Study population: Dynamic chest radiographs of 10 patients (Abnormal,
n=4; Normal,
n=6) were obtained in this study.
The patients had been diagnosed to be normal (51–84 years old; mean,
58 years old; M:F=2:4) or abnormal (31–73 years old; mean,
69 years old; M:F=4:0) in terms of pulmonary blood circulation based on findings on CT,
lung perfusion scintigraphy,
and other clinical findings.
Approval for the study was obtained from our institutional review board,
and the patient gave his written informed consent to participation in this study.
Image acquisition: Dynamic chest radiographs were obtained using a portable dynamic FPD system (CXDI-50RF; Canon,
Japan) (Figure 1),
which was capable of both radiography and fluoroscopy.
Fifty images were obtained in 10 seconds on exposure conditions as follows: 120 kV,
0.1 mAs/frame,
SID 1.5 m,
5.0 frames per second (fps) (Figure 2).
Imaging was performed during respiration,
in the standing position,
and posteroanterior (PA) direction.
The total exposure dose was less than the amount of those in conventional chest radiography in the two projections (PA + LA).
Low pixel values were related to dark areas in the images,
and these in turn were related to high X-ray translucency in this study.
The matrix size was 2208 × 2688 pixels,
the pixel size was 160 × 160 mm,
the field of view was 35 × 43 cm,
and the grayscale range of the images was 16 bits.
Visualization of blood cirulation: Figure 3 shows the overall scheme of our algorithm.
The images composing a whole cardiac phase during breath-holding were determined based on the movement of diaphragm and left ventricular wall,
and one cardiac cycle was selected for the analysis.
For visual evaluation,
inter-frame differences were calculated and mapped on the original image as blood circulation vector map (Figure 4) [10,11].
The absolute values were then summed through a single cardiac phase,
and the results were visualized in color scale as a blood circulation map (Figure 5).
Statistical analysis: Regions of interest (ROIs) were manually located on the lung and cardiac areas,
and the maximum differences in pixel values during a single cardiac phase were measured in each area.
To address the relationship among the hilar and peripheral regions at the hilar level,
right and left in symmetrical positions at each lung area,
and three lung areas at each lung,
we examined the null hypothesis that there were no differences among the measurement results of each area by Wilcoxon test (Figure 6).
Clinical evaluation: To evaluate the diagnostic performance,
the resulting images were compared to perfusion scintigrams in patients with pulmonary blood flow impairments.