Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Oct 6.
Published in final edited form as: Radiol Phys Technol. 2022 Jan 29;15(1):45–53. doi: 10.1007/s12194-021-00648-w

Quantitative analysis of changes in lung density by dynamic chest radiography in association with CT values: a virtual imaging study and initial clinical corroboration

Teruyo Sugiura 1,2, Rie Tanaka 2, Ehsan Samei 3, William Paul Segars 3, Ehsan Abadi 3, Kazuo Kasahara 4, Noriyuki Ohkura 4, Masaya Tamura 5, Isao Matsumoto 5
PMCID: PMC9536504  NIHMSID: NIHMS1838254  PMID: 35091991

Abstract

Dynamic chest radiography (DCR) identifies pulmonary impairments as decreased changes in radiographic lung density during respiration (Δpixel values), but not as scaled/standardized computed tomography (CT) values. Quantitative analysis correlated with CT values is beneficial for a better understanding of Δpixel values in DCR-based assessment of pulmonary function. The present study aimed to correlate Δpixel values from DCR with changes in CT values during respiration (ΔCT values) through a computer-based phantom study. A total of 20 four-dimensional computational phantoms during forced breathing were created to simulate both CT and projection images of the same virtual patients. The Δpixel and ΔCT values of the lung fields were correlated on a regression line, and the inclination was statistically evaluated to determine whether there were significant differences among physical types, sex, and breathing methods. The resulting conversion expression was also assessed in the DCR images of 37 patients. The resulting Δpixel values for 30/37 (81%) real patients, 6/7 (86%) normal controls, and 24/30 (80%) chronic obstructive pulmonary disorder patients were within the range of ΔCT values ± standard deviation (SD) reported in a previous study. In addition, no significant differences were detected for each condition of thoracic breathing, suggesting that the same regression line inclination values measured across the entire lung can be used for the conversion of Δpixel values, providing a quantitative analysis that can be correlated with ΔCT values. The developed conversion expression may be helpful for improving the understanding of respiratory changes using radiographic lung densities from DCR-based assessments of pulmonary function.

Keywords: Dynamic chest radiography, Four-dimensional extended cardiac-torso phantom, Radiographic lung density, Computed tomography value, Chronic obstructive pulmonary disease

1. Introduction

Pulmonary diseases, such as pneumonia and chronic obstructive pulmonary disease (COPD) have been increasing worldwide in recent years [1]. The severity of pulmonary diseases is generally determined using spirometer-based respiratory function tests. Differential severity is assessed using the pneumonia severity index [2] and A-DROP [3]. Accordingly, the localization and distribution of peripheral airway lesions and alveolar destruction, as well as evaluation of complications are essential for long-term management of pulmonary diseases. In this setting, functional lung imaging provides a major benefit by delivering combined morphological and functional information about the lungs.

Multiple modalities are now available for functional lung imaging, including nuclear medicine [4, 5], computed tomography (CT) [6-12], cine magnetic resonance imaging (MRI) [13-15], and four-dimensional ventilated-CT (4DCT) [16, 17]. Several studies have quantitatively assessed pulmonary function and lung conditions based on respiratory changes in CT values (ΔCT values), which are measured by subtracting CT values obtained on both inspiratory and expiratory CT scans [11]. The ΔCT values were caused by temporal changes in lung density during respiration, and decreased ΔCT values suggest decreased lung function. Kauczor et al. [11] examined the relationship between pulmonary function tests and ΔCT values of the lung fields (n = 155, range 18–86 years, median 58 years, male:female = 105:50), and reported that quantitative analysis of CT values was useful for interpreting lung conditions and detecting pulmonary dysfunction. They also demonstrated that the inspiratory mean lung density and the increase in expiratory attenuation could differentiate between obstructive and restrictive ventilatory impairment in normal participants, and that the best results were achieved from the scans obtained at full expiratory position (p < 0.05). Recent advances in imaging technology have enabled 4DCT imaging with low radiation exposure [17], providing functional information based on continuous ΔCT values obtained during respiration. While these 4DCT scans are useful for the evaluation of pulmonary function, image-based respiratory medicine could be simplified to improve patient irradiation, examination time, and medical expenses if information such as ΔCT values could be more readily obtained on a routine basis.

Several attempts have been made to assess lung function using radiographic imaging of lung densities [18-22], including dynamic chest radiography (DCR), a recent flat-panel detector (FPD) technology which has enabled a simple and convenient method for functional X-ray imaging [23]. Several reports have been published on the feasibility of DCR-based assessment of pulmonary function based on radiographic lung densities, which change during a respiratory cycle and can be measured by subtracting pixel values in sequential images (Δpixel values) [24-28]. If DCR could be quantitatively evaluated, it would be useful for interpreting the severity of pulmonary dysfunction related to the lung composition. Although previous clinical and animal studies have demonstrated high correlations between Δpixel values, radioisotope accumulation rates [24], and tidal volume [28], the use of DCR is currently limited to relative assessments, because pixel values are not scaled and standardized, such as CT values. For a better understanding of the respiratory changes indicated by radiographic lung densities, it is beneficial to associate Δpixel with ΔCT values within the lung fields of the same patient. Although inspiratory and expiratory CT scan data are necessary to determine ΔCT values, both scans are not routinely performed. Moreover, if both scans were routinely acquired, patients would receive an additional radiation dose. Therefore, in the present study, we focused on the 4D extended cardiac-torso (XCAT) phantom, a virtual human phantom created by computer simulation [29].

The XCAT is a library 4D computer-based anthropomorphic phantom that includes thousands of defined anatomical structures and parameterized models for cardiac and respiratory motions. This configuration enables optimization of scan parameters as well as evaluation of the diagnostic performance of various modalities, including chest CT and radiography [29-34]. Using the XCAT phantoms, we can simulate chest CT and DCR of the same subject at any respiratory phase without exposing patients to radiation. The purpose of the present study was to correlate Δpixel values in the lung fields of DCR images with corresponding ΔCT values for a better understanding of respiratory changes detected as changes in radiographic lung densities from DCR-based assessment of pulmonary function.

2. Materials and methods

2.1. Specifications of XCAT phantoms

XCAT phantoms are based on visible human male and female anatomical data (freely distributed by the U.S. National Library of Medicine) [35] and 3D brain MRI data, combined with motion data from 4DCT for beating heart and respiration. The phantom contains models of the heart and respiratory systems in motion and has been used in a variety of simulations for multimodal imaging [29-34]. The pulmonary vessels and bronchi of the XCAT phantom are reproduced up to the fifth generation [30], and lung density for the phantoms changes according to lung volume during respiration. In the present study, we set the inflated lung density to 0.26 g/cm3, which is the average lung density of a healthy adult [36]. The following data files were output: (1) a non-uniform rational B-splines (NURBS) anatomy file to be used for subsequent image projection; (2) image files in units of linear attenuation coefficient; (3) simulated CT images (matrix size: 256 × 256 pixels; slice thickness: 0.3125 cm; 500–550 slices; 32-bit real grayscale); and (4) an activity phantom with the lungs set to specified values (Fig. 1). A 256 × 256 matrix size was used to reduce the computation time in this simulation study and because high spatial resolution was not necessary for the calculation of average ΔCT values across the entire lungs.

Fig. 1.

Fig. 1

Open in a new tab

Simulated computed tomography images for each type of 4D extended cardiac-torso phantom. From left to right: normal male, overweight male, obese male, and normal female. BMI body mass index

2.2. Phantom preparation

Figure 2 shows the overall scheme of the approach. The present study involved 20 XCAT phantoms of different body types and sexes (male:female = 15:5). The body types were classified into 3 groups according to the World Health Organization (WHO) body mass index (BMI [kg/m2]) classification criteria: normal (18.5–24.9 kg/m2), overweight (25.0–29.9 kg/m2), and obese (30.0–34.9 kg/m2) [37]. XCAT phantoms of adult males (five normal, five overweight, five obese) and females (five normal) with a normal heart rate (60 beats/min) and a complete respiratory cycle were generated using the XCAT program developed at Duke University [29-33]. To simulate forced breathing, diaphragm displacement and chest-wall motion in the anteroposterior (AP) direction were set to 4 cm and 1.5 cm, respectively. These functional parameters were determined based on those measured on DCR images of 80 real patients seen in our hospital, which were not used in the clinical verification in this study. To verify the effects of chest-wall motion on Δpixel values in the lung field, we also generated an XCAT phantom without chest wall motion, i.e., with a maximum AP displacement of 0 cm. The influence of breast shadows on Δpixel values was investigated using five male and five female XCAT phantoms with normal body types.

Fig. 2.

Fig. 2

Open in a new tab

Overall scheme of our process for the measurement of respiratory changes in radiographic lung density (Δpixel values) and computed tomography (ΔCT values) in the lung fields of a computer-based phantom. The rectangle and rhombus represent the procedure and data, respectively. NURBS non-uniform rational B-splines

In standard DCR for respiratory function analysis, imaging is conducted using a pulsed X-ray at a rate of 15 frames per second (frames/s) over a complete respiratory cycle of approximately 10 s. Similarly, a total of 150 XCAT phantom respiratory phases were generated to simulate a complete respiratory cycle of 10 s, after which 60 frames from maximum exhalation to inhalation over 4 s were extracted for the present study. Assuming that a perfect bone suppression image-processing technique was applied to the resulting projection images, bone structures such as the ribs, clavicles, and scapulae were removed from the XCAT phantoms before the image projection process [38, 39].

Image noise, scatter, and beam hardening effects were not included in 4DCT imaging of the XCAT phantoms, because ΔCT values were calculated from the average CT values across the entire lungs. For the simulated CT images (i.e., noise-free 4DCT imaging of the XCAT phantoms), the lung areas were determined automatically using the activity phantom (binary images) with the lungs defined as pixel value = 0, whereas all other organs were defined as pixel value = 1.

2.3. Creation of projection images

An X-ray simulator was used to acquire projection images of the virtual patient under the same imaging conditions as the actual DCR (100 kV; 0.2 mAs/frame; 15 frames/s; source-to-image receptor distance = 2.0 m) [30]. The XCAT phantoms were projected in the posteroanterior (PA) direction to obtain dynamic chest radiographs of the virtual patient during respiration (matrix size: 512 × 512 pixels [half of the normal matrix size of 1024 × 1024 pixels]; pixel size: 0.834 mm; 32-bit real grayscale), as shown in Fig. 3. In this simulation study, half the normal matrix size was acceptable, because Δpixel values were calculated from the average pixel values measured across the entire lungs. To calculate the Δpixel values of the lung field, we selected frames with diaphragm displacements of −1, −2, −3, and −4 cm from the base frame (0 cm at the time of maximum exhalation), as shown in Fig. 4. To facilitate the pixel value measurement, the projection images were rescaled into 16-bit grayscale images. For all projection images, lung regions were segmented using a U-Net-based segmentation algorithm implemented using Keras with a TensorFlow backend [40], as described in a previous study [41]. This step provided automatic lung segmentation with a Sorenson Dice coefficient of 0.94 [42]. In the event that the automated segmentation failed, it was manually corrected by a single radiologist using the free image processing software ImageJ (ImageJ 1.52p, National Institutes of Health, Bethesda, MD, USA) and MITK (MITK-2016.11.0-win64, German Cancer Research Center, Heidelberg, Germany).

Fig. 3.

Fig. 3

Open in a new tab

Creation of projection images for a given extended cardiac-torso phantom with respiration and heart motion

Fig. 4.

Fig. 4

Open in a new tab

Selection of respiratory phases in the study. We used the frames with a diaphragm displacement of −1, −2, −3, and −4 cm (marked with black points) from the base frame (0 cm at the maximum exhalation time)

2.4. Measurement of Δpixel and ΔCT values

As shown in Fig. 5, the average pixel values across the entire lung were measured on the projection images of the XCAT phantom to calculate the Δpixel values for respiration. Each Δpixel value was calculated as the difference between pixel values during respiration and those at maximum inspiration, i.e., at a diaphragm displacement of 0 cm. The average linear attenuation coefficient of the lungs was measured as the average of the values for all slices of the simulated CT images. This measurement was then converted to CT values to calculate ΔCT values in Hounsfield units (HU).

Fig. 5.

Fig. 5

Open in a new tab

Measurement of the average pixel values and the average linear attenuation coefficient in the lung field on the frames that are selected in Fig. 4. a Dynamic chest radiography and label images to measure the average pixel values, b simulated computed tomography and activity phantom images to measure the average linear attenuation coefficient

ΔCT values were calculated as the differences between CT values during respiration and maximum inspiration. Here, “μtissues” and “μwater” represent the average linear attenuation coefficients in the lung field measured on the simulated CT images and those calculated for a water material, i.e., 0.0481 [per pixel], respectively. The equation (regression line) of the relationship between Δpixel and ΔCT values in the lung field was calculated for each of the 20 XCAT phantoms.

2.5. Evaluation of virtual patients

The fidelity of the XCAT phantom was assessed by comparing the measured ΔCT values with those previously reported in normal lung fields (103 ± 49 HU) [11]. A paired t test was performed to determine whether a statistical difference existed between normal males and females in the inclination of the Δpixel and ΔCT value regression lines, with or without chest wall motion, and between the different body types of male phantoms. Statistical significance was set at P < 0.05.

2.6. Clinical evaluation

The present study was approved by the Medical Ethics Review Committee of Kanazawa University (approval no. 1729). The physician in charge of the study provided the patients with a full explanation of the imaging procedures and obtained their written consent.

The resulting conversion expression was also assessed in DCR images of 37 real patients who visited the departments of respiratory medicine and thoracic surgery at our hospital. Of the 37 patients, 7 (range 53–81 years, median 63 years, male:female ratio = 4:3) with no abnormal pulmonary function tests and clinical findings were classified as the normal controls, and 30 (range 40–85 years, median 72 years, male:female = 25:5) with suspected COPD based on pulmonary function tests were classified as the COPD group.

The diagnostic criteria for COPD were as follows: forced expiratory volume in one second (%FEV1) < 70% and a predicted vital capacity (%VC) > 80% [43]. The measured Δpixel values were correlated with the ΔCT values using the resulting conversion expression combined with chest wall motion. The obtained ΔCT values were then compared with the ΔCT values reported in a previous study (normal: 103 ± 49 HU; obstructive impairment: 52 ± 40 HU) [11].

3. Results

3.1. Results of XCAT phantom

The mean CT values at maximum inspiration and expiration in the lung fields measured in XCAT phantoms were −638.53 ± 62.58 HU and −513.12 ± 57.81 HU, respectively. These measurements resulted in a ΔCT value of 125 ± 31 HU, which was almost within the normal variation of ΔCT values measured in normal controls reported in a previous study (103 ± 49 HU) [11].

3.2. Correlation of Δpixel to ΔCT values

The present preliminary study successfully correlated Δpixel with ΔCT values. Figure 6 shows the relationship between Δpixel and ΔCT values from 20 XCAT phantoms with four different diaphragm displacements. The ΔCT values without chest wall motion were lower than those with chest wall motion. However, the change in ΔCT values was not reflected in the Δpixel values. There was a significant difference in the regression line inclinations between the phantoms with and without chest wall motion (P < 0.05). The r value was 0.72 in the case of those with chest wall motion, which was higher than 0.65 in the case without chest wall motion. Here, the larger regression line inclinations represent smaller Δpixel values based on certain ΔCT values.

Fig. 6.

Fig. 6

Open in a new tab

Relationship between Δpixel and ΔCT values across the entire lung. a With and b without chest wall motion, respectively

Figure 7 shows a box-and-whisker plot showing the variations in regression line inclinations for males and females for each body type. We found no significant differences between normal men and women in regression line inclinations. In the case of those with chest wall motion, there were no significant differences in regression line inclinations between males of different body types and between normal females and males of any body type. However, there was a significant difference between males of different body types without chest wall motion.

Fig. 7.

Fig. 7

Open in a new tab

Box-and-whisker plots showing the variations in regression line inclinations. a With and b without chest wall motion, respectively. The error bar indicates the maximum value in the upper part and the minimum value in the lower part. N.S. not significant

3.3. Clinical evaluation

Using the conversion expression obtained from the XCAT phantom with chest wall motion, the ΔCT values in the lung fields of real patients were converted to 101.96 ± 52.10 HU and 77.57 ± 42.38 HU in the normal controls and the COPD group (N.S.), respectively. We also confirmed that Δpixel values in 81% (30/37) of real patients, 86% (6/7 patients) of normal controls, and 80% (24/30 patients) of COPD patients were correctly correlated with the range of ΔCT values, i.e., the average values ± SD (normal, 103 ± 49 HU; obstructive disease, 52 ± 40 HU), as reported in a previous study [11].

4. Discussion

To better understand the respiratory changes observed in radiographic lung image densities, Δpixel values in the lung fields of DCR images were correlated with the ΔCT values from the lung fields of 4DCT images acquired from computer-based phantoms. We found a significant difference in regression line inclinations between scans with and without chest wall motion, along with an increased r value of those with chest wall motion. These results suggest that inter-individual variations were decreased by chest wall motion. The increased ΔCT values can be explained by the increased Δlung volume due to chest wall displacement in the AP direction. However, the change in the ΔCT values was not likely to be revealed in the Δpixel values owing to the increase in body thickness (i.e., prolonged X-ray path length through the body) in the inspiratory phase, resulting in increased X-ray attenuation. This could be the cause for the increased regression line inclinations of those with chest wall motion, as shown in Fig. 6a. However, there was a significant difference between the different body types of those without chest wall motion. No significant differences were found in regression line inclinations with chest wall motion, suggesting that inter-population variations were also decreased by the chest wall motion. In contrast, we found no significant differences in the regression line inclinations between normal males and females. All the relevant findings suggested that a common conversion expression could be utilized for Δpixel values across an entire lung with DCR, regardless of patient sex or body type, under the condition of thoracic respiration with chest wall motion.

In the present study, 37 cases were used to validate the resulting conversion expression. Of the 37 patients, 6/7 patients (86%) in the normal control group and 24/30 (80%) in the COPD group had converted ΔCT values within the range of intrapulmonary ΔCT values reported in a previous study (normal 103 ± 49 HU; obstructive disease 52 ± 40 HU) [11]. These results suggest that the developed conversion expression was effective for correlating ΔCT values with Δpixel values. Although there were no significant differences in the converted ΔCT values between the normal and COPD groups, we confirmed that the converted ΔCT values in the COPD group tended to be lower than those in the normal group. This result was consistent with the imaging findings on inspiratory and expiratory CT of patients with COPD [11].

The present study had several limitations. First, the conversion expression developed in this study is limited to the application of mean values across entire lung fields. Physiological evidence has confirmed that the ventilation per unit volume of the lung is greatest at the base of the lung and decreases closer to the apex, and this is known as “regional differences in ventilation” [23]. Further studies are needed to develop a conversion expression for subdivided lung regions so that local evaluations of lung properties based on Δpixel values can be performed. Second, in the present study, there was a 2-week interval between the date of pulmonary function tests and DCR. Therefore, a misclassification of DCR (as normal or COPD) due to changes in patient conditions could have occurred, resulting in out-of-range converted ΔCT values in seven clinical cases. Third, the ΔCT values measured in the XCAT phantoms were 125 ± 31 HU, which was not entirely within the reference values from previous studies (103 ± 49 HU) [11]. The difference in the average ΔCT values may be caused by a difference in the inspired volume and imaging conditions. Although little effect on clinical use could be attributed to a large normal variation, further studies are needed to address the reason for and effect of this mismatch. Fourth, the present results did not consider image noise. The variability of CT values under clinical conditions depends on the imaging conditions (kV and/or mA). Therefore, it is necessary to investigate the effect of imaging conditions on the conversion expression. Fifth, only a small number of XCAT phantoms were investigated in the present study. Further studies with larger numbers of XCAT phantoms, including assessment of developed conversion expressions compared to inspiratory/expiratory CT images or 4DCT images from the same patient, are required before the developed conversion expression can be implemented in a clinical setting.

5. Conclusion

We correlated respiratory Δpixel with ΔCT values in the lung field using simulated CT and DCR images of XCAT phantoms. The present study demonstrated the feasibility of quantitatively evaluating pulmonary function based on respiratory changes in radiographic lung densities measured as Δpixel values on DCR images. The developed conversion expression could be helpful for a better understanding of pulmonary function based on DCR.

Acknowledgements

The authors sincerely thank the staff from the departments of respiratory medicine, thoracic surgery, and radiology at Kanazawa University Hospital for providing the data described above.

Funding

The devices used in our clinical study were also provided by Konica Minolta, Inc., Tokyo, Japan. This work was supported in part by Grants-in-Aid from the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) KAKENHI (#16K10271, 19K12841), the Shimadzu Corporation for Science and Technology, and the Tateishi Science and Technology Foundation.

Footnotes

Conflict of interest Our institution has received a research grant from Konica Minolta, Inc., Tokyo, Japan.

Informed consent The present study was approved by the institutional review board of Kanazawa University (registration number 1729). Written informed consent was obtained from all the participants.

References

  • 1.Mathers CD, Loncar D. Projections of global mortality and burden of disease from 2002 to 2030. PLoS Med. 2006;28:209–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Aujesky D, Fine MJ. The pneumonia severity index: a decade after the initial derivation and validation. Clin Infect Dis. 2008;1(47 Suppl 3):S133–9. 10.1086/591394. [DOI] [PubMed] [Google Scholar]
  • 3.Kohno S, Seki M, Watanabe A, CAP Study Group. Evaluation of an assessment system for the JRS 2005: a-DROP for the management of CAP in adults. Intern Med. 2011;50(11):1183–91. [DOI] [PubMed] [Google Scholar]
  • 4.Mortensen J, Berg RMG. Lung scintigraphy in COPD. Semin Nucl Med. 2019;49(1):16–21. [DOI] [PubMed] [Google Scholar]
  • 5.Kawai H, Kawakami T, Tsujimoto M, et al. Prediction of pulmonary function after major lung resection using lung perfusion scintigraphy with single-photon emission computed tomography/computed tomography. Fujita Med J. 2020;6(2):37–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Magkanas E, Voloudaki A, Bouros D, et al. Pulmonary sarcoidosis: correlation of expiratory high-resolution CT findings with inspiratory patterns and pulmonary function tests. Acta Radiol. 2009;42(5):494–501. [PubMed] [Google Scholar]
  • 7.Lee E, Seo JB, Lee HJ, et al. Quantitative assessment of global and regional air trappings using non-rigid registration and regional specific volume change of inspiratory/expiratory CT scans: studies on healthy volunteers and asthmatics. Korean J Radiol. 2015;16(3):632–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Gawlitza J, Trinkmann F, Scheffel H, et al. Time to exhale: additional value of expiratory chest CT in chronic obstructive pulmonary disease. Can Respir J. 2018;2018(7):1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Maglione M, Montella S, Mollica C, et al. Lung structure and function similarities between primary ciliary dyskinesia and mild cystic fibrosis: a pilot study. Ital J Pediatr. 2017;43(1):34–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Kalef-Ezra J, Karantanas A, Tsekeris P. CT measurement of lung density. Acta Radiol. 1999;40(3):333–7. [DOI] [PubMed] [Google Scholar]
  • 11.Kauczor HU, Hast J, Heussel CP, et al. CT attenuation of paired HRCT scans obtained at full inspiratory/expiratory position: comparison with pulmonary function tests. Eur Radiol. 2002;12:2757–63. [DOI] [PubMed] [Google Scholar]
  • 12.Chen H, Chen R, Guan Y, et al. Correlation of pulmonary function indexes determined by low dose MDCT with spirometric pulmonary function tests in patients with chronic obstructive pulmonary disease. Am J Roentgenol. 2004;202(4):711–8. [DOI] [PubMed] [Google Scholar]
  • 13.Beek EJR, Dahmen AM, Stavngaard T, et al. Hyperpolarised 3He MRI versus HRCT in COPD and normal volunteers. PHIL trial. Eur Respir. 2009;34:1311–21. [DOI] [PubMed] [Google Scholar]
  • 14.Egger C, Gerard C, Vidotto N, et al. Lung volume quantified by MRI reflects extracellular-matrix deposition and altered pulmonary function in bleomycin models of fibrosis: effects of SOM230. Am J Physiol Lung Cell Mol Physiol. 2014;306(12):1064–77. [DOI] [PubMed] [Google Scholar]
  • 15.Roach DJ, Cremillieux Y, Serai SD, et al. Morphological and quantitative evaluation of emphysema in chronic obstructive pulmonary disease patients: a comparative study of MRI with CT. J Magn Reson Imaging. 2016;44:1656–63. [DOI] [PubMed] [Google Scholar]
  • 16.Yamamoto T, Kabus S, Lorenz C, et al. Pulmonary ventilation imaging based on 4-dimensional computed tomography: comparison with pulmonary function tests and SPECT ventilation images. Int J Radiat Oncol Biol Phys. 2014;90(2):414–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Yamashiro T, Moriya H, Tsubakimoto M, et al. Continuous quantitative measurement of the proximal airway dimensions and lung density on four-dimensional dynamic-ventilation CT in smokers. Int J Chron Obstruct Pulmon Dis. 2016;11:755–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Fujita H, Doi K, MacMahon H. Basic imaging properties of a large image intensifier—TV digital chest radiographic system. Investig Radiol. 1987;22:328–35. [DOI] [PubMed] [Google Scholar]
  • 19.Lam KL, Chan HP, MacMahon H, et al. Dynamic digital substruction evaluation of regional pulmonary ventilation with nonradioactive xenon. Invest Radiol. 1990;25:728–35. [DOI] [PubMed] [Google Scholar]
  • 20.Kiuru A, Svedstrom E, Kuuluvainen I. Dynamic imaging of pulmonary ventilation. Description of a novel digital fluoroscopic system. Acta Radiol. 1991;32:114–9. [PubMed] [Google Scholar]
  • 21.Kiuru A, Svedstrom E, Korvenranta H. Dynamic pulmonary imaging: performance properties of a digital fluoroscopy system. Med Phys. 1992;19:467–73. [DOI] [PubMed] [Google Scholar]
  • 22.Tanaka R, Sanada S, Fujimura M, et al. Development of functional chest imaging with a dynamic flat-panel detector (FPD). Radiol Phys Technol. 2008;1:137–43. [DOI] [PubMed] [Google Scholar]
  • 23.Tanaka R Dynamic chest radiography: flat-panel detector (FPD) based functional X-ray imaging. Radiol Phys Technol. 2016;9:139–53. [DOI] [PubMed] [Google Scholar]
  • 24.Tanaka R, Sanada S, Fujimura M, Yasui M, Tsuji S, Kayashi N, Okamoto H, Nanbu Y, Matsui O. Ventilatory impairment detection based on distribution of respiratory-included changes in pixel values in dynamic chest radiography: a feasibility study. Int J CARS. 2011;6(1):103–10. [DOI] [PubMed] [Google Scholar]
  • 25.Yamada Y, Ueyama M, Abe T, et al. Difference in the craniocaudal gradient of the maximum pixel value change rate between chronic obstructive pulmonary disease patients and normal subjects using sub-mGy dynamic chest radiography with a flat panel detector system. Eur J Radiol. 2017;92:37–44. [DOI] [PubMed] [Google Scholar]
  • 26.Tanaka R, Matsumoto I, Tamura M, et al. Comparison of dynamic flat-panel detector-based chest radiography with nuclear medicine ventilation-perfusion imaging for the evaluation of pulmonary function: a clinical validation study. Med Phys. 2020;47(10):4800–9. [DOI] [PubMed] [Google Scholar]
  • 27.Tanaka R, Tani T, Nitta N, et al. Pulmonary function diagnosis based on respiratory changes in lung density with dynamic flat-panel detector imaging: an animal-based study. Invest Radiol. 2018;53(7):417–23. [DOI] [PubMed] [Google Scholar]
  • 28.Tanaka R, Tani T, Nitta N, Tabata T, Matsutani N, Muraoka S, Yoneyama T, Sanada S. Pulmonary function diagnosis based on respiratory changes in lung density with dynamic flat-panel detector imaging: an animal-based study. Invest Radiol. 2018;53(7):417–23. [DOI] [PubMed] [Google Scholar]
  • 29.Abadi E, Segars WP, Tsui BMW, et al. Virtual clinical trials in medical imaging: a review. J Med Imaging. 2020;7(4):042805. 10.1117/1.JMI.7.4.042805. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Segars WP, Sturgeon G, Mendonca S, et al. 4D XCAT phantom for multimodality imaging research. Med Phys. 2010;37(9):4902–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Segars WP, Mahesh M, Beck TJ, et al. Realistic CT simulation using the 4D XCAT phantom. Med Phys. 2008;35:3800–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Segars WP, Tsui BMW. MCAT to XCAT: the evolution of 4-D computerized phantoms for imaging research. Proc IEEE Inst Electr Electron Eng. 2009;97(12):1954–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Segars WP, Bond J, Frush J, Hon S, Eckersley C. Population of anatomically variable 4D XCAT adult phantoms for imaging research and optimization. Med Phys. 2013;40(4):043701 (p 11). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Tanaka R, Samei E, Segars WP, Abadi E, Matsumoto I, Tamura M, Takata M, Yamashiro T. Preoperative assessment of pleural invasion and adhesion of lung tumors with dynamic chest radiography: a virtual clinical trial. Med Phys. 2020;48(4):1616–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Visible Human Male and Female datasets, National Library of Medicine. http://www.nlm.nih.gov/research/visible/visible_human.html.
  • 36.ICRU. International commission on radiation units and measurement, photon, electron, proton and neutron interaction data for body tissues. ICRU Report No 46. Bethesda: ICRU; 1992. [Google Scholar]
  • 37.Obesity: preventing and managing the global epidemic. Report of a WHO Consultation on obesity. Geneva, 3–5, June 1997. [PubMed] [Google Scholar]
  • 38.Suzuki K, Abe H, MacMahon H, Doi K. Image-processing technique for suppressing ribs in chest radiographs by means of massive training artificial neural network (MTANN). IEEE Trans Med Imaging. 2006;25:406–16. [DOI] [PubMed] [Google Scholar]
  • 39.Knapp J, Worrell S. Feature based neural network regression for feature suppression. U.S. Patent Number, 204,292 B2, June 12, 2012.
  • 40.Olaf R, Fischer P, Brox T. U-net: convolutional networks for biomedical image segmentation. In: International conference on medical image computing and computer-assisted intervention. Cham: Springer; 2015. [Google Scholar]
  • 41.Kitahara Y, Tanaka R, Holger R, Oda H, Mori K, Kasahara K, Matsumoto I. Lung segmentation based on a deep learning approach for dynamic chest radiography. In: Proc. SPIE 10950. 2019. [Google Scholar]
  • 42.Skourt BA, Hassani AE, Majda A. Lung CT image segmentation using deep neural networks. Proc Comput Sci. 2018;127:109–13. [Google Scholar]
  • 43.Pauwels RA, Buist AS, Calverley PM, Jenkins CR, Hurd SS, GOLD Scientific Committee. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease. NHLBI/WHO Global Initiative for Chronic Obstructive Lung Disease (GOLD) Workshop summary. Am J Crit Care Respir Med. 2001;163:1256–76. [DOI] [PubMed] [Google Scholar]

RESOURCES