Visual analysis of pulmonary blood flow in pulmonary circulation assessment: differences between two variant algorithms for processing dynamic images

Introduction

Recent advancements in medical imaging technology have enabled the assessment of spatial and temporal information derived from the time-resolved anatomical imaging of pulmonary perfusion and hemodynamics (1). Compared to three-dimensional (D) computed tomography, pulmonary arteriography, and pulmonary perfusion scintigraphy (PPS) with ionizing radiation exposure (2), magnetic resonance imaging (MRI) has enabled the qualitative and quantitative evaluation of pulmonary circulation through pulmonary perfusion and hemodynamic imaging (1). Advancements in MRI technology, notably in 4D-flow MRI, phase-contrast MRI, and arterial spin-labeling MRI, have enabled the quantitative evaluation of various parameters without using contrast agents. These include blood flow within the cardiac chambers or pulmonary arteries (3), hemodynamic indices (such as average flow, average velocity, and acceleration time) (4), pulmonary perfusion throughout the entire lung field (5), local pulmonary transit time (6), and left-to-right blood flow ratio (BFR) (7). However, capital investment and inspection time constraints prevent its widespread implementation. Therefore, minimally invasive, low-cost, easy-to-perform techniques are required.

Dynamic chest radiography (DCR), which can be performed simultaneously with standard chest radiography with low radiation exposure, has recently emerged as a potential solution in clinical settings (8,9). The correlation between the magnitude of the pixel value changes in moving images and degree of pulmonary blood circulation has been confirmed in animal experiments (10). Consequently, dynamic perfusion digital radiography (DPDR) using images obtained by DCR enables visual qualitative evaluation of pulmonary blood flow distribution using color-coded changes in pixel values and quantitative assessment using the left-right BFR without the injection of contrast agents (8,11). The prototype dynamic imaging system (Konica Minolta, Inc., Tokyo, Japan) uses two algorithms to generate the DPDR: cross-correlation calculation processing (CCC-pro) (11,12) and reference frame subtraction processing (RFS-pro) (13). The cross-correlation value from the degree of waveform correlation between the changes in the pixel values in the pulmonary and ventricular regions was evaluated for CCC-pro.

In contrast, the temporal changes in the pixel values from the end-diastolic phase for each pixel were evaluated for RFS-pro. In previous reports, perfusion images processed by either algorithm could visually diagnose pulmonary thromboembolism (10,14). Furthermore, a strong correlation has been reported between each algorithm and the PPS for left-right BFR, a quantitative evaluation (11,12). Although different aspects of pulmonary blood flow have been evaluated using two different algorithms, there have yet to be any reports discussing the differences in blood flow evaluation resulting from them.

This study aimed to evaluate the differences in the analysis of pulmonary circulation using blood flow imaging processed from dynamic images using two different algorithms. Pulmonary perfusion imaging using DCR, which can be easily performed with low radiation exposure and a short examination time, is highly useful in clinical practice. Therefore, it is essential to clarify the appropriate use of these algorithms. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-24-152/rc).

Methods

Patients

Among patients scheduled for radical resection of primary lung cancer at Shiga University of Medical Science from May 2018 to December 2020, we consecutively enrolled 100 patients who met the following criteria: ability to hold breath as instructed, no history of thoracic surgery, ≥20-years-old, and no risk of adverse events due to irradiation. This cohort underwent preoperative DCR as a standard procedure. However, PPS was not mandatory and was performed only in cases deemed necessary by outpatient physicians. In this cross-sectional study, we simultaneously selected patients who underwent DCR and PPS to confirm the differences in pulmonary perfusion imaging using two different algorithms. We analyzed 46 patients who underwent both preoperative DCR and PPS, excluding one patient who was diagnosed with an inflammatory lung tumor instead of lung cancer postoperatively (Figure 1).

Figure 1 Flow diagram of this study. The final analysis included 46 patients who underwent simultaneous preoperative DPDR and PPS. DPDR, dynamic perfusion digital radiography; PPS, pulmonary perfusion scintigraphy.

Table 1 shows the clinical variables collected from the patients’ electronic medical records. Data from this patient population have been previously used in related studies (11,12). This study was conducted per the Declaration of Helsinki (as revised in 2013). This study was approved by the Institutional Review Board of the Shiga University of Medical Science (No. CRB 5180008; October 10, 2017). Written informed consent was obtained from all individual participants.

DCR imaging protocol

Using a prototype system (Konica Minolta, Inc.) comprising an indirect-conversion flat-panel detector (PaxScan, 4343CB, Varex Imaging Corporation, Salt Lake City, UT, USA), an X-ray tube (RAD-94/B-130H, Varian Medical Systems, Inc., Palo Alto, CA, USA), and a pulsed X-ray generator (EPS45RF, EMD Technologies, Saint-Eustache, Canada) we performed pulmonary perfusion analysis (11,12). According to automatic voice guidance, the patients were scanned at a pulse rate of 15 fps during various breathing patterns while sitting. The study used 7 s images during breath-holding. The entrance surface dose for this breath-holding scan was 0.56 mGy, below the International Atomic Energy Agency guidance level for both posterior-anterior and lateral chest radiographs.

Pulmonary perfusion analysis

All images were analyzed using image processing software for DCR (Prototype Model; Konica Minolta, Inc.), which incorporated two algorithms for imaging pulmonary blood flow using pixel value changes extracted from regions of interest (ROIs) set in arbitrary lung and ventricular regions (Figure 2A). The ROI for the lung field covered the entire lung. In contrast, the ROI for the ventricular region was manually delineated with 50×50-pixel rectangular ROIs, following specific placement rules to avoid overlap and ensure proper boundary delineation. One algorithm, CCC-pro (Figure 2B), was used for pulmonary blood flow analysis using the following method (11,12). After removing the pixel value changes corresponding to the respiratory cycle with a high-pass filter, the cross-correlation value from the degree of waveform correlation between the changes in pixel values in the pulmonary and ventricular regions was calculated and color-coded to visualize pulmonary blood flow over the entire lung field. Another algorithm, RFS-pro (Figure 2C), was implemented using the following method (13). After extracting the pixel value changes corresponding to the heartbeat cycle using a bandpass filter, temporal changes in the pixel values from the end-diastolic phase for each pixel were calculated and color-coded to visualize the dynamic perfusion images.

Figure 2 Two different algorithms for imaging pulmonary blood flow. (A) Temporal changes in pixel values for the ROIs set in arbitrary lung and ventricular regions are calculated from sequential images. (B) In CCC-pro, the cross-correlation value from the degree of waveform correlation between the changes in the pixel values in the pulmonary and ventricular regions is calculated as follows. Step 1: the value change for each pixel is calculated from dynamic images. Step 2: the pixel value changes, corresponding to a respiratory cycle, are removed using a high-pass filter. A cutoff frequency of 0.85 Hz extracts periodic pixel value changes, corresponding to a cardiac cycle. Step 3: the degree of correlation between each pixel in the lung field and reciprocal pixel value change in the ventricular region ROI is calculated as a CCv. Step 4: the phase is shifted by one frame, and the CCv is calculated for 31 frames. Each frame’s CCv (−1.0 to 1.0) was color-coded and visualized as DPDR. When blood flow is present, the CCv is high; when blood flow is absent, the CCv is low. (C) In RFS-pro, each pixel’s temporal value change from the end-diastolic phase is calculated as follows. Step 1: the value change for each pixel is calculated from dynamic images. Step 2: the timing of the minimum pixel value, which represents the maximum blood volume on the ventricle ROI, is defined as the end-diastolic phase. Step 3: the temporal value change from the end-diastolic phase is calculated and visualized as DPDR. An increase in blood volume from the end-diastolic phase is represented by red to yellow. A decrease is represented by blue to cyan. Black means no change. ROI, region of interest; CCC-pro, cross-correlation calculation processing; CCv, cross-correlation value; RFS-pro, reference frame subtraction processing; DPDR, dynamic perfusion digital radiography.

BFR measurement

To compare the distribution of lung perfusion between the DCR and PPS groups, each lung was divided into upper, middle, and lower areas, and six areas were assessed.

In CCC-pro, each pixel’s maximum cross-correlation calculation value (MaxCCv) is calculated for all frames, thereafter the sum of the MaxCCv values for the six ROIs is calculated (SumMaxCCv). The BFR was defined as follows in Eq. [1].

BFRineachROI=SumMaxCCvineachROISumMaxCCvinalllungfields[1]

In RFS-pro, each pixel’s minimum reference frame subtraction value (MinRFSv) is calculated for all frames, thereafter the sum of the MinRFSv values for the six ROIs is calculated (SumMinRFSv). The BFR in each ROI was defined as follows in Eq. [2].

BFRineachROI=SumMinRFSvineachROISumMinRFSvinalllungfields[2]

PPS

PPS was performed using a dual-head variable-angle gamma camera (Discovery630; GE Healthcare Life Sciences, Buckinghamshire, England) with high-resolution, low-energy collimators. Each patient was initially administered half of the 200-MBq 99mTc microalbumin/99mTc-macroaggregate solution in the prone position, followed by the remaining half with the patient in the supine position. Subsequently, planar scans were acquired and the images were processed for quantitative perfusion analysis. Geometric mean values were calculated from both lungs’ anterior-to-posterior and posterior-to-anterior projections and divided into six ROIs.

Pixel values histogram

We compared the histograms that illustrated changes in the vertical pixel values for blood flow images generated using two different variants of the DPDR algorithm and PPS in the bilateral lung fields. In DPDR, the frame with the highest blood flow distribution in the lung field was selected from all dynamic images. Blood-flow images were generated from the same frame using two different processing algorithms. The contours of the lung fields were determined by referencing a chest radiograph obtained during inspiratory breath-holding, and histograms were generated using image processing software for the prototype model (Konica Minolta, Inc.). In PPS, a histogram is generated by setting a region surrounding the dots of the lung field in the frontal planar reverse image.

Statistical analysis

Statistical analyses were performed using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria). First, we assessed the normality of continuous data using the Shapiro-Wilk test. Subsequently, correlations between the BFRs calculated using the two variants of the DPDR algorithm and PPS pre-surgery were evaluated using Pearson’s correlation and linear regression analysis. Additionally, Bland-Altman analysis (15) was used to compare the BFRs calculated using the two DPDR algorithms and PPS variants. Limits of agreement were drawn as the mean difference between predicted and measured values ±2 standard deviations (SD) of the differences. Categorical variables are presented as [n (%)]. All statistical analyses were two-tailed, and statistical significance was set at P<0.05.

Results

Clinical characteristics

Table 1 shows the clinical characteristics of the 46 patients obtained from their electronic medical records. The mean age of the patients was 72.91 years. Forty patients (87.0%) were male. Among the 14 patients with respiratory comorbidities, 12 had a history of chronic obstructive pulmonary disease (COPD) treatment, and interstitial pneumonia was detected in three patients, including one with overlapping COPD. Sixteen patients (34.8%) had stage 2 COPD as measured by spirometry. Twenty-eight patients had cardiovascular comorbidities, primarily hypertension.

Correlation among blood flow distributions using two variants of DPDR algorithms and PPS

Figure 3A-3C shows the linear regression analysis of the affected side-to-total ratio obtained using the two DPDR algorithms and PPS variants. Although CCC-pro and RFS-pro correlated (R=0.718, P<0.01), a stronger correlation was observed between CCC-pro and PPS (R=0.859, P<0.01) than between RFS-pro and PPS (R=0.548, P<0.01). Bland-Altman plots were used to assess the consistency of each BFR, and over 90% of the plots for any measurement methods combination fell within the limits of agreement, indicating substantial agreement (Figure 3D-3F). However, systematic errors were absent in the relationship between RFS-pro and PPS but were present in other relationships. CCC-pro to PPS analysis revealed a proportional error and mean difference of −0.20% for 91.3% of values within 2 SDs of the mean (Figure 3D). In comparison, analysis of CCC-pro to RFS-pro showed no mean difference for 95.7% of values within 2 SDs of the mean and a proportional error (Figure 3F).

Figure 3 Comparison between BFR from DPDR with two variants of DPDR algorithm and PPS. (A-C) Correlation between the BFR on the affected side obtained from two variants of DPDR algorithms and PPS. (D-F) Bland-Altman analysis between the affected side BFR was obtained from two DPDR algorithms and PPS variants. The black line indicates the mean, dotted lines indicate the limit of agreement, and broken lines indicate mean ±2 SD. CCC-pro, cross-correlation calculation processing; PPS, pulmonary perfusion scintigraphy; RFS-pro, reference frame subtraction processing; LOA, limit of agreement; BFR, blood flow ratio; DPDR, dynamic perfusion digital radiography; SD, standard deviation.

Relationship between BFRs calculated from each method in six-division lung fields (Figure 4)

Figure 4 Scatter plot of blood flow ratios calculated from each method in six-division lung fields. (A) CCC-pro and PPS. (B) RFS-pro and PPS. (C) CCC-pro and RFS-pro. CCC-pro, cross-correlation calculation processing; RU, right upper; RM, right middle; RL, right lower; LU, left upper; LM, left middle; LL, left lower; PPS, pulmonary perfusion scintigraphy; RFS-pro, reference frame subtraction processing.

Regarding the relationship between PPS and CCC-pro, PPS was dominant in the middle and lower lung fields, whereas CCC-pro was dominant in the upper lung field in both lungs. Regarding the relationship between the PPS and RFS-pro, both methods showed a high BFR in the middle lung field, with RFS-pro being dominant. The BFR was higher in the upper lung field in RFS-pro and lower lung field in PPS, without any significant left-right difference. Furthermore, both methods showed a high BFR in the middle lung field, especially RFS-pro, on the left side, and the BFR in the lower lung field was slightly higher in PPS. Regarding the relationship between CCC-pro and RFS-pro, both methods showed similar blood flow distributions in the upper and lower lung fields, and both methods had high BFR in the middle lung field, with RFS-pro being dominant.

Comparison of blood flow imaging histograms

Figure 5 illustrates the representative histograms obtained using each imaging method. In all the images, the lower left lung field exhibited low pixel values because of the small size of the lung field caused by the cardiac shadow. In CCC-pro, pixel values were similarly high in the upper and middle lung fields, whereas those in the paradiaphragmatic and paracardiac regions were low because of the effects of the heartbeat. In RFS-pro, the pulmonary artery and vein or the aorta in the hilar region showed significantly high pixel values.

Figure 5 Comparison of histograms based on perfusion images obtained from each method. The vertical axis of the histogram represents the level of the corresponding lung field, and the horizontal axis represents the pixel value (the pixel value increases in the direction of the arrow). The center shows the original pulmonary perfusion image, whereas the left and right sides display the corresponding histograms of the lung fields. T, top; B, bottom; CCC-pro, cross-correlation calculation processing; RFS-pro, reference frame subtraction processing; PPS, pulmonary perfusion scintigraphy.

Considering the pixel value overshoot in the hilar region, the shape of the histogram was more similar to that of the PPS in RFS-pro than in CCC-pro. Comparing the pixel value changes in the right upper lung field for each imaging method, a difference in the gradient of the histogram was observed depending on body posture.

Comparison of actual pulmonary blood flow images

Factors affecting pulmonary blood flow were observed in actual pulmonary blood flow images. Video 1A shows the cases in which the left-to-right BFRs of CCC-pro, RFS-pro, and PPS were similar. In RFS-pro, the pulmonary artery region, which had the greatest impact on the BFR, was located symmetrically, resulting in no difference in the BFR between the left and right sides. In contrast, Video 1B shows cases in which the left-to-right BFRs of CCC-pro, RFS-pro, and PPS were different. In RFS-pro, the left main pulmonary artery was subtracted by overlapping it with the cardiovascular shadow. Therefore, the high pixel value of the right main pulmonary artery affected the BFR.

Discussion

This study examined whether images processed using two dynamic imaging system algorithms were suitable for pulmonary blood flow analysis. Regarding the BFR of the affected lungs of patients scheduled for lung cancer surgery, a high correlation was observed in CCC-pro between the two image processing algorithms and PPS, which were used for blood flow evaluation. However, the Bland-Altman analysis revealed a slightly higher number of cases included in the limits of agreement for RFS-pro than for CCC-pro, and no systematic error was observed. When comparing the PPS with each algorithm using histograms, the image from RFS-pro reflected the body position and was more similar to that of PPS, especially when considering the effect of pixel value overshoot in the pulmonary vessels at the hilum. The potential implication of this overshoot on the imaging of the hilar vessels suggests a possible effect on the ratio of pulmonary blood flow between CCC-pro and RFS-pro.

These two algorithms differ in their processing methods. CCC-pro measures the correlation of heartbeats between the ventricular region and any lung-field ROI. In contrast, RFS-pro captures changes in pixel values in any lung-field ROI during the cardiac cycle. Considering that changes in pixel values reflect the pulmonary arterial blood flow volume (10,16), RFS-pro is expected to represent the pulmonary blood flow volume more accurately than CCC-pro. However, when comparing the BFRs, strong correlations with PPS were not observed. PPS counts and images of the distribution of radioisotopes trapped in the capillaries indicate that arteries with diameters more extensive than those of the capillaries are not reflected in the images (17).

Conversely, RFS-pro strongly reflects the pixel value changes in the main pulmonary artery, where the blood flow volume is the highest during systole, representing the blood flow volume more accurately. Nonetheless, the course of the pulmonary artery, overlapped with the heart and great vessels, and twisting of the body during imaging affects the blood flow evaluation in RFS-pro. Since pulsation is maintained even at the capillary level, similar to that in arterioles (18), the assessment of CCC-pro for visualizing correlations of vascular pulsation reflects the blood flow distribution from the center to the periphery, particularly in the presence or absence of blood flow. Despite the patient being seated, the high histogram in the upper lung field did not directly reflect the blood flow volume. These differences in the images’ blood flow distribution and BFR could be attributed to the factors above. Therefore, CCC-pro reflects blood flow distribution of from the central to peripheral regions, particularly the presence or absence of blood flow; RFS-pro reflects blood volume, precisely the peak value of instantaneous blood volume; and PPS reflects blood distribution in the peripheral regions. Although, BFR comparison favored CCC-pro with improved processing methods for hilar vessels in RFS-pro, an evaluation combining blood flow distribution and BFR is possible.

The potential and usefulness of lung blood flow evaluation using DCR-based DPDR were previously reported (11,12). DCR-based DPDR is minimally invasive and requires a low radiation dose, making it suitable for repeated imaging and early postoperative evaluation of changes in lung blood flow (8,9). However, while DPDR demonstrates excellent qualitative assessment, quantitative blood flow evaluation for each side’s blood flow presents challenges. Hence, DPDR is primarily a qualitative assessment of the left-to-right BFR. A previous study reported a high correlation between the BFR in the left and right pulmonary arteries measured by MRI and pulmonary blood flow scintigraphy (7). However, recently, performing quantitative blood flow evaluation in lung fields using ROI without needing contrast agents is possible (5,6). Free-breathing phase-resolved functional lung MRI allows the assessment of ventilation and perfusion of the lung field (19). Integrating MRI assessment with DPDR lung perfusion imaging facilitates a more detailed evaluation of early postoperative lung resection and a more straightforward assessment of circulatory dynamics and respiratory pathophysiological changes in other conditions. Further progress in dynamic imaging systems will contribute significantly to these developments.

This study has some limitations. First, the dynamic imaging system used in this investigation was a prototype device installed within our facility, which may have imaging conditions and device specifications different from those of commercially available systems. Although a direct comparison of the impact of differences in imaging conditions and specifications on image quality has yet to be conducted, RFP-pro may underestimate the blood flow owing to signal attenuation near the chest wall. Therefore, caution should be exercised when evaluating the study outcomes. A possible way to gain further insight would be to compare pulmonary blood flow imaging and BFR between the prototype and commercially available systems using the actual dynamic images used in this study. Second, this was a single-center study with a small sample size.

Further research, employing commercially available systems, is required to validate our findings. Furthermore, a more extensive multicenter study incorporating ground-truth data is imperative to facilitate the adoption of this device in mainstream clinical settings. Finally, by assessing the adaptation of the two DPDR algorithms variants and their evolution, we anticipate that this research will contribute to a better understanding of respiratory physiology and pathophysiology associated with pulmonary circulation.

Conclusions

Pulmonary circulation evaluation using two different dynamic imaging system algorithms revealed that CCC-pro was useful for quantitative BFR evaluation. In contrast, RFS-pro enabled the accurate assessment of blood flow distribution in lung fields. Further improvements are required to establish algorithms to facilitate a more versatile analysis of pulmonary blood flow.

Acknowledgments

We would like to thank Editage (www.editage.jp) for English language editing. Katsunori Miyata, a radiological technician at our institution, contributed substantially by assisting with DCR.

Funding: This work was supported by Konica Minolta, Inc.

Footnote

Reporting Checklist: The author has completed the STROBE reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-24-152/rc

Conflicts of Interest: The author has completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-24-152/coif). J.H. received a research grant from Konica Minolta Inc. The author has no other conflicts of interest to declare.

Ethical Statement: The author is accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The study was approved by the Institutional Review Board of Shiga University of Medical Science (No. CRB 5180008; October 10, 2017). Written informed consent was taken from all individual participants.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Ohno Y, Hanamatsu S, Obama Y, Ueda T, Ikeda H, Hattori H, Murayama K, Toyama H. Overview of MRI for pulmonary functional imaging. Br J Radiol 2022;95:20201053. [Crossref] [PubMed]
2. Johns CS, Swift AJ, Hughes PJC, Ohno Y, Schiebler M, Wild JM. Pulmonary MR angiography and perfusion imaging-A review of methods and applications. Eur J Radiol 2017;86:361-70. [Crossref] [PubMed]
3. Cain MT, Schäfer M, Ross LK, Ivy DD, Mitchell MB, Fenster BE, Bull TM, Barker AJ, Vargas D, Hoffman JRH. 4D-Flow MRI intracardiac flow analysis considering different subtypes of pulmonary hypertension. Pulm Circ 2023;13:e12307. [Crossref] [PubMed]
4. Tsuchiya N, Iwasawa T, Ogura T, Yamashiro T, Yara S, Fujita J, Murayama S. Pulmonary flow assessment by phase-contrast MRI can predict short-term mortality of fibrosing interstitial lung diseases. Acta Radiol 2020;61:1350-8. [Crossref] [PubMed]
5. Addo DA, Kang W, Prisk GK, Tawhai MH, Burrowes KS. Optimizing human pulmonary perfusion measurement using an in silico model of arterial spin labeling magnetic resonance imaging. Physiol Rep 2019;7:e14077. [Crossref] [PubMed]
6. Moore JE, Cerne JW, Pathrose A, Veer M, Sarnari R, Ragin A, Carr JC, Markl M. Quantitative Assessment of Regional Pulmonary Transit Times in Pulmonary Hypertension. J Magn Reson Imaging 2023;57:727-37. [Crossref] [PubMed]
7. Hafiz R, Kocaoglu M, Trout AT. Comparison of phase contrast magnetic resonance imaging and scintigraphy for determination of split pulmonary blood flow in children and young adults with congenital heart disease. Pediatr Radiol 2023;53:2040-7. [Crossref] [PubMed]
8. Tanaka R. Dynamic chest radiography: flat-panel detector (FPD) based functional X-ray imaging. Radiol Phys Technol 2016;9:139-53. [Crossref] [PubMed]
9. Hata A, Yamada Y, Tanaka R, Nishino M, Hida T, Hino T, Ueyama M, Yanagawa M, Kamitani T, Kurosaki A, Sanada S, Jinzaki M, Ishigami K, Tomiyama N, Honda H, Kudoh S, Hatabu H. Dynamic Chest X-Ray Using a Flat-Panel Detector System: Technique and Applications. Korean J Radiol 2021;22:634-51. [Crossref] [PubMed]
10. Miyatake H, Tabata T, Tsujita Y, Fujino K, Tanaka R, Eguchi Y. Detection of Pulmonary Embolism Using a Novel Dynamic Flat-Panel Detector System in Monkeys. Circ J 2021;85:361-8. [Crossref] [PubMed]
11. Hanaoka J, Shiratori T, Okamoto K, Kaku R, Kawaguchi Y, Ohshio Y, Sonoda A. Reliability of dynamic perfusion digital radiography as an alternative to pulmonary perfusion scintigraphy in predicting postoperative lung function and complications. J Thorac Dis 2022;14:3234-44. [Crossref] [PubMed]
12. Hanaoka J, Yoden M, Hayashi K, Shiratori T, Okamoto K, Kaku R, Kawaguchi Y, Ohshio Y, Sonoda A. Dynamic perfusion digital radiography for predicting pulmonary function after lung cancer resection. World J Surg Oncol 2021;19:43. [Crossref] [PubMed]
13. Yamasaki Y, Abe K, Kamitani T, Hosokawa K, Hida T, Sagiyama K, Matsuura Y, Baba S, Isoda T, Maruoka Y, Kitamura Y, Moriyama S, Yoshikawa H, Fukumoto T, Yabuuchi H, Ishigami K. Efficacy of Dynamic Chest Radiography for Chronic Thromboembolic Pulmonary Hypertension. Radiology 2023;306:e220908. [Crossref] [PubMed]
14. Yamasaki Y, Abe K, Hosokawa K, Kamitani T. A novel pulmonary circulation imaging using dynamic digital radiography for chronic thromboembolic pulmonary hypertension. Eur Heart J 2020;41:2506. [Crossref] [PubMed]
15. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307-10. [Crossref] [PubMed]
16. Tanaka R, Sanada S, Tsujioka K, Matsui T, Takata T, Matsui O. Development of a cardiac evaluation method using a dynamic flat-panel detector (FPD) system: a feasibility study using a cardiac motion phantom. Radiol Phys Technol 2008;1:27-32. [Crossref] [PubMed]
17. Tow DE, Wagner HN Jr, Lopez-Majano V, Smith EM, Migita T. Validity of measuring regional pulmonary arterial blood flow with macroaggregates of human serum albumin. Am J Roentgenol Radium Ther Nucl Med 1966;96:664-76. [Crossref] [PubMed]
18. LEE GD. DUBOIS AB. Pulmonary capillary blood flow in man. J Clin Invest 1955;34:1380-90. [Crossref] [PubMed]
19. Pöhler GH, Klimeš F, Behrendt L, Voskrebenzev A, Gonzalez CC, Wacker F, Hohlfeld JM, Vogel-Claussen J. Repeatability of Phase-Resolved Functional Lung (PREFUL)-MRI Ventilation and Perfusion Parameters in Healthy Subjects and COPD Patients. J Magn Reson Imaging 2021;53:915-27. [Crossref] [PubMed]