| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 All authors: Department of Radiology, St. Marianna University School of Medicine, 2-16-1 Sugao, Miyamae-Ku, Kawasaki, Kanagawa, 216-8511, Japan.
Received January 10, 2005;
accepted after revision March 8, 2005.
Address correspondence to S. Matsuoka
(shinma{at}d9.dion.ne.jp).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Twenty-six patients with PE and 11 patients without PE who underwent CTA were evaluated with frequency histograms. We obtained quantitative parameters such as mean lung attenuation, median lung attenuation, SD, skewness, kurtosis, and the proportion of lung attenuation except for the median ± 50 H (P ± 50 H). Lung attenuation was also assessed visually and scored. The relationship between those histogram parameters, or visual score, and PaO2 was evaluated. CTA scores for evaluation of the degree of pulmonary artery obstruction were obtained, and the relation with PaO2 was assessed.
RESULTS. No significant differences were found in mean lung attenuation and median lung attenuation between patients with and without PE. Meanwhile, SD, skewness, kurtosis, and P ± 50 H were significantly different between patients with and without PE (p = 0.0003, 0.0071, 0.0047, and 0.0028, respectively) and significantly correlated with PaO2 (r = -0.770, 0.797, 0.786, -0.871, respectively). Significant differences were found in visual scores between patients with and without PE (p < 0.0001). There were significant but relatively low correlations between CTA score and arterial blood gas levels (r = -0.442, p = 0.03).
CONCLUSION. In patients with acute PE, heterogeneity in lung attenuation is more prominent than in patients without PE.
Keywords: CT • CT angiography • embolism • lung
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In patients with acute PE, various pathophysiologic changes such as regional alterations of pulmonary blood flow [3, 4], bronchoconstriction [5-7], edema [8], and infarction [9, 10] are seen. Lung attenuation values on CT images are affected by reduced pulmonary perfusion [11-14], bronchoconstriction [15], and increased intravascular volume or edema [16, 17]. These pathophysiologic changes are also recognized in nonembolic regions [18] and result in ventilation-perfusion (V/Q) ratio inequality [3, 19-22]. Thus, theoretically, lung attenuation in patients with acute PE may become heterogeneous. However, several investigators have reported that heterogeneous lung attenuation, termed "mosaic perfusion," is a less frequent and uncommon feature of acute PE [23, 24]. Unfortunately, in those previous studies, mosaic perfusion was assessed with visual and subjective evaluations. To our knowledge, quantitative analysis of lung attenuation in patients with acute PE has not been assessed.
Frequency histograms of CT scans of the lung provide quantitative indexes—such as mean lung attenuation, skewness, and kurtosis—for the evaluation of the distribution of lung attenuation values [25, 26]. Skewness is a measure of the degree of asymmetry of a distribution. Kurtosis is a measure of whether the data are peaked or flat relative to a normal distribution. This quantitative analysis has been used for the evaluation of interstitial pulmonary fibrosis, and the quantitative indexes have been shown to correlate with pulmonary function [26]. Considering the pathophysiologic characteristics of acute PE, we hypothesized that the frequency histograms of lung attenuation in patients with PE differ from those in patients without PE (referred to hereafter as "non-PE patients"). In addition, many patients with acute PE develop clinically significant hypoxemia. Although the exact mechanism by which this hypoxemia occurs is controversial, the potential mechanism of hypoxemia in acute PE can primarily be explained by V/Q ratio inequality [3, 19-22], which causes heterogeneous regional alterations of ventilation and perfusion. Therefore, we also hypothesized that there may be a relation between distribution of lung attenuation values and the degree of hypoxemia in patients with acute PE.
The purpose of this study was to investigate the distribution of lung attenuation values on CTA with quantitative histogram analysis in patients with acute PE in comparison with non-PE patients. In addition, we assessed the relation between the degree of hypoxemia and the distribution of lung attenuation values compared with the degree of arterial obstruction scored using CTA.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In the 130 patients who had no sign of PE on CTA, 58 patients underwent CT venography of the lower extremities at the same time. Of those patients, we selected 15 patients who also had a normal D-dimer level. We excluded three patients who had prior cardiopulmonary disease (n = 1) or obvious abnormal parenchymal lesions (n = 2). One patient had a history of smoking, so that patient was also excluded. Thus, 11 patients (five men and six women; mean age, 56 years; age range, 39-87 years) were included in the non-PE group.
CT Scanning
CTA was performed on one of two MDCT scanners: Aquilion (Toshiba Medical
Systems), with a collimation of 4 x 1 mm, a table speed of 8 mm per
rotation, and a 0.5-sec rotation time (PE group, n = 16; non-PE
group, n = 7), or Asteion (Toshiba Medical Systems), with a
collimation of 4 x 2 mm, a table speed of 16 mm per rotation, and a
0.75-sec rotation time (PE group, n = 10; non-PE group, n =
5), 120 kV, and 200-250 mA using a 512 x 512 matrix. CTA was performed
from 1 cm above the top of the aortic arch to the diaphragm at deep
inspiration during a single breath-hold. A bolus of 100 mL of iodinated
contrast material was injected at a rate of 3 mL/sec with an automatic
injector. The bolus-tracking method was used for optimizing pulmonary artery
opacification. A region of interest was placed on the main pulmonary artery,
and scanning was initiated at a preset level of opacification. With this
protocol, we consistently achieved high and uniform contrast enhancement
throughout the thorax in all patients. For each patient, the raw data were
retrospectively reconstructed with a section thicknesses of 2 mm.
Evaluation of CT Findings
Quantitative assessment of lung attenuation—CT images were
transferred to a PC for quantitative analysis of lung attenuation. The CT
images with 2-mm section thickness and 10-mm spacing were selected from among
all of the CT images of each patient, and these CT images were segmented using
a semiautomatic image-processing program (ImageJ [version 1.31], a public
domain Java image-processing program inspired by the National Institutes of
Health program Image for the Macintosh [Apple Computer]). This program uses a
semiautomatic threshold technique to isolate the lungs from other tissues and
structures and selects all pixels between -500 and -1,024 H. Minimal user
intervention by one radiologist was required to exclude nonlung structures
that satisfy the threshold criteria, such as the trachea, blood vessels, and
large bronchi near the hilum. CT slices with a significant interlobar fissure
area or extensive artifacts caused by the inflow of high concentrations of
contrast medium in the superior vena cava were removed from the analysis.
We used a 16-bit histogram plug-in, the Java programming language, which is
available on the ImageJ Web site
[27], to obtain frequency
histograms. Gray-scale frequency histograms for the combined lung parenchyma
of each subject were plotted. Because lung size varied from one patient to
another, the number of pixels was preferentially expressed as a percentage of
the total number of pixels. Values for the mean lung attenuation, median lung
attenuation, SD, skewness, and kurtosis of the gray-scale frequency histogram
for each subject were calculated using statistical software (JMP software
[version 5.0.1], SAS Institute). Skewness describes the degree of asymmetry of
a histogram; a histogram with a long tail to the right has a positive skewness
value, and a completely symmetric distribution has a skewness value of zero.
Kurtosis describes how sharply peaked a histogram is; a histogram that is more
peaked than a normal distribution has a positive kurtosis value, and a normal
distribution has a kurtosis of zero. Skewness is given by the following
equation:
 |
 |
Visual assessment of lung attenuation—The CT images obtained with 2-mm section thickness and 10-mm spacing were selected from among all of the CT images of each patient. The same CT images were viewed as those for quantitative assessment. These CT images were evaluated by two experienced radiologists independently without knowledge of clinical data, and final decisions about the findings were reached by consensus. Regional attenuation differences between normal and abnormal lung parenchyma are often small [28]; thus, window settings for the detection of subtle lung attenuation differences (width, 250 H; center, median lung density) were applied. The patterns of lung attenuation on single-detector images were divided into two categories: 1, homogeneously normal or gravity-dependent gradient lung attenuation with no obvious difference in attenuation values or patterns between the right and left lungs; and 2, geographic regions of heterogeneous lung attenuation or attenuation values or patterns that differ between the right and left lung parenchyma. Category 1 was assigned 0 points, and category 2 was assigned 1 point. In each patient, the visual score was calculated by adding scores for each slice and then dividing by slice number.
CTA—CTA was evaluated on a monitor in a scroll-through or cine mode with the standard window settings for viewing CTA (width, 500 H; center, 80 H) at our institution. Two experienced radiologists without knowledge of the clinical data independently evaluated the images, and final decisions about the findings were reached by consensus. PE was diagnosed when either complete or partial filling defects within the main, lobar, or segmental arteries were identified. CTA scores described in the literature [1] were applied to evaluate the degree of pulmonary artery obstruction.
To define the CT obstruction index, we regarded the arterial tree of the
lung as having 18 segmental arteries: three to the right upper lobes, two to
the left upper lobes, two to the middle lobe and to the lingula, five to the
right lower lobes, and four to the left lower lobes. The presence of an
embolus in a segmental artery was scored as 1 point, and emboli in the most
proximal arterial level were scored a value equal to the number of segmental
arteries arising distally. To provide additional information about the
residual perfusion distal to the embolus, a weighting factor was assigned to
each value, depending on the degree of vascular obstruction. This factor was
equal to 0 when no thrombus was observed; 1, when partially occlusive thrombus
was observed; or 2, with total occlusion. Thus, the maximal CT obstruction
index was 36 per patient. Isolated subsegmental embolus was considered as a
partially occluded segmental artery and was assigned a value of 1. The
percentage of vascular obstruction was calculated by dividing the patient
score by the maximal total score and by multiplying the result by 100.
Therefore, the CT obstruction index can be expressed as follows:
 |
ABG Measurements
In the present study, we evaluated patients with PE who had ABG levels
measured while breathing room air. All measurements of ABG were obtained with
a blood gas analyzer (ABL520, Radiometer) during the 24 hr before or after
CTA. The following physiologic variable was measured: partial pressure of
O2 (PaO2).
Statistical Analysis
Comparisons of the mean for mean lung attenuation, median lung attenuation,
SD, skewness, kurtosis, and P ± 50 H between the PE group and the
non-PE group were performed using the Mann-Whitney U test. Comparison
of the visual score of lung attenuation between the PE group and the non-PE
group was also performed using the Mann-Whitney U test. The linear
regression analysis and Spearman's rank correlation analysis were used to
evaluate the relationship between, first, PaO2 and the
mean lung attenuation, median lung attenuation, SD, skewness, kurtosis, or P
± 50 H; second, PaO2 and the visual lung
attenuation score; third, PaO2 and the CTA score; and,
fourth, the CTA score and the mean lung attenuation, median lung attenuation,
SD, skewness, kurtosis, or P ± 50 H. All statistical analyses were
performed using JMP 5.0.1 software. Data are expressed as mean ± SD.
For all statistical analyses, a p value of less than 0.05 was
considered significant.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The results of quantitative analysis of lung attenuation are shown in Table 1. The mean values of the mean lung attenuation and median lung attenuation for the PE group and non-PE group were not significantly different (p = 0.273 and p = 0.158, respectively). Other measured parameters—SD, skewness, kurtosis, and P ± 50 H—of the frequency histogram were significantly different between patients with PE and those without PE. The mean value of the SD of the PE group (87.1 ± 8.4) was significantly higher than that of non-PE group (78.9 ± 2.9) (p = 0.0003). The mean value of skewness of the PE group (1.36 ± 0.5) was significantly lower than that of the non-PE group (1.82 ± 0.2) (p = 0.0071). The mean value of kurtosis of the PE group (2.04 ± 1.7) was significantly lower than that of the non-PE group (3.78 ± 1.1) (p = 0.0047). The mean value of P ± 50 H in the PE group (42.2% ± 10.2%) was significantly greater than that of the non-PE group (31.7% ± 4.0%) (p = 0.0028). Typical frequency histograms are shown in a patient with PE (Figs. 1A and 1B) and one without PE (Figs. 2A and 2B). All PE and non-PE patients had a unimodal distribution histogram. Non-PE patients had the histogram with a skewed long tail to the right, toward increasing gray-scale density. The patients with PE, however, had a distribution that was more symmetric and flatter (lower kurtosis) than that of the non-PE patients.
|
|
|
|
|
The Relations Between Measured Histogram Parameters and ABG Measurements
The correlation between the measured parameters of the frequency histogram
and measurements of PaO2 is shown in
Table 2. The mean
PaO2 was 67.7 ± 10.8 mm Hg. No significant
correlation was found between the mean lung attenuation and
PaO2 (r = 0.280, p = 0.39). Similarly,
there was no significant correlation between the median lung attenuation and
PaO2 (r = 0.309, p = 0.23). In
contrast, significant correlations were found between the other measured
parameters (SD, skewness, kurtosis, and P ± 50 H) of the frequency
histogram and ABG levels (Figs.
3A,
3B,
3C, and
3D). A significant inverse
relationship between the SD and PaO2 (r =
-0.770, p = 0.001) was found. Significant positive relationships were
found between PaO2 and both skewness (r =
0.797, p < 0.0001) and kurtosis (r = 0.786, p
< 0.0001). The values of the P ± 50 H were highly correlated with
PaO2. The values of P ± 50 H were increased with
decreasing PaO2 levels (r = -0.871, p
< 0.0001).
|
|
|
|
|
Visual Scores of Lung Parenchyma
The same CT slices that were evaluated for the analysis of distribution
histogram of lung attenuation were used for the visual score assessment. The
mean visual scores for lung attenuation were 0.60 ± 0.16 (range,
0.28-0.88) in the PE group and 0.22 ± 0.16 (range, 0-0.60) in the
non-PE group. There was a significant difference between the PE and non-PE
groups (p < 0.0001). Significant correlations were found between
visual score and PaO2 (r = -0.709, p =
0.004) (Fig. 4).
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Recently, visualization and quantification of pulmonary perfusion based on various CT techniques have been introduced [11-14]. Contrast-enhanced dynamic helical CT densitometry allows the assessment of at least segmental perfusion defects in patients with acute PE [11-14]. The segments with reduced perfusion showed a significantly lower mean CT attenuation value on the enhanced scans than the segments with normal perfusion. Im et al. [15] reported that they did not find lung attenuation changes in pigs during the 28 days immediately after experimentally induced PE. However, they evaluated lung attenuation with unenhanced CT. IV contrast administration produced a rise in the overall mean lung attenuation [29]. Thus, attenuation differences between PE and normal areas could not be detected with unenhanced CT. Although the change of the attenuation value by the perfusion defect is subtle, it may be related to some extent to the heterogeneity of the lung attenuation in patients with PE.
In patients with acute PE, generalized or local bronchoconstriction has been invoked [5-7]. The proposed mechanism includes bronchoactive amines, such as serotonin and prostaglandins, that are released during platelet aggregation in the thrombus, as has been shown in canine lungs [30], or a change in parasympathetic nervous system tension, which controls bronchial smooth-muscle tension [18]. In an experimental study, bronchial obstruction resulted in lobular areas of low attenuation on high-resolution CT [15]. Arakawa et al. [31] studied 15 patients with PE and confirmed that clots were identified on CTA in 90.6% of the arteries proximal to the areas of decreased attenuation. They also reported that 71.9% of the areas with decreased attenuation on inspiratory scans showed air-trapping on expiratory scans. Therefore, decreased blood flow and bronchoconstriction may lead to the existence of low-attenuation areas in patients with acute PE.
Lung parenchyma distal to a pulmonary thromboembolus is pathologically normal or shows only mild atelectasis and minimal intraalveolar hemorrhage or edema. Especially in the early stage, histologic examination shows only intraalveolar hemorrhage and edema, with intact alveolar wall. Grossman et al. [32] found increased lung attenuation after arterial occlusion in dogs. In addition, pathophysiologically the blood flow increase may be generated in non-PE regions [4], and increased intravascular volume increases lung attenuation [16]. These pathologic and physiologic changes might lead, in part, to increases in lung attenuation.
In this study, the heterogeneity of lung attenuation in patients with acute PE was confirmed even by visual analysis. However, several investigators have reported that mosaic perfusion or the presence of low-attenuation area is not a common feature of acute PE [23, 24, 33]. Coche et al. [23] compared lung parenchymal findings of 88 patients with suspected acute PE who underwent CTA. Those researchers found mosaic perfusion in 12% of patients with acute PE and in 10% of patients without this finding. Shah et al. [24] evaluated parenchymal findings in 28 patients with acute PE and found only 7% with mosaic perfusion. However, Arakawa et al. [31] found mosaic perfusion in four (44.4%) of nine patients with acute PE.
The difference of frequency in these studies may result from variability in patient selection; these studies may differ from our study because we used a special setting for window level and window width. Groell et al. [11] reported that the segment with reduced perfusion showed a lower attenuation value on enhanced CT than segments with normal perfusion. However, their attempts to detect parenchymal perfusion defects by visual assessment have been unsuccessful because of subtle differences between defects and normal perfusion. We analyzed lung parenchyma with enhanced contrast imaging using a narrow window setting. Thus, a subtle attenuation difference could be detected in our study. The existence of mosaic perfusion in patients with acute PE may be missed with the usual window setting for lung parenchyma. However, there are various problems in the visual assessment of lung attenuation. We used the median value of lung attenuation as the center level, but we cannot assert that this method was the most appropriate one. Future investigation, including the reproducibility of our method, is required.
In our PE group, significant correlations were found between PaO2 and SD, skewness, kurtosis, or P ± 50 H. Therefore, the degree of heterogeneity in lung attenuation may relate to hypoxia. Hypoxemia is commonly reported in patients with acute PE. The exact mechanisms by which hypoxemia occurs remain in dispute. In general, hypoxemia is mainly explained by V/Q ratio inequality [3, 19-22]. The high proportion of the V/Q ratio is explained by the preferential reduction in perfusion within the region of the emboli. The low V/Q ratio is mainly due to a shifting of blood flow away from the region of the emboli toward regions with normal ventilated units [3, 19-21]. The heterogeneity of lung attenuation on CTA might reflect these pathophysiologic changes in acute PE.
Quantitative CT parameters, such as mean lung attenuation, skewness, and kurtosis, can be obtained from frequency histograms of the lung [25, 26]. This method has been applied to the quantitative evaluation for interstitial pulmonary fibrosis, and those parameters correlate with results of pulmonary function tests [26]. In our PE patients, these parameters of frequency histograms of lung attenuation differed from those of non-PE patients. The distribution of the lung attenuation values in the PE patients was characterized by a simultaneous higher proportion of lung attenuation at both extremities of the histogram, with subsequent widening of the SD and flattening of the middle peak. Quantitative analysis of lung scintigraphy scans based on the frequency histograms of the V/Q ratio was applied in patients with PE, and the histograms of the V/Q distribution showed widening of the second moment of the distribution and flattening of the middle peak, comparable to the results of our histogram analysis [34].
With regard to the distribution of the V/Q ratio using multiple perfusions of inert gases in patients with acute PE, data have also shown an abnormal distribution of lung V/Q ratios to alveolar units, with low V/Q ratios, and have helped explain the abnormalities of blood gas exchange [3, 19]. Undoubtedly, lung attenuation values and the V/Q ratio are based on the different pathophysiologic mechanisms. Low-attenuation areas on enhanced CT in patients with acute PE may relate mainly to decreased blood flow and bronchoconstriction. However, histogram parameters in our PE patients correlated with the ABG levels and might relate to abnormalities in blood gas exchange.
In previous studies, CTA scores were used for the quantitative evaluation of acute PE severity, and those scores correlated with clinical data such as oxygen saturation and echocardiographic findings [1, 2]. With regard to the importance of the size of embolic clots, experimental studies have shown that embolus size significantly influences gas exchange characteristics [35]. We adopted the method of Qanadli et al. [1] considering the size of the embolic clots. However, in our PE patients, there was relatively a low correlation between the CTA score and ABG levels. The method of quantitative assessment of lung attenuation can be used to evaluate pulmonary blood flow without influencing the existence or size of the thrombus. Furthermore, there were no significant correlations between the CTA score and the SD, skewness, kurtosis, or P ± 50 H. These parameters may reflect the reserve capacity of pulmonary function in patients with PE. By combining CTA and quantitative analysis of lung attenuation, a comprehensive and noninvasive diagnosis of thoracic structure and function is feasible with a single technique in patients with acute PE.
There are various problems with quantitative analysis of lung attenuation in patients with acute PE. First, we could not clarify the relation of lung attenuation and ventilation or blood flow in this study. Low-attenuation areas might reflect hypoperfusion, bronchoconstriction, or both. In our PE patients, few patients underwent V/Q scintigraphy because the diagnosis of PE was based on CTA examinations in most cases. To clarify the relation of bronchoconstriction, the combined use of expiration CT will be considered in the future. Second, there was considerable variability in the lung density map even among healthy individuals [29]. In addition, an attenuation gradient lung is normally present, with the most dependent lung regions being densest. However, indexes of the quantitative attenuation analysis were significantly different between the PE and non-PE groups. Third, quantitative measurements of lung attenuation are known to be affected by several factors, such as variation in CT techniques [36]. These variations in our study might have influenced the results to some extent. Although several methods were applied to correct lung attenuation for the quantitative histogram analysis of lung disease [25, 26], to our knowledge, standardization of CT densitometry has not been achieved completely. Thus, we did not correct the gray-scale values measured in current study. Further work needs to be done to develop an appropriate correction factor for lung attenuation.
Several other limitations of this study exist. First, we excluded patients with severe PE such as those requiring mechanical ventilation or having pulmonary infarction. In the complete obstruction of a unilateral pulmonary artery, lung attenuation may lower uniformly. Second, we could not conclude whether the threshold value that evaluates the low-attenuation area was appropriate. Third, although all subjects were nonsmokers, we could not exclude the possibility of lung attenuation being affected in subjects with subclinical small airways disease.
In conclusion, the parameters of frequency histograms of lung attenuation on contrast-enhanced CT were significantly different between patients with and without PE. In patients with acute PE, heterogeneity in lung attenuation was more prominent, which may be due to alteration of perfusion and ventilation distribution. In addition, significant correlations were found between those parameters and hypoxia. The degree of heterogeneity in lung attenuation may relate to abnormal gas change. By combining CTA and quantitative analyses of lung attenuation, diagnosis and severity assessment of acute PE are feasible with a single technique. Moreover, quantitative CT analyses of lung attenuation may be helpful for pathophysiologic analysis of patients with acute PE.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
