Development of a non-linear dual-energy technique in chest radiography

Highlights

• Feasibility study of developed conventional non-linear dual-energy technique for applications of chest radiography system.

• Qualitative and quantitative comparison with conventional methods of dual energy technology for improvement of quantum noise.

• We have successfully applied improved dual energy technology to detect chest diseases.

Abstract

The dual-energy (DE) technique is useful in chest radiography because it can separate anatomical structures, such as bone and soft tissue. The standard log subtraction, simple smoothing of the high-energy image, anti-correlated noise reduction, and general linear noise reduction algorithms have been used as conventional DE techniques to separate bone and soft tissue. However, 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. In this study, we propose a non-linear DE technique that requires the calculation of the coefficient in advance by using a calibration phantom. The calibration phantom is composed of aluminium and PMMA to calculate non-linear coefficients using the quadratic fitting model for soft tissue and bone. The results demonstrated that the non-linear DE technique exhibited a higher contrast-to-noise ratio, signal-to-noise ratio, and figure of merit at 60/70 kV_p and 130 kV_p. In addition, it demonstrated better performance and image quality than conventional DE techniques in terms of material decomposition capability. In conclusion, diagnostic accuracy in chest radiography is expected to increase through the non-linear DE technique.

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.