Development of a non-linear dual-energy technique in chest radiography
Introduction
Lung cancer is currently the leading cause of cancer deaths worldwide with an extremely high mortality rate; the rate of death continues to grow globally. In addition to the significant impact on patients and their families, the economic costs of cancer are enormous in terms of direct medical care for treatment and loss of human capital from early death. Because early diagnosis of lung cancer is difficult, the number of deaths from lung cancer is higher than those from other malignant diseases. Therefore, early detection of lung cancer is essential because of its low survival rate.
Digital chest radiography has been shown to have relatively low sensitivity to the detection of pulmonary nodules and is limited by incomplete sensitivity owing to complex anatomical overlapping. According to Mountain et al. (1987), radiography is sorely insufficient for the detection of early-stage diseases, missing approximately 50% of nodules less than 10 mm in diameter. This is because the visibility of the human lungs is severely affected by the clavicle and ribs, and small vertebrae and rib fractures can be partially masked by soft tissues, such as the heart and diaphragm (Brody et al., 1981).
However, digital radiography, which provides anatomical information in various regions with a simple protocol, is widely used in many clinical situations. Since the development of X-ray technology, chest radiography has been considered the gold standard for the evaluation of human anatomical structures for medical purposes. Radiography is the most commonly used diagnostic tool in clinical practice owing to its simple implementation and cost effectiveness (Garmer et al., 2000). Because of the importance of radiography in diagnosis, dual-energy (DE) radiography has been proposed and is being actively studied in many groups. This study investigates the use of DE imaging in radiography, including in early lung cancer detection. This is a worthwhile topic to review, considering decades of research and recent developments in flat-panel detector (FPD)-based DE imaging systems, theoretical and experimental methodologies for performance evaluation, and a variety of noise-reduction algorithms.
DE techniques have been reported useful in many applications including thoracic, cardiac, breast, and interventional imaging (Ducote et al., 2006; Fraser et al., 1986; Kappadath et al., 2004; Molloi et al., 1989). DE in chest radiography has been suggested for generating a material selected image through the combination of two radiation images acquired in different X-ray energy spectra. The DE X-ray absorption method is one of the most widely used methods for quantitative measurement of human body composition and has been widely used to diagnose osteoporosis and evaluate the composition of whole bodies in various clinical situations.
Despite the availability of DE techniques, the main problem with DE technology is that noise can be amplified, which can reduce diagnostic accuracy. One of these DE techniques is weighted log subtraction after the acquisition of low-energy and high-energy images. This amplifies the quantum noise, although the anatomical noise associated with a given material composition can be improved. Decades of FPD technology development has reduced the noise effect of DE images; however, the reduction in quantum noise is still an important challenge. Increasing the X-ray exposure is one solution to compensate for quantum noise, but patient exposure to radiation is a cause of concern. Therefore, the exposure parameter must deliver the appropriate radiation dose from the DE radiography. One group showed that nodule detection on chest radiography was severely limited by quantum noise, which includes anatomical and structural noise (Samei et al., 1999). If the quantum and anatomical noise in DE chest radiography could be reduced, the sensitivity of DE radiography would be significantly increased. To overcome this problem in DE chest radiography, DE techniques such as simple smoothing of high-energy images (SSH) and anti-correlated noise reduction (ACNR) were developed. In special cases, the general linear noise reduction algorithm (GLNR) has been reported as a superior technique for noise reduction and efficient lesion detection (Richard and Siewerdsen, 2008). However, the ability to quantitatively evaluate and compare these noise reduction algorithms in clinical practice has been insufficient. In addition, conventional DE techniques cannot accurately decompose anatomical structures because these techniques are based on the assumption that X-ray imaging proceeds in a linear relationship. This relationship can cause quantum noise, as well as anatomical loss of normal tissue and difficulty in detecting lesions. These techniques require optimisation by the filter to separate the spectrum because of spectral overlap. However, spectral filtration can reduce image quality because this separation cannot be completely accurate without overlap (Primak et al., 2009). In this study, we propose a non-linear DE technique, which requires the use of a calibration phantom to calculate the coefficient in advance. Finally, we compared conventional DE techniques with the non-linear DE technique.
Section snippets
Experiments set-up
A digital radiography system (INNOVISION, DK Medical Systems, Co., Ltd. South Korea) was used to acquire projection data. The maximum kVp rating of the X-ray tube is 150 kVp, and the effective focus is 0.6/1.2 mm. The target angle is 12 with an anode rotation speed of 3450 RPM. The size of the detector is 17 in 17 in (43.0 cm 43.0 cm), and the size of the pixel matrix is 3072 3072. The detector was composed of an a-Si TFT with a photodiode and CsI:Tl/Gd2O2S:Tb. The pixel pitch and
Results
Fig. 4 shows the energy optimised images using the non-linear DE technique. The upper DE images were obtained from 60/130 kVp, and the lower DE image were obtained from 70/130 kVp. With the high energy fixed at 130 kVp, the low energy was qualitatively separated at 60 and 70 kVp.
Fig. 5(a) shows that FOM value associated with the contrast-to-noise (CNR) was the highest at 60 kVp. Fig. 5(b) shows that the FOM values associated with the signal-to-noise ratio (SNR) was the highest at 70 kVp. In our
Discussion
A previous study has demonstrated the excellent performance of chest radiography in conventional DE methods (Lee et al., 2016), however quantitative measurements using a simple chest phantom were not assessed in detail for anatomical overlap and quantum noise. In this study, advanced DE techniques were compared by adding GLNR, and non-linear DE methods were proposed using a chest phantom similar to the human body to improve the usefulness of DE in chest radiography.
A digital radiography system
Conclusion
In summary, we proposed a new DE method in clinical chest radiography and compared it with conventional DE algorithms, using noise suppression to solve problems such as quantum noise. We demonstrated the feasibility of the non-linear DE method by efficiently improving the image quality, including quantum noise in radiography system for clinical application. The DE is also performed below the average dose measured in actual clinical radiography to solve dose problems caused by two exposures. The
CRediT authorship contribution statement
Minjae Lee: Conceptualization, Methodology, Software, Writing - original draft, Investigation. Donghoon Lee: Data curation, Resources, Writing - review & editing. Hyemi Kim: Validation, Visualization, Investigation, Supervision. Sunghoon Choi: Resources. Dohyeon Kim: Formal analysis.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (NRF-2017R1A2B2001818).
References (24)
- et al.
Optimization of chest radiographic imaging parameters: a comparison of image quality and entrance skin dose for digital chest radiography systems
Clin. Imag.
(2012) - et al.
Digital radiography image quality: image acquisition
J. Am. Coll. Radiol.
(2007) - et al.
Dose optimization in chest radiography: system and model characterization via experimental investigation
IEEE Trans. Instrum. Meas.
(2013) - et al.
A method for selective tissue and bone visualization using dual energy scanned projection radiography
Med. Phys.
(1981) Statistically defined backgrounds: performance of a modified nonprewhitening observer model
J. Opt. Soc. Am. A
(1994)- et al.
Optimization of a flat‐panel based real time dual‐energy system for cardiac imaging
Med. Phys.
(2006) - et al.
Calcification in pulmonary nodules: detection with dual-energy digital radiography
Radiology
(1986) - et al.
Digital radiography versus conventional radiography in chest imaging
Am. J. Roentgenol.
(2000) - et al.
Theoretical optimization of dual‐energy x‐ray imaging with application to mammography
Med. Phys.
(1985) - et al.
An algorithm for noise suppression in dual energy CT material density images
IEEE Trans. Med. Imag.
(1988)
Quantitative evaluation of dual-energy digital mammography for calcification imaging
Phys. Med. Biol.
Generalizing the MTF and DQE to include x‐ray scatter and focal spot unsharpness: application to a new microangiographic system
Med. Phys.
Cited by (5)
Adaptive dual-energy algorithm based on pre-calibrated weighting factors for chest radiography
2022, Physics in Medicine and BiologyUltra-Low-Dose Spectral CT Based on a Multi-level Wavelet Convolutional Neural Network
2021, Journal of Digital ImagingOptimization of a calibration phantom for quantitative radiography
2021, Medical Physics