CTa    
 Sensitivity (%)96.296.294.398.1
 Specificity (%)100.0100.0100.0100.0
 Accuracy (%)96.996.995.498.5
Chest radiographsb    
 Sensitivity (%)69.8 (71.7)86.869.873.6
 Specificity (%)83.3 (66.7)100.091.7100.0
 Accuracy (%)72.3(70.8)89.273.878.5

Note—For n = 56 phantoms.

a
Values in parentheses are with additional maximum intensity projection (MIP), which are shown only where they differed from values without additional MIP.
b
Values in parentheses are with bone subtraction (BS), which are shown only where they differed from values without BS.
Per-phantom analysis for chest radio-graphs—The mean sensitivity among all readers for the conventional chest radio-graphs in two planes was 75% ± 7% (without bone subtraction). The specificity was 93.8% ± 6.9% with a diagnostic accuracy of 78.5% ± 6.6% (for reader-specific results, see Table 3). Bone subtraction did not offer a significant improvement for lung lesion detection; the sensitivity with the additional use of bone suppression was 75.5% ± 6.7%, the specificity was 89.6% ± 13.7%, and the accuracy was 78.1% ± 7%.
Comparison of the per-phantom analysis for chest radiographs and microdose CT—Comparing the mean sensitivity of microdose CT and chest radiographs yielded highly significant advantages in favor of microdose CT (96.2% vs 75%; p = 10–6). The use of MIP along with microdose CT or bone subtraction with chest radiographs did not influence these results.
False-positive nodules for microdose CT and chest radiographs—When microdose CT was used, a total of two false-positive nodules in the per-nodule analysis among the four readers were detected. There were no additional false-positive lesions found using MIPs. Therefore, false-positive rate was 0.77% per phantom. On conventional chest radiographs without dual-energy subtraction, the four readers scored 21 false-positive lesions together (8.1% per phantom; compared with microdose CT, p = 7 × 10–5). The false-positive rate increased by two false-positive lesions after applying dual-energy subtraction to the chest radiographs (8.8% per phantom; compared with microdose CT, p < 10–5).
Interobserver variability—The mean interrater agreements (κ) of microdose CT without MIP were 1 ± 0, 1 ± 0, 0.48 ± 0.05, 0.23 ± 0.1, and 0.40 ± 0.14 for 12 mm, 10 mm, 8 mm, 5 mm, and all nodules together, respectively. With the aid of MIP, the inter-rater agreement changed only for the smaller nodules; the mean kappa value for the 8-mm nodules increased to 0.77 ± 0.33, and that for the 5-mm nodules decreased to 0.20 ± 0.1. The interrater agreement for chest radiography with or without dual-energy subtraction was always between 0.5 (moderate agreement) and 0.66 (substantial agreement), except for the case of the 5-mm nodules; with or without dual-energy subtraction, the readers did not detect these small nodules. Therefore, kappa values with and without dual-energy subtraction were not measurable.
The mean intrareader variability of micro-dose CT for all nodules was moderate, with a level of 0.55 ± 0.3. This was higher than the mean interreader variability of 0.40 ± 0.14.
Peripheral versus central location—There was no significant difference in the microdose CT sensitivity for nodule location between 0.81 peripherally and 0.85 centrally (p = 0.21). In contrast, chest radiography yielded significantly different sensitivities for the peripheral and central nodules, which were 53% ± 6% and 40% ± 10% (p = 0.0029), respectively.
Effective dose calculations—The mean effective radiation dose for the chest radiograph in two planes, including dual-energy bone subtraction, was 0.242 mSv (0.119 mSv for the posteroanterior projection and 0.123 mSv for the lateral projection). For microdose CT, the applied effective dose was 0.1323 mSv for the spiral scan (including CT radiograph). The actual DLP for the CT planning topogram was 5 mGy · cm; in contrast, the DLP for the diagnostic scan was 4 mGy · cm.

Discussion

The current study shows that a microdose chest CT scan with an even lower dose than that of chest radiographs in two planes can be achieved in a phantom-based setting. Micro-dose CT showed better lung nodule detection compared with chest radiographs, regardless of the combination of chest radiographs with bone subtraction and compared with micro-dose CT with or without MIP (Fig. 5). The mean and individual diagnostic accuracies of microdose CT were significantly better than with chest radiography. Furthermore, the false-positive findings in microdose CT were significantly lower compared with chest radiographs with dual-energy subtraction. The higher false-positive rate using dual-energy subtraction compared with conventional radiographs without dual-energy subtraction may be related to readers' experience.
Fig. 5 —Comparison of lung nodule detection between modalities including bone subtraction and varying reconstruction kernels.
A, Regular posteroanterior chest radiograph.
B, Same radiographic view as A with bone subtraction enables identification of suprahilar lung nodule in right upper lobe.
C–E, Microdose CT images show better lesion conspicuity of nodule seen in B using reconstruction kernels I30f (C), I50f (D), and I70f (E). Lesion is evident with all reconstruction algorithms. Note increasing image noise with harder reconstruction kernels.
Regarding the size- and location-dependent analysis of nodule sensitivity, CT's sensitivity was always better than that of chest radiography (Table 2). These findings are in accordance with current literature [22].
The use of the bone subtraction technique in conjunction with conventional radiography was unable to increase the diagnostic accuracy of chest radiography in the current study. This finding is in contrast with the findings of previous studies [2325]. Szucs-Farkas et al. [25] were able to show significantly better detection rates with bone suppression images for inexperienced radiologists. Given the comparable testing conditions, dependence on reader experience may explain this finding; the readers from our study had at least 2 years of experience with chest imaging.
In a population where radiation dose and dose reduction represents a prime consideration, the utilization of dual-energy subtraction for lung lesion detection is questionable except when a benefit can be expected. In our opinion, a central question to be raised is who should perform chest lesion screening. Nevertheless, on the basis of previous data, dual-energy subtraction may benefit less experienced radiologists, enhancing their lesion recognition [23, 25]. Inexperienced readers can also achieve better lesion conspicuity with microdose CT.
In microdose CT, additional MIP improved the nodule sensitivity value for nodules as small as 5 mm (from 45.6% ± 21.1% to 75% ± 17.8%). Consequently, the effect of MIP on the nodule detection rate decreased with nodules measuring 8, 10, and 12 mm (Table 2).
MIP reconstructions did not improve the diagnostic accuracy in the phantom-based analysis; the lesion conspicuity was identical to that in the absence of MIPs. In contrast, the per-nodule analysis of the MIP reconstructions significantly enhanced the radiologists' lesion recognition (83.6% ± 9% without MIP vs 92.5% ± 6% with MIP; p < 0.001). This finding is in agreement with results in the literature indicating that the use of MIP images increases the sensitivity for lung lesion recognition [26].
The reconstruction kernel analysis showed favorable signal-to-noise ratio and contrast-to-noise ratio for the smooth soft-tissue kernel, I30f. However, the image noise was significantly higher using the sharp kernels, I50f and I70f. The minimal spatial resolution of the smooth kernel did not affect the readers' lesion detection, and the subjective rating of the image quality by the readers was highest for I30f. The interobserver variability with microdose CT was best for nodules measuring 12 and 10 mm. For nodule diameters of 8 and 5 mm, the interobserver variability was moderate. MIP was unable to significantly improve this finding. Probably because of the nonvisibility of the 5-mm nodules, chest radiography showed improved interrater variability relative to that of microdose CT.
In 1990, ICRP introduced the quantity effective dose to provide a value representing the radiation dose received by a number of specified organs and the type of applied radiation. Effective dose is the sum of all equivalent doses to the body, reflecting the stochastic radiation damage of tissues. Because effective dose mirrors the effects of ionizing radiation in general, it can be applied to investigate and compare radiation dose, independently of the modality used (CT or conventional radiography).
The effective dose of CT can easily be assessed using the DLP provided in the scan protocol. The DLP is calculated from the CT dose index (CTDI), which is derived from phantom-based dose estimations and mathematic simulations performed with Monte Carlo simulations [27].
Because of the single unidirectional exposure in conventional radiography, the amount of ionizing radiation can be measured with three quantities [19]: air kerma, entrance surface dose, and dose-area product (DAP). Of these quantities, DAP best reflects the effective dose. Furthermore, DAP can be simply converted from air kerma and surface dose when the beam geometry is known. Similar to dose estimation in CT, the DAP is then adjusted with organ weighting factors, tube voltage, and the filter used, resulting in the effective dose.
The average effective radiation dose for posteroanterior and lateral images of the chest is reported to be 0.1 mSv [28]. The mean level of radiation for radiographs in two planes with bone subtraction was significantly higher (0.242 mSv) in the current study. Although an increased dose of 14% for a single-shot, dual-energy chest radio-graph in posteroanterior orientation has been reported [29], the present value falls within the upper range of the acceptable dose for chest radiographs. In contrast, the effective dose for microdose CT was almost half the dose of single-shot, dual-energy radiography. Specifically, the dose was 0.13 mSv, which corresponds more closely to the aforementioned dose of regular chest radiographs of 0.1 mSv. Therefore, one could assume that for the same dose level, the performance of microdose CT is better than regular chest radiographs for nodule detection.
In 2008, Mettler et al. [28] reported an average effective dose for chest CT examinations of 7 mSv (range, 4.0–18.0 mSv). Neroladaki et al. [30] performed ultralow-dose chest CT scans with an effective radiation dose of 0.16 ± 0.006 mSv, similar to our test results (0.13 mSv). Neroladaki et al. used a pitch factor of 0.984:1 beam pitch. Additionally, we applied a high-pitch protocol with a factor of 2, which allowed further reduction of the dose [31].
Ultralow-dose scanning protocols for chest CT generally consist of high-pitch settings, lower tube voltage values (80–100 kVp), and lower tube currents (10–25 mAs). In the study population of Neroladaki et al. [30], settings of 100 kVp and 10 mAs provided an effective dose similar to that in the present protocol (80 kVp and 6 mAs). Reducing the tube current by 40% resulted in an effective dose reduction of 19% (0.16 vs 0.13 mSv). This phenomenon may present a subsiding effect on dose reduction in the ultralow-dose scanning range.
In the newest Siemens scanners, the lowest settings for the tube current and voltage are 80 kVp and 6 mAs. Below these threshold values, quantum starvation will produce noise artifacts, resulting in insufficient image quality. Perhaps with continued technical development, further dose reduction will be possible.
Finally, the chest CT scan consists of the following two major components: the CT radiograph (topogram and scout images) and the helical scan itself. The CT radiograph is mandatory for utilizing automated tube current modulation programs. In the actual study setting, the scout image was obtained using 100 kVp and 40 mAs, which resulted in an effective dose of 0.074 mSv. Compared with the helical scan itself, which has study preferences of 80 kVp and 6 reference mAs (0.056 mSv), these values appear high. It would be worthwhile to identify techniques that allow the omission of CT radiographs for microdose CT in a screening setting. The risk of overscanning in the z-axis without a topogram must be resolved in future studies.

Limitations

The general limitations for phantom-based studies are self-evident—the results must be confirmed in clinical studies. Given the promise of the results presented here, our institution is preparing to implement microdose protocols in clinical trials. Although microdose CT has the potential to replace chest radiography for nodule detection, it will not replace standard chest CT for patients with interstitial lung disease or those with mediastinal, heart, pleural, or chest wall disease. Furthermore, no foreign materials were located in the scan range (e.g., chest tubes, wires, catheters); such devices could potentially produce artifacts, such as beam hardening, that would result in increased radiation dose. Additionally, obese patients will require adapted dose regimens; the hypothesized study population did not include obese patients with a body mass index (weight in kilograms divided by the square of height in meters) above 25.
The dimensions of the anthropomorphic chest phantom are based on an averagesized Asian male torso (adult weight, 70 kg), which might not be reflective of other populations. Finally, we only included solid lung nodules; therefore, ground-glass opacities were not taken into account.

Conclusion

The use of microdose CT for the detection of nodules and other lung disease with a dose comparable to that of conventional chest radiography is feasible when there is no indication for a standard CT. In a phantom-based setting, the accuracy of lung nodule detection with microdose CT is significantly better than that of radiography. These preliminary results indicate that microdose CT has the potential to replace chest radiography for lung nodule detection.

Footnotes

Supported by the Bernese Cancer League, the Jubilee Foundation Swisslife, and the Swiss Fight Against Cancer Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Based on a presentation at the Radiological Society of North America 2013 annual meeting, Chicago, IL.

References

1.
Robert Koch Institut, Zentrum für Krebsregisterdaten. Lungenkrebs (Bronchialkarzinom): ICD-10 C33-34. www.krebsdaten.de/Krebs/DE/Content/Krebsarten/Lungenkrebs/lungenkrebs.html. Accessed December 16, 2014
2.
Jemal A, Siegel R, Xu J, Ward E. Cancer statistics, 2010. CA Cancer J Clin 2010; 60:277–300
3.
Howlander N, Noone AM, Krapcho M, et al., eds. SEER cancer statistics review, 1975–2008. National Cancer Institute website. seer.cancer.gov/archive/csr/1975_2008/. Published 2011. Accessed December 16, 2014
4.
Bach PB, Mirkin JN, Oliver TK, et al. Benefits and harms of CT screening for lung cancer: a systematic review. JAMA 2012; 307:2418–2429
5.
National Lung Screening Trial Research Team. Results of initial low-dose computed tomographic screening for lung cancer. N Engl J Med 2013; 368:1980–1991
6.
Boiselle PM. Computed tomography screening for lung cancer. JAMA 2013; 309:1163–1170
7.
Christe A, Heverhagen J, Ozdoba C, Weisstanner C, Ulzheimer S, Ebner L. CT dose and image quality in the last three scanner generations. World J Radiol 2013; 5:421–429
8.
Kovalchik SA, Tammemagi M, Berg CD, et al. Targeting of low-dose CT screening according to the risk of lung-cancer death. N Engl J Med 2013; 369:245–254
9.
Naidich DP, Bankier AA, MacMahon H, et al. Recommendations for the management of sub-solid pulmonary nodules detected at CT: a statement from the Fleischner Society. Radiology 2013; 266:304–317
10.
Kruger R, Flynn MJ, Judy PF, Cagnon CH, Seibert JA. Effective dose assessment for participants in the National Lung Screening Trial undergoing posteroanterior chest radiographic examinations. AJR 2013; 201:142–146
11.
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
12.
Christe A, Szucs-Farkas Z, Huber A, et al. Optimal dose levels in screening chest CT for unimpaired detection and volumetry of lung nodules, with and without computer assisted detection at minimal patient radiation. PLoS ONE 2013; 8:e82919
13.
Shah PK, Austin JH, White CS, et al. Missed non-small cell lung cancers: radiographic findings of potentially resectable lesions evident only in retrospect. Radiology 2003; 226:235–241
14.
Quekel LG, Kessels AG, Goei R, van Engelshoven JM. Miss rate of lung cancer on the chest radiograph in clinical practice. Chest 1999; 115:720–724
15.
Detterbeck FC, Homer RJ. Approach to the ground-glass nodule. Clin Chest Med 2011; 32:799–810
16.
Khawaja RD, Singh S, Gilman M, et al. Computed tomography (CT) of the chest at less than 1 mSv: an ongoing prospective clinical trial of chest CT at submillisievert radiation doses with iterative model image reconstruction and iDose4 technique. J Comput Assist Tomogr 2014; 38:613–619
17.
Hansell DM, Bankier AA, MacMahon H, McLoud TC, Müller NL, Remy J. Fleischner Society: glossary of terms for thoracic imaging. Radiology 2008; 246:697–722
18.
Le Heron JC. Estimation of effective dose to the patient during medical x-ray examinations from measurements of the dose-area product. Phys Med Biol 1992; 37:2117–2126
19.
ICRP. The 2007 recommendations of the International Commission on Radiological Protection: ICRP publication 103. Ann ICRP 2007; 37:1–332
20.
Deak PD, Smal Y, Kalender WA. Multisection CT protocols: sex- and age-specific conversions factors used to determine effective dose from dose-length product. Radiology 2010; 257:158–166
21.
Fleiss JL. Measuring nominal scale agreement among many raters. Psychol Bull 1971; 76:378–383
22.
de Hoop B, Schaefer-Prokop C, Gietema HA, et al. Screening for lung cancer with digital chest radiography: sensitivity and number of secondary work-up CT examinations. Radiology 2010; 255:629–637
23.
Szucs-Farkas Z, Patak MA, Yuksel-Hatz S, Ruder T, Vock P. Improved detection of pulmonary nodules on energy-subtracted chest radiographs with a commercial computer-aided diagnosis software: comparison with human observers. Eur Radiol 2010; 20:1289–1296
24.
White CS, Flukinger T, Jeudy J, Chen JJ. Use of a computer-aided detection system to detect missed lung cancer at chest radiography. Radiology 2009; 252:273–281
25.
Szucs-Farkas Z, Schick A, Cullmann JL, et al. Comparison of dual-energy subtraction and electronic bone suppression combined with computer-aided detection on chest radiographs: effect on human observers' performance in nodule detection. AJR 2013; 200:1006–1013
26.
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
27.
Kalender WA. Computed tomography, 3rd ed. Erlangen, Germany: Publicis, 2011:175–224
28.
Mettler FA, Huda W, Yoshizumi T, Mahesh M. Effective doses in radiology and diagnostic nuclear medicine: a catalog. Radiology 2008; 248:254–263
29.
Kuhlman JE, Collins J, Brooks GN, Yandow DR, Broderick LS. Dual-energy subtraction chest radiography: what to look for beyond calcified nodules. RadioGraphics 2006; 26:79–92
30.
Neroladaki A, Botsikas D, Boudabbous S, Becker CD, Montet X. Computed tomography of the chest with model-based iterative reconstruction using a radiation exposure similar to chest X-ray examination: preliminary observations. Eur Radiol 2013; 23:360–366
31.
De Zordo T, von Lutterotti K, Dejaco C, et al. Comparison of image quality and radiation dose of different pulmonary CTA protocols on a 128-slice CT: high-pitch dual source CT, dual energy CT and conventional spiral CT. Eur Radiol 2012; 22:279–286

FOR YOUR INFORMATION

The comprehensive book based on the ARRS 2014 annual meeting categorical course on The Radiology M and M Meeting: Misinterpretations, Misses, and Mimics is now available! For more information or to purchase a copy, see www.arrs.org.

Information & Authors

Information

Published In

American Journal of Roentgenology
Pages: 727 - 735
PubMed: 25794062

History

Submitted: March 21, 2014
Accepted: June 11, 2014

Keywords

  1. chest CT dose
  2. CT dose reduction
  3. lung lesion detection
  4. microdose CT imaging

Authors

Affiliations

Lukas Ebner
Department of Diagnostic, Interventional and Pediatric Radiology, Hospital and University of Bern Inselspital, Freiburgstrasse 10, Bern CH-3010, Switzerland.
Institute of Forensic Medicine, University of Zurich, Zurich, Winterthurstrasse 190/52, CH-8057 Zurich, Switzerland.
Yanik Bütikofer
Department of Diagnostic, Interventional and Pediatric Radiology, Hospital and University of Bern Inselspital, Freiburgstrasse 10, Bern CH-3010, Switzerland.
Daniel Ott
Department of Diagnostic, Interventional and Pediatric Radiology, Hospital and University of Bern Inselspital, Freiburgstrasse 10, Bern CH-3010, Switzerland.
Adrian Huber
Department of Diagnostic, Interventional and Pediatric Radiology, Hospital and University of Bern Inselspital, Freiburgstrasse 10, Bern CH-3010, Switzerland.
Julia Landau
Department of Diagnostic, Interventional and Pediatric Radiology, Hospital and University of Bern Inselspital, Freiburgstrasse 10, Bern CH-3010, Switzerland.
Justus E. Roos
Department of Radiology, Duke University Medical Center, Durham, NC.
Johannes T. Heverhagen
Department of Diagnostic, Interventional and Pediatric Radiology, Hospital and University of Bern Inselspital, Freiburgstrasse 10, Bern CH-3010, Switzerland.
Andreas Christe
Department of Diagnostic, Interventional and Pediatric Radiology, Hospital and University of Bern Inselspital, Freiburgstrasse 10, Bern CH-3010, Switzerland.

Notes

Address correspondence to L. Ebner ([email protected]).

Metrics & Citations

Metrics

Citations

Export Citations

To download the citation to this article, select your reference manager software.

Articles citing this article

View Options

View options

PDF

View PDF

PDF Download

Download PDF

Media

Figures

Other

Tables

Share

Share

Copy the content Link

Share on social media