70 kV computed tomography of the thorax: valence for computer-assisted nodule evaluation and radiation dose

Abstract

Background

Computed tomography (CT) is the gold standard for evaluation of pulmonary nodules and is at the same time responsible for the majority of the collective effective dose.

Purpose

To evaluate radiation dose and efficacy of computer-assisted detection (CAD) for solid pulmonary nodules in low dose chest CT performed at 70 kV.

Material and Methods

CAD was performed upon chest CT with 70 kV and 100 kV (gold standard) at manufacture’s recommended tube current of 87 mAs (collimation, 64 × 0.6 mm). Detection rate for pulmonary nodules and size measurements of both techniques were compared to each other. Radiation dosage in terms of effective dose (E) was measured using an Alderson-Rando Phantom.

Results

Seventy-four patients with 301 solid nodules were included in the study. CAD detection rate was similar for 70 kV (94.7%) and 100 kV (92.4%). Mean transversal nodule diameter was 5.5 mm for 70 kV and 5.7 mm for 100 kV with an average volume of 0.12 mL (both techniques). Derived from the phantom measurements patient examinations resulted in an E of 0.51 mSv (70 kV) versus 2.02 mSv (100 kV).

Conclusion

70 kV low-dose chest CT is suitable for CAD based lung nodule analysis at a fraction of the radiation burden of the standard technique. Since the measurements are highly accurate, 70 kV CT could be used for detection of pulmonal lesions as well as follow-up studies.

Keywords

Chest CT CAD radiation dose 70 kV pulmonary nodules

Introduction

Helical computed tomography (CT) of the thorax represents the current gold standard for the evaluation of pulmonary nodules due to quick acquisition of a three-dimensional image set in a high spatial resolution. While most small nodules are of benign origin, nodule characteristics such as solid appearance, a diameter of >5 mm, and a traceable growth rate raise suspicion for malignancy (1 –3). Considering average tumor doubling times and patient risk profiles, follow-up examinations are therefore usually performed within 3–12 months (4). For about a decade, radiologists have been able to make use of dedicated software applications (computer-assisted detection [CAD]) which search for nodules and perform accurate size measurements (5). However, increased concern is raised among patients and physicians since CT is known to be responsible for the majority of ionizing radiation exposure in today’s medicine (6,7). Apart from using automated tube current modulation techniques and iterative reconstruction algorithms, scientific efforts focus on dose savings in chest CT by reducing tube voltage and current (8,9).

Recently a new X-ray tube for CT devices (Straton MX, Siemens Healthcare, Erlangen, Germany) that can perform CT examinations at voltages of 70 kV, has become commercially available, promising a new perspective in low-dose imaging. The purpose of this study was to evaluate 70 kV chest CT regarding radiation dosage and effectiveness of CAD-based solid nodule detection.

Material and Methods

Patients, examination technique

Data from patients who underwent chest CT for cancer staging between May and June 2012 in a 64-slice scanner (Somatom Definition AS, Siemens Healthcare, Erlangen, Germany) were retrospectively evaluated for this study. The image sets used for this analysis were taken from 74 consecutive patients (46 men, 28 women) with an average age of 59 years (range, 19–83 years). Instead of the standard protocol of an 80 kV non-enhanced CT followed by a contrast-enhanced CT scan (100 kV), patients were examined with 70 kV for the non-enhanced scan and then with the usual 100 kV contrast-enhanced scan. The local institutional review board waived the need for informed consent.

By using the manufacturer’s default, the tube current of both scans was set to 87 mAs with the following parameters: activated tube current modulation; rotation time, 0.33 s; pitch value, 0.6; and collimation, 64 × 0.6 mm. Prior to the CT scan, a scout view (100 kV, 35 mA, frontal view) was prepared to plan the examination range. To minimize the risk of motion artifacts image acquisition was performed in end-inspirational breath-hold in caudo-cranial direction. Both data-sets were reconstructed from raw data in the transverse direction (slice thickness, 1.5 mm; increment, 0.7 mm) with a hard, edge enhancing kernel (i70) in an image matrix of 512 × 512. Iterative image reconstruction “SAFIRE” (Siemens Healthcare, Erlangen, Germany) was activated by default (strength: 3).

Computer-assisted nodule detection, background noise

Lung CAD was performed on a dedicated multimodality workstation (Leonardo-Syngo MMWP V.36 A, Siemens Healthcare, Erlangen, Germany) running the software application “Oncology”. Apart from CAD, this software features real-time multiplanar image reconstruction for manual evaluation of the data-set. By default the software was set to evaluate solid nodules only, since the authors are aware of many false-positive results of CAD tools when evaluating ground glass apparent consolidations. First, the 100 kV CT studies were manually analyzed by three readers with 3–8 years of experience in chest radiology in consensus (BS, BB, and MK) regarding the number and location of the pulmonary nodules (lung window, center, –500 HU; width, 2000 HU). Then, both data-sets (70 kV and 100 kV) were analyzed again, this time using the CAD function “automatic lesion detection”. The software automatically displayed the location of the nodules, measured the maximum transverse diameter, and performed volumetric analysis. All CAD results were reassessed by the three readers regarding quantitative false-positive and false-negative findings and the results of both kV-techniques were compared to each other. Lesions that were not detected by CAD were manually tagged and added to the CAD list for size evaluation. According to the recommendations of the Fleischner Society, the nodules were divided into four groups: <4 mm, 4–6 mm, 6–8 mm, and >8 mm (4). If a nodule measured 4 mm in the 100 kV data-set and 3.9 mm in the 70 kV data-set, the lesion was classified in the nodule group “4–6 mm”, since CAD of the 100 kV technique was seen as the gold standard.

To measure background noise, a 10 mm region of interest (ROI) in circular shape was placed in air in front of the right chest. The lung window was used to exclude accidental measurements of the patient's sheet. Standard deviation of HU of three measurements was noted and the average was defined as the background noise. The mean result of both examination techniques was compared to assess the difference in average background noise.

Phantom measurements

The anthropomorphic Rando Phantom (The Phantom Laboratory, Salem, NY, USA) was utilized to determine radiation exposure. One hundred and twenty-six pre-calibrated lithiumfluoride thermoluminescent dosimeters (TLD, The Harshaw Chemical Company, Cleveland, OH, USA) were placed in boreholes of predefined anatomical regions of the head, neck, and trunk (3 TLDs per boreholder). Both the 70 kV and 100 kV examinations were performed upon the phantom which was placed on the examination table in the isocentral supine position. To achieve an appropriate radiation dosage, a similar to the previously performed patient examinations CTDi_vol value was taken (for the phantom measurements to keep the previously chosen CTDi_vol). The TLDs were irradiated five times in order to achieve a significant dosage. After pre-annealing the rods at 1008℃ for 10 min, they were read using a dedicated analytical aperture made by the same manufacturer as the TLDs. Each measured result was divided by 5 to achieve the TLD dose. With knowledge of the different organ doses, the effective doses (E) were calculated by following the recommendations of the International Commission On Radiation Protection (ICRP) (10).

Patient effective dose estimation

The dose length products (DLP) of the CT studies were automatically documented after each CT examination in a dose protocol sheet. DLP is a reasonable parameter to estimate patient effective dose (E) by multiplication with a standardized factor k (E = DLP × k). For this study, the effective patient doses were ascertained using the conversion factors that were determined from the phantom measurements.

Statistical analysis

Mean values and ranges were calculated and box and whisker diagrams were calculated for descriptive statistics. Sensitivity and positive predictive value of CAD performance was calculated and paired samples t-test was used to analyze differences in means. Differences in patient effective doses of both techniques were analyzed by using the Wilcoxon Signed Ranks test, since normal distribution was excluded via Kolmogorov-Smirnov test. A P value of α < 0.05 was considered a significant difference.

Results

A total number of 301 pulmonary nodules could be detected in 74 patients by the CAD software. Overall sensitivity regarding CAD based nodule detection for the 70 kV technique was 94.7% (285 nodules of 301 total nodules) and positive predictive value was 0.71 (141 false-positive nodules). The 100 kV technique had a sensitivity of 92.4% (278 of 301 detected) while the positive predictive value was 0.67 (182 false-positive nodules). CAD measurements regarding transverse nodule size in the standard 100 kV data-set resulted in a mean diameter of 5.7 mm (range, 1–25 mm; standard deviation, 3.44 mm), while the mean lesion size measured with the 70 kV technique was mathematically smaller with 5.5 mm (range, 1–25 mm; standard deviation, 3.39 mm; P < 0.01). Volumetry of the pulmonary nodules in the 100 kV data-set resulted in an average of 0.12 mL for the 100 kV data-set (range, 0.005–4.120 mL) and the 70 kV data-set (range, 0.005–4.358 mL; P = 0.607). Table 1 shows details regarding the CAD performance of both examination techniques for the different nodule sizes.

Table 1.

Performance of automated nodule detection and measurement of both examination techniques in the study population (n = 74 patients).

Nodule size (n = 301)	70 kV technique	100 kV technique
<4 mm (n = 75)	Nodules found	n = 64 (85%)	n = 61 (81%)
	Nodules missed	n = 13	n = 14
	Transversal diameter	3.3 mm	2.8 mm
	Bolumetry	0.166 mL	0.158 mL
4–6 mm (n = 119)	Nodules found	n = 113 (95%)	n = 114 (96%)
	Nodules missed	n = 6	n = 5
	Transversal diameter	4.6 mm	4.4 mm
	Volumetry	0.347 mL	0.323 mL
6–8 mm (n = 66)	Nodules found	n = 66 (100%)	n = 63 (95%)
	Nodules missed	n = 0	n = 3
	Transversal diameter	6.5 mm	6.4 mm
	Volumetry	0.721 mL	0.730 mL
>8 mm (n = 41)	Nodules found	n = 41 (100%)	n = 40 (98%)
	Nodules missed	n = 0	n = 1
	Transversal diameter	12.2 mm	12.1 mm
	Volumetry	0.603 mL	0.616 mL

The measurement of background noise showed a significantly higher (P < 0.01) standard deviation of HU in the 70 kV protocol (110.5 HU; range, 47.8–177.0 HU) versus the 51.9 HU (range, 40.1–111.8 HU) compared to the 100 kV protocol (see Fig. 1 for illustration of the images’ graininess).

Fig. 1.

Computer-assisted measurement of pulmonary nodules at 70 kV (a) and at 100 kV (b) (1.5 mm slice thickness; image center, –500 HU; image width, 2000 HU). The measured transversal diameter and volumetry differed slightly.

Phantom measurements

Radiation doses found in the phantom study were considerably lower when using the 70 kV technique than the 100 kV technique with a DLP being 38.2 mGycm (70 kV) and 132.2 mGycm (100 kV), respectively. TLD measurements resulted in an effective dose value of 0.51 mSv with 70 kV compared to 1.98 mSv with 100 kV. Then a conversion factor, k, could be developed for both techniques to calculate effective doses (E) for the patient studies (Table 2).

Table 2.

Examination parameters and effective dose measurements for thoracic CT at 70 kV and 100 kV (phantom study).

Tube voltage (kV)	70	100
Tube current (mAs)	87
Collimation (mm)	64 × 0.6
Tube current modulation	Deactivated
Scan length (cm)	33
E (mSv)	0.51	1.98
DLP (mGy x cm)	38.2	132.2
Conversion factor E/DLP	0.0134	0.0150

DLP, dose length product; E, effective dose.

Patient radiation dose

The 100 kV protocol showed significantly higher values of the DLP in comparison to the 70 kV technique (P < 0.001). The average DLP for the 70 kV technique was 38.4 mGy × cm (range, 16–96 mGy ×cm), while the DLP in the 100 kV group was 134.7 mGy ×cm (range, 62–313 mGy ×cm). The mean E value of all examinations was 0.51 mSv (70 kV) compared to 2.02 mSv (100 kV, P < 0.001). The minimum E was 0.02 mSv (resp. 0.93 mSv), and the maximum effective dose was 1.29 mSv (resp. 4.70 m Sv) (Fig. 2).

Fig. 2.

This figure shows the effective dose (E) values of chest CT of both examination techniques. The difference was statistically significant (P < 0.01).

Discussion

Responding to the ever increasing collective dose of patients who undergo X-ray-based imaging, a major challenge for radiologists currently is to keep the radiation at the lowest level necessary. It is a known fact that the reduction of photon energy (tube voltage) is a major contribution in decreasing the radiation burden in chest CT imaging (8,9). However, the loss of diagnostic validity must be avoided by all means, since the accurate and reproducible evaluation of pulmonal lesions is crucial for patient health. The purpose of this study was to analyze chest CT with the recently introduced 70 kV technique regarding CAD performance and radiation exposure.

As expected, the images acquired with the 70 kV technique suffered from increased graininess due to the reduced X-ray power. However, CAD-based nodule detection performed well with the proposed low-dose technique at a comparable, yet slightly increased sensitivity and specificity. The reason for the marginal better result of the 70 kVp series might be from physical nature: due to the decreased tube voltage, attenuation of tissues increase and therefore the increasing difference to lung tissue might turn the balance towards the low kVp protocol. Yet this is hypothetical since a measurement of varying attenuations of these nodules was not part of the study. Also, size evaluation of the lesions regarding the maximum transverse diameter and measured volume was almost identical for the four different nodule groups. Although the size measurements between both techniques reached statistical significant differences, we believe that these results may not be of clinical disadvantage: the CAD measured differences were so small (range, 0.1–0.5 mm) that the spatial resolution of 1.5 mm might be responsible for the limited accuracy. In comparison to the available data in the literature, CAD performance in this study was within standard precision (11,12). However, the results presented here showed a certain amount of false-positive performance and missed a number of nodules: about 5% of the lesions were not detected automatically, which underlines the strong need for manual re-evaluation of the lung parenchyma by the radiologist. Furthermore, the limited positive predictive value found in our study certifies this statement. However, in general CAD-based size measurements are likely to be superior to manual analysis, since even the same readers tend to vary size analysis by up to 30%, especially for lesions <5 mm (11 –13). In particular, for follow-up examinations of suspicious lesions, CAD is a powerful tool to objectively identify slowly-growing small pulmonary lesions. Interestingly the computer-assisted nodule performance was marginally worse in the 100 kV protocol, the result was constant when several examination studies were repeated. This difference could be caused by the increased difference in density between (air-filled) lung parenchyma and solid nodules, which is a consequence of a decreased tube voltage in CT imaging. The comparable diagnostic validity found in both techniques of this study suggests that the 70 kV technique is suitable for CAD-based follow-up examinations that investigate the growth behavior of lung nodules.

Based on the results from the TLD measurements, the new conversion factors (k) for the estimation of a study population’s effective dose (E) were utilized. Since E is not a physical parameter but an estimation of radiation-risk due to irradiated tissue, k is typically used in daily clinical practice for the estimation of E from the given DLP. However, derived from these results and data from the available literature, using a generic k value may led to imprecise values and may result in an underestimation of E of up to 25% (Table 3) (14 –19). As shown in the available literature as well as in our study, these differences are seen not only among different CT devices but also between different examination techniques on the same CT device (e.g. single vs. dual-source technique, spiral vs. sequential mode, and 70 kV vs. 100 kV). Therefore, specific and thus more appropriate k values should be researched and utilized in the future to estimate the patient’s radiation risk in different CT studies more precisely.

Table 3.

Different values of established conversion factors (k) that are used to estimate normalized effective dose values (E) in thoracic CT studies.

Conversion factor k = E/DLP	Examination protocol
Our study	0.0134	70 kV chest CT
Our study	0.0150	100 kV chest CT
Sommer (2010) (14)	0.0186	120 kV chest CT
	0.0188	120 kV chest CT (dual source)
	0.0180	120 kV cardiac CT
Paul (2012) (16)	0.0146	120 kV chest CT
	0.0151	100 kV chest CT
	0.0167	120 kV chest CT (dual source)
	0.0146	140/100 kV chest CT (dual energy)
Bongartz (2000) (19)	0.017	Derived from broad factors
Bongartz (2004) (18)	0.019	Derived from broad factors
Shrimpton (2004) (15)	0.014	Mean value of three different CT tubes
Jessen (1999) (17)	0.014	Derived from broad factors

DLP, dose length product.

The average E of the 70 kV technique in this study was 0.51 mSv, which is in contrast to the standard protocol result of 2.02 mSv. Compared to the data in the literature, the radiation burden is therefore lower by 50–90% (8,16,20). Especially in combination with improved iterative raw data reconstruction techniques (e.g. ASIR, SAFIRE, AIDR, iDose) that suppress image noise effectively, 70 kV imaging could be utilized for various fields of anatomic areas (8,21 –23).

Limitations of the current study include the uncertainty of performance of the 70 kV technique for soft tissue discrimination in low-contrast areas, such as mediastinal lymph nodes or muscle tissue. Moreover, the influence of artifact-causing metal implants (e.g. spine screws, metal plates, heart pacer) on the diagnostic validity is still unclear, but it is expected to be worse for the low kV technique. Furthermore, even if the examination-technique-adapted conversion factors of this study are completely accurate for estimating the effective patient dose, the method is still merely an approximation, since the values were derived from a standardized phantom body. The fact that non-enhanced (70 kV) and iodine contrast-enhanced (100 kV) data-sets were compared may have led to an overestimation of the results of the 100 kV scan since increasing the attenuation of certain nodules might have improved the overall CAD performance. Finally, the influence of patient’s body mass index was not part of this study, yet it is unclear whether the low dose protocol may be adequate for obese patients.

Comparing the 70 kV technique to chest X-ray imaging, the effective patient dose is only about three times higher, yet with the benefit of an isovolumetric data-set in submillimeter spatial resolution.

In conclusion, the study shows that 70 kV chest CT is suitable for CAD-based lung nodule detection and evaluation while causing significantly lower radiation burden to the patient.

Future analysis might now be necessary to re-evaluate the benefit of chest CT screening in contrast to the concern associated with the currently diminished radiation-based stochastic risks.

Footnotes

Conflicts of interest

JM Kerl JM is consultant of Siemens Healthcare. M Sedlmair is a research scientist of Siemens Healthcare.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

References

Ost

Fein

Feinsilver

. Clinical practice. The solitary pulmonary nodule. N Engl J Med 2003; 348: 2535–2542.

Pastorino

Bellomi

Landoni

. Early lung-cancer detection with spiral CT and positron emission tomography in heavy smokers: 2-year results. Lancet 2003; 362: 593–597.

Swensen

Jett

Hartman

. CT screening for lung cancer: five-year prospective experience. Radiology 2005; 235: 259–265.

MacMahon

Austin

Gamsu

. Guidelines for management of small pulmonary nodules detected on CT scans: a statement from the Fleischner Society. Radiology 2005; 237: 395–400.

Beyer

Wormanns

Novak

. [Clinical evaluation of a software for automated localization of lung nodules at follow-up CT examinations]. Rofo 2004; 176: 829–836.

Amis

Jr Butler

Applegate

. American College of Radiology white paper on radiation dose in medicine. J Am Coll Radiol 2007; 4: 272–284.

Smith-Bindman

Lipson

Marcus

. Radiation dose associated with common computed tomography examinations and the associated lifetime attributable risk of cancer. Arch Intern Med 2009; 169: 2078–2086.

Baumueller

Winklehner

Karlo

. Low-dose CT of the lung: potential value of iterative reconstructions. Eur Radiol 2012; 22: 2597–2606.

Tang

Wang

. Effect of low tube voltage on image quality, radiation dose, and low-contrast detectability at abdominal multidetector CT: Phantom Study. J Biomed Biotechnol 2012; 2012: 130169–130169.

10.

International Commission on Radiological Protection. 2007 recommendations of the International Commission on Radiological Protection. Publication no. 103, Ottawa, ON: Ann ICRP, 2007.

11.

Armato

3rd Li

Giger

. Lung cancer: performance of automated lung nodule detection applied to cancers missed in a CT screening program. Radiology 2002; 225: 685–692.

12.

Lee

Goo

Lee

. The potential contribution of a computer-aided detection system for lung nodule detection in multidetector row computed tomography. Invest Radiol 2004; 39: 649–655.

13.

Arimura

Suzuki

. Computer-aided detection of peripheral lung cancers missed at CT: ROC analyses without and with localization. Radiology 2005; 237: 684–690.

14.

Sommer

Schenzle

Becker

. Saving dose in triple-rule-out computed tomography examination using a high-pitch dual spiral technique. Invest Radiol 2010; 45: 64–71.

15.

Shrimpton

. Assessment of patient dose in CT. In: EUR. European guidelines for multislice computed tomography funded by the European Commission 2004: contract number FIGMCT2000-20078-CTTIP, Luxembourg: European Commission, 2004.

16.

Paul

Banckwitz

Krauss

. Estimation and comparison of effective dose (E) in standard chest CT by organ dose measurements and dose-length-product methods and assessment of the influence of CT tube potential (energy dependency) on effective dose in a dual-source CT. Eur J Radiol 2012; 81: e507–e512.

17.

Jessen

Shrimpton

Geleijns

. Dosimetry for optimisation of patient protection in computed tomography. Appl Radiat Isot 1999; 50: 165–172.

18.

Bongartz

Golding

Jurik

. European guidelines for multislice computed tomography, Brussels: European Commission, 2004.

19.

Bongartz

Golding

Jurik

. European Guidelines on quality criteria for computed tomography, Brussels: European Commission, 2000.

20.

Bauer

Kramer

Renker

. Dose and image quality at CT pulmonary angiography-comparison of first and second generation dual-energy CT and 64-slice CT. Eur Radiol 2011; 21: 2139–2147.

21.

Pontana

Duhamel

Pagniez

. 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.

22.

Singh

Kalra

Shenoy-Bhangle

. Radiation dose reduction with hybrid iterative reconstruction for pediatric CT. Radiology 2012; 263: 537–546.

23.

Tang

. Evaluation of dose reduction and image quality in chest CT using adaptive statistical iterative reconstruction with the same group of patients. Br J Radiol 2012; 85: 906–911.

70 kV computed tomography of the thorax: valence for computer-assisted nodule evaluation and radiation dose – first clinical results

Abstract

Background

Purpose

Material and Methods

Results

Conclusion

Keywords

Introduction

Material and Methods

Patients, examination technique

Computer-assisted nodule detection, background noise

Phantom measurements

Patient effective dose estimation

Statistical analysis

Results

Phantom measurements

Patient radiation dose

Discussion

Footnotes

Conflicts of interest

Funding

References