Content uploaded by Luna Gargani
Author content
All content in this area was uploaded by Luna Gargani on Nov 28, 2017
Content may be subject to copyright.
Giacomo Baldi
Luna Gargani
Antonio Abramo
Luigia D’Errico
Davide Caramella
Eugenio Picano
Francesco Giunta
Francesco Forfori
Lung water assessment by lung
ultrasonography in intensive care: a pilot study
Received: 22 January 2012
Accepted: 10 August 2012
Copyright jointly held by Springer and
ESICM 2012
G. Baldi ())A. Abramo F. Giunta
F. Forfori
Department of Surgery, University of Pisa,
Via Paradisa, 2, 56100 Pisa, Italy
e-mail: giacomo.baldi@gmail.com
Tel.: ?39-50-997605
Fax: ?39-50-997609
L. Gargani E. Picano
Institute of Clinical Physiology,
National Research Council, Pisa, Italy
L. D’Errico D. Caramella
Department of Oncology, Transplants
and New Medical Technologies,
University of Pisa, Pisa, Italy
Abstract Objective: To investi-
gate the accuracy of lung
ultrasonography (LUS) in the quanti-
fication of lung water in critically ill
patients by using quantitative com-
puted tomography (CT) as the gold
standard for the determination of lung
weight. Methods: Twenty consecu-
tive patients admitted to an intensive
care unit who underwent chest CT as
a step in their clinical management
were evaluated within 4 h by LUS.
Lung weight, lung volume, and
physical lung density were calculated
from the CT scans using ad hoc
software. Semiquantitative ultrasound
assessment of lung water was per-
formed by determining the ultrasound
B-line score, defined as the total
number of B-lines detectable in an
anterolateral LUS examination.
Results: Good correlations were
found between the B-line score and
lung weight (r=0.75, p\0.05) and
density (r=0.82, p\0.01), that
only marginally increased when the
lung density of the first 10 mm of
subpleural lung tissue was evaluated
(r=0.83, p\0.01). Moreover, val-
ues of subpleural lung density were
not significantly different from values
of the whole lung density
(0.34 ±0.11 vs. 0.37 ±0.16 g/ml,
p=ns). Very good correlations were
found between the B-line score and
both the weight (r=0.85, p\0.01)
and the density (r=0.88, p\0.01)
of the upper lobes. The weight of the
lower lobes was not correlated with
the B-line score (r=0.14, p=ns).
Conclusions: Lung ultrasound
B-lines are correlated with lung
weight and density determined by
CT. LUS may provide a reliable,
simple and radiation-free lung densi-
tometry in the intensive care setting.
Keywords Lung ultrasonography
Extravascular lung water
Lung densitometry
Lung ultrasound B-lines
Intensive care
Introduction
A clinically significant increase in lung water is a frequent
condition in intensive care patients and is usually sec-
ondary to acute heart failure, acute lung injury or acute
respiratory distress syndrome (ARDS) [1]. A quantitative
measurement of extravascular lung water (EVLW) would
be extremely useful for the clinical management of these
conditions, both as an index of severity and to guide
treatment [1–4]. Historically, the most frequently used
examination for this purpose in the clinical arena has been
chest radiography, which may also allow semiquantitative
assessment of pulmonary edema using a specific scoring
systems [5]. Unfortunately, portable chest radiography in
the critical care setting often yields inaccurate images [6,
7]. Thermodilution or dye-dilution techniques, although
invasive, have been shown not to be accurate in esti-
mating values of EVLW [8–10]. The gold standard for
Intensive Care Med
DOI 10.1007/s00134-012-2694-x ORIGINAL
noninvasive EVLW assessment is represented by quanti-
tative computed tomography (CT), which unfortunately is
both expensive and often not feasible, since it requires the
transfer of a critical patient to the radiology department;
moreover, it employs ionizing radiation [11].
In recent years, lung ultrasonography (LUS) has been
proposed as a versatile tool for the assessment of some
pulmonary and pleural diseases [12–16]. The lung, for-
merly believed to be poorly accessible to ultrasound, has
instead revealed a rich sonographic semeiotics, enabling
differentiation of several conditions such as pneumotho-
rax [17,18], hemothorax [19], pleural effusions [20,21],
lung consolidations [22,23] and pulmonary interstitial
syndrome [12–16,24–30]. Pulmonary edema investigated
by LUS is characterized by the appearance of sonographic
signs known as B-lines (previously called ultrasound lung
comets) [13]. B-lines are defined as discrete laser-like
vertical hyperechoic reverberation artifacts that arise from
the pleural line, extend to the bottom of the image without
fading, and move synchronously with pleural movement.
These artifacts are considered the sonographic sign of the
pulmonary interstitial syndrome [15]. They seem to
originate from the altered interface between alveolar air
and thickened interlobular septa, although the actual
physical genesis has not yet been confirmed [25].
B-lines were initially investigated as a sonographic
marker for the differential diagnosis between cardiogenic
and noncardiogenic dyspnea, showing very good accuracy
[24]. Subsequently, B-lines have been shown to correlate
not only with chest radiographic scoring for edema [27],
but also with EVLW and with pulmonary capillary wedge
pressure [28]. In ARDS, LUS signs are perceived very
early [29], with a sensitivity of 98 % and a specificity of
88 % in diagnosing the presence of the interstitial syn-
drome as seen on CT, performing better than both
auscultation and chest radiography [7].
The aim of our study was to further investigate the
correlation between the number of B-lines and the excess
of EVLW, by comparing LUS with CT. We also aimed to
establish the performance of LUS in the assessment of
EVLW in the whole lung, since LUS is especially able to
evaluate the more peripheral lung regions.
Materials and methods
Patient population
We evaluated 20 consecutive in-patients admitted to the
intensive care unit (ICU) of the Hospital Santa Chiara
(University of Pisa), who were scheduled to undergo a chest
CT scan as a step in their clinical management. Exclusion
criteria were age less than 18 years, LUS not feasible (e.g.,
chest trauma, skin burns, etc.) and lack of written informed
consent. The study was approved by the Ethics Committee
of Pisa. Written informed consent was obtained from the
patients according to the Declaration of Helsinki or from
the patient’s relatives if the patient was mechanically
ventilated. Every patient underwent a chest CT scan fol-
lowed by a LUS examination. Two patients were examined
twice and one patient three times, as they needed additional
CT scans during their stay in the ICU. Precautions were
taken to perform CT and LUS examination under similar
conditions. First, to reduce the effects of ongoing diuretic
therapy or evolution of the underlying condition on lung
water, the LUS and CT scan were temporally separated by
no more than 4 h (mean delay 2.7 ±1.5 h), and second
when the chest CT was performed in a mechanically ven-
tilated patient, the same ventilation parameters were used
during the LUS examination.
CT image segmentation
Each patient underwent a spiral CT scan of the chest
without administration of contrast agent. The CT scans
were performed on a 16-detector row Toshiba Aquilion.
CT resolution and acquisition time could not be stan-
dardized; however, the lung variables evaluated (weight
and density) are not greatly influenced by these CT
parameters.
The determination of lung weight, lung volume and
lung density required segmentation of the lungs, which is
a very time-consuming task when performed manually.
Therefore sought a faster method by developing ad hoc
software (Fig. 1). We based our work on existing tools,
namely the ITK toolkit [31] an open-source software
program for segmenting and filtering medical images. We
tailored the ITK region growing algorithms for the lung
segmentation task. To the best of our knowledge, this is
the first study using automatic CT segmentation to
extrapolate lung weight [32–35].
Lung weight and lung volume determination
We calculated lung weight and lung volume by summing
the weight and volume of every voxel in the segmented
areas as previously described [35]. We also classified lung
regions according to their aeration: hyperinflated, nor-
mally aerated, poorly aerated and non-aerated zones [25].
Lung ultrasonography
Each patient underwent LUS to assess for the presence of
B-lines, as previously described [13]. The linear 10-MHz
probe of commercially available machines (Philips Sonos
7500, Siemens Acuson Sequoia, Esaote MyLab50) was
employed. A series of scans was performed with the patient
in the supine position, applying the probe perpendicular to
the skin over intercostal spaces along anatomical reference
lines. More precisely, we scanned the anterolateral hemi-
thoraxes along the parasternal, midclavicular, anterior
axillary and mid-axillary lines. The left hemithorax was
scanned from the second to the fourth intercostal space,
whereas the right lung was scanned from the second to the
fifth intercostal space, giving a total of 28 scanning sites
(Fig. 2). At each scanning site the number of B-lines was
determined and summed to yield the B-line score.
The edematous lung shows B-lines as vertical narrow
bands originating from the pleural line and extending to the
bottom of the image (Fig. 3a). Evaluation of the number of
B-lines has been previously described [16]. A B-line score
of B5 was considered a normal sonographic pattern, since a
few B-lines can be present in healthy subjects, especially
above the diaphragm [12]. We considered a full white
screen of confluent B-lines (corresponding to ten B-lines) as
a white lung pattern [16]. The white lung pattern was not
frequent in our study population and was found mainly in
patient 16, and corresponded to the presence of ground-
glass opacity on the CT scan.
Regional density
Aeration and lung weight vary between different regions
of the lung, especially in ARDS [36]. Since we assessed
the B-line scores exclusively in the anterior and lateral
lung regions, we further analyzed CT scans to extract
weight, volume and density of the dependent (lower
lobes) and nondependent (upper lobes) regions. We also
estimated by mathematical simulation of the distance
traveled by the ultrasound beam as it is dissipated by the
subpleural lung tissue density (SLD).
Statistical analysis
Continuous variables are expressed as means ±SD or as
medians (25th, 75th percentiles) as appropriate. Differ-
ences in LUS and CT parameters were tested by the
Mann-Whitney nonparametric test. We used nonpara-
metric Spearman’s correlation coefficient and linear
regression analysis to search for correlations between
B-line score and the several variables derived by our ad
hoc software. In particular, we tested for correlations
between B-line score and lung weight, lung density, lung
gas volume and volumes of the differently aerated areas.
We also repeated the regression analysis for both left and
right lungs, for both upper and lower lobes and for every
subpleural layer constructed from the CT scans. A
pvalue of 0.05 was considered statistically significant.
We used R software, version 2.11.1, for the statistical
analysis.
Fig. 1 Automatic lung segmentation (green right lung, red left
lung, blue trachea and main bronchi). aThe output of the
segmentation algorithm in a patient with ARDS. Lower lobe
segmentation had to be performed manually due to intense edema
and atelectasis. bConstruction of subpleural layers in a control
patient: the 10-mm thick layer is shown for both lungs. Note how it
starts from the parasternal regions and contours the lungs up to the
mid-axillary line. c3D visualization of the subpleural layers in
brepresenting the portions of lung explored by LUS
Fig. 2 Data collection table
used for the LUS examination
with the scanning site positions
over the patient’s thorax
Results
Study population
The demographic and clinical characteristics of the
patients are listed in Table 1. In the 13 patients with
pathological pulmonary conditions seen on the chest CT
scans, the causes of respiratory failure were: pneumonia
(in 5), ARDS (in 5), atelectasis (in 2), and pleural effusion
(in 1). In seven patients the chest CT scan was normal.
One patient had emphysema.
CT measurements
The lowest CT resolution was associated with a voxel
volume of 1.797 mm
3
and the highest with a voxel volume
of 0.618 mm
3
with a mean volume of 1.21 ±0.31 mm
3
.
Some of the CT scans showed artifacts due to the urgency
of the examination, mechanical ventilation or the presence
of chest drain tubes. However, every patient was correctly
positioned with arms above the head and such artifacts
only prolonged the time required for lung segmentation,
without affecting the accuracy of the lung weight and
volume calculations.
Lung ultrasonography
LUS examinations were always performed in less than
10 min, with a feasibility of 100 %. The median B-line
score was 20 (interquartile range 5–27). We evaluated the
performance of LUS in differentiating between patholog-
ical and normal lungs. Considering a B-line score of B5as
the cut-off between pathological and normal lungs, we
estimated the sensitivity and specificity of LUS for
identifying edematous lungs in ICU patients. The sensi-
tivity was 94 %, corresponding to 16 correctly classified
examinations from a total of 17 edematous lungs accord-
ing to the CT scans. Patient 13 was wrongly classified as
nonpathological, because the posterior consolidation and
effusion were not detected on the anterolateral examina-
tion. Such a consolidation can be quite significant as has
been shown previously by the increased weight of the
lower lobes with respect to healthy controls [36]. The
specificity of LUS was 100 % and every nonpathological
patient was correctly classified.
Lung weight, lung density and B-line score
A significant linear correlation was found between B-line
score and lung weight (r=0.75, p\0.05; Fig. 4a). The
regression plots showed differences between patients with
medium lung weight/low B-line score (e.g., patient 3) and
patients with medium lung weight/high B-line score (e.g.,
patient 2). To account for these differences we took into
consideration the respective patient’s lung volume.
Patient 3 had an emphysematous lung with no edema and
a high volume (4,765 ml), whereas patient 2 had ARDS
with massive pleural effusions and a consequently
reduced volume (1,912 ml). Therefore we searched for a
correlation between B-line score and lung density
(defined as the ratio between lung weight and lung vol-
ume), which was also found to be significant (r=0.82,
p\0.01; Fig. 4b) with an increased rvalue. Considering
single lungs, a better correlation was found between right
lung density and right B-line score (r=0.83, p\0.01)
than between left lung density and left B-line score
(r=0.79, p\0.01). This can be explained by the fact
that the right lung is better explored by LUS, whereas on
Fig. 3 LUS scan (left)
performed at the third
intercostal space along the right
anterior axillary line in a patient
with ARDS. The pleural line
(P) is markedly hyperechogenic
and gives rise to five B-lines
(C); Sskin, Mmuscles. Note the
irregularity of the pleural line
from which the central B-line
originates; this is a frequent
sign in ARDS. On the
corresponding CT image
(right), the portion of lung
corresponding to the LUS
image on the left is shown
(dashed circle). Note how it
corresponds to an edematous
region that extends to the pleura
the left side, the presence of the heart may limit the
evaluation of the lung. We also searched for a correlation
between B-line score and the percentages of differently
aerated tissue. The best correlation coefficient was found
for poorly aerated zones (r=0.84, p\0.01).
Subpleural density and B-line score
For each CT scan, we estimated the maximum subpleural
distance reached by the ultrasound beam (d
M
), assuming
an ultrasound speed of 1,540 m/s in tissues and 350 m/s
in air. The median value for d
M
is 6 mm with a minimum
of 2 mm in normal lungs and a maximum of 10 mm in
ARDS. Then, for each patient we considered only the
portion of lung parenchyma starting from the pleura up to
such distance and calculated its density. SLD represents
the actual density the ultrasound beam reacts to during
LUS. The correlation between B-line score and SLD was
slightly better (r=0.83, p\0.01) than the correlation
between B-line score and total lung density (Fig. 6a).
Moreover, the mean difference between SLD and total
lung density was less than 0.03 g/ml.
Weight and density of the upper and lower lobes
A significant correlation was found between B-line score
and the weight of the upper lobes (r=0.85, p\0.01;
Fig. 5a). On the other hand, no correlation was found
between B-line score and the weight of the lower lobes
(r=0.14, p=ns; Fig. 5b). The correlation between
B-line score and the density of the upper lobes was
slightly stronger (r=0.88, p\0.01; Fig. 6b) than the
correlation between B-line score and SLD.
The results of the CT scans and LUS examinations are
summarized in Table 2.
Discussion
This study confirmed that LUS is a reliable tool for
evaluating lung water, showing a very good correlation
with lung density as assessed by CT. To the best of our
knowledge, this is the first study to evaluate the correla-
tion between B-lines and lung density derived from chest
CT. Our findings are consistent with previously reported
findings showing that B-lines are the sonographic sign of
pulmonary interstitial syndrome [15], and that the origin
of B-lines is linked to the thickening of interlobular septa
due to edema or fibrosis [37]. Previous studies have
compared LUS findings to chest CT signs of pulmonary
edema, both qualitatively [12] and semiquantitatively [38,
39]. In this study, we achieved greater accuracy in the
determination of lung water by evaluating CT-derived
Table 1 Main demographic and study variables in the individual patients (ventilatory parameter values are reported for non-control patients)
Patient Age
(years)
Sex Reason for
admission
Cause of
respiratory failure
BMI
(kg/m
2
)
PaO
2
/FiO
2
(mmHg)
PaCO
2
(mmHg)
Duration of prior
mechanical
ventilation (h)
Tidal
volume
(ml)
PEEP
(mbar)
Final
outcome
1 76 M Esophageal rupture Massive atelectasis 32.5 216 45 7 588 12 Died
2 77 M Pneumonia ARDS 35.6 131 48 21 594 17 –
3 57 M Emphysema – – – – – – – –
4 67 M Gastric aspiration Pneumonia 33.6 275 42 12 600 13 –
578F– – – – – – – – –
6 62 M Neutropenic leukemia ARDS 38.2 155 47 15 654 15 Died
764M– – – – – – – – –
864F– – – – – – – – –
9 80 F Sepsis after mastectomy Atelectasis 27.3 – – – – – –
10 61 M Sepsis after biliary stenting Pneumonia 31.7 224 44 52 586 12 Died
11 58 M Pneumonia ARDS 24.8 116 49 32 500 14 –
12 73 M Esophagectomy Pneumonia 25.9 340 40 8 486 10 –
13 73 F Pancreatitis Pleural effusion 34.2 – – – – – –
14 72 M Sepsis after biliary surgery Pneumonia 21.8 280 43 45 448 12 Died
15 38 F Tracheoesophageal fistula Pneumonia 22.4 380 42 4 480 10 –
16 69 F CMV infection ARDS 41.3 110 49 36 684 17 Died
17 80 M Urosepsis ARDS 23.6 184 47 22 512 15 –
18 57 M – – – – – – – – –
19 45 F – – – – – – – – –
20 61 M – – – – – – – – –
quantitative lung volume and lung weight measurements.
We improved on the methodology of lung segmentation
employed previously [34,35] by developing semiauto-
matic segmentation software.
We found that the B-line score correlates better with
lung density than with lung weight. Lung density repre-
sents the information carried by both lung weight and
lung volume and can therefore detect clinical conditions,
where only one of those variables changes significantly.
For example, an increase in lung volume due to an
evacuative thoracentesis should in principle not modify
lung weight very much, but should instead lead to a
reduction in lung density, which is responsible for the
better respiratory performance. On the other hand, a
decrease in lung volume due to a pneumothorax should
increase lung density without modifying lung weight too
much. Lung weight alone would not be able to explain or
detect these clinical conditions.
This correlation poses the question as to how LUS can
be accurate even though it actually explores only a sample
of the lung surface. We propose a twofold interpretation.
On the one hand, as we showed in the SLD analysis, the
Fig. 4 Positive linear correlations between B-line score and lung weight (a) and lung density (b). There is a better correlation between
B-line score and lung density. Dashed lines are the regression lines
Fig. 5 Correlations between B-line score and the weights of the upper lobe (a) and lower lobe (b). There is a positive linear correlation
for the upper lobes, but the correlation for the lower lobes is not significant. Dashed line is the regression line
ultrasound beam penetrates more deeply into edematous
lungs, therefore becoming more informative. On the other
hand, the total density of the lungs does not differ too
much from the subpleural density detected by LUS
(±0.03 mg/dl). In other words, EVLW can be accurately
determined by LUS, because its increase is often caused
by a diffuse condition affecting both the central and
peripheral zones of the lung in a similar fashion [36,40].
Fig. 6 Positive linear correlations between B-line score and SLD (a) and upper lobe density (b). Dashed line is the regression line
Table 2 Analysis of the LUS examination and CT scan results
Patient Aeration
loss pattern
on CT
Lung
weight
(g)
Upper lobe
weight (g)
Lower lobe
weight (g)
Lung
volume
(ml)
Lung density
(g/ml)
Volume of
poorly aerated
regions (%)
B-line
score
d
M
(mm)
d
1
a
Focal 1,006 604 402 2,214 0.45 15 25 6
1
b
Focal 890 490 401 2,395 0.37 14 23 6
2 Diffuse 1,304 913 391 1,912 0.68 76 48 10
3 – 1,381 829 552 4,765 0.29 11 3 7
4 Patchy 1,594 845 749 4,129 0.39 24 27 7
5 – 813 341 472 4,492 0.18 6 2 4
6 Diffuse 1,924 1,231 693 3,462 0.56 33 37 8
7 – 730 285 445 5,151 0.14 6 4 3
8 – 631 259 372 5,149 0.12 4 5 4
9 Focal 758 394 364 2,324 0.33 16 14 7
10
a
Focal 1,620 1,264 356 2,533 0.64 42 56 10
10
b
Focal 1,088 566 522 3,307 0.33 16 21 4
11 Diffuse 2,163 1,406 757 4,265 0.50 40 41 10
12 Focal 859 369 490 1,945 0.44 12 6 7
13 Focal 1,282 603 679 3,576 0.36 14 4 5
14
a
Focal 1,428 900 528 2,932 0.49 29 20 10
14
b
Focal 1,250 850 400 2,725 0.46 27 14 6
14
c
Focal 822 485 337 1,867 0.44 17 11 6
15 Patchy 1,121 695 426 3,203 0.33 18 9 4
16 Diffuse 2,342 1,757 586 4,983 0.47 37 48 8
17 Diffuse 1,484 1,143 341 3,712 0.40 35 43 7
18 – 552 226 326 4,245 0.13 5 2 3
19 – 573 275 298 4,176 0.14 6 2 2
20 – 641 301 340 3,982 0.16 7 3 3
Mean – 1,177 698 479 3,476 0.37 21 18 6.1
SD – 497 414 155 1,070 0.16 17 16 2.4
a, b, c
First, second and third examinations in the same patient
d
Distance reached by the simulated ultrasound beam
Moreover, in ARDS an increased lung density is often
present in the upper lobes [36], explaining the good cor-
relation between B-line score and both the upper lobe
density and SLD.
Clinical implications
The implementation of LUS as a tool for a dynamic
evaluation of EVLW would have important clinical
implications. LUS is a bedside, easy-to-learn, rapid,
semiquantitative, non-ionizing, repeatable technique that
is appealing compared to other more accurate methods of
lung water assessment such as thermodilution or CT.
Furthermore, the possibility of obtaining real-time infor-
mation on EVLW that can be directly evaluated by the
intensivist and readily correlated with the patient’s clini-
cal context would allow early diagnosis, serial follow-up,
and eventually therapy tailoring.
The diagnostic power of LUS can also be employed to
assess pulmonary edema during its early stages, thus
helping to decrease the potential mortality of lung con-
ditions where diagnostic timing is crucial (e.g., sepsis-
induced ARDS). LUS can also be used in the differential
diagnosis of respiratory failures by rapidly ruling out an
interstitial syndrome, in which the B-line score is not in
the significant range. Lastly, LUS can be used as a tool to
guide therapy. LUS has been recently shown to be very
accurate in the assessment of lung reaeration after anti-
biotic therapy in pneumonia [39] and in PEEP titration
during lung recruitment maneuvers [38]. There are still
many clinical settings where LUS may be a promising
approach, such as monitoring of fluid resuscitation [30]
and diuretic therapy [41].
Study limitations
Our study had some limitations. First, the number of
enrolled patients was too small to give conclusive results.
Another limitation was the nonexhaustive sample of lung
conditions we analyzed. Moreover, although many studies
have demonstrated that intra- and interobserver variability
are reasonable if sonographers are properly trained [27,
42,43], the evaluation of the number of B-lines can suffer
from operator dependency. CT scans were analyzed using
ad hoc software and were not validated using the gravi-
metric gold standard, but the latter has been tested against
LUS, and a very good correlation was found [44]. The
maximum of 4 h as the time delay between the CT scan
and LUS examination is not negligible, since variation in
lung water is a dynamic process. Unfortunately, it was not
possible logistically to perform LUS immediately after
the CT scan because the CT facilities were located in a
different department. However, no patient had a signifi-
cant fluid shift, either as a result of fluid administration or
diuretic therapy. We employed a linear probe because it
was the only available probe in our ICU (generally
employed for US-guided vascular access). The linear
probe is not the preferred one to assess B-lines, which are
better visualized using a convex, microconvex or a car-
diac probe. However, linear probes have been used for the
assessment of B-lines and showed a good correlation with
the phased-array probe [45].
Conclusions
Our findings strengthen the existing evidence that LUS is
a reliable tool for the semiquantitative assessment of
EVLW in the ICU. We also showed that LUS is corre-
lated better with lung density than with lung weight.
However, further studies with larger populations are
needed to confirm these preliminary data.
Addendum: data analysis
Lung weight and volume
These variables are derived by summing voxel volumes
(V
v
) and attenuation values (A
v
) over the voxels pertain-
ing to the lungs. Therefore, the difference in the
calculated lung weight and density between a fine-grained
and a coarse-grained CT scan is due to the coarser
approximation of the contour lung regions only. The
average voxel volume was 1.21 ±0.31 mm
3
.
To calculate the weight of a voxel we applied a pre-
viously described method [33–35]. The correctness of the
calculation is based on some working assumptions: (1)
there is a linear correlation between the A
v
in Hounsfield
units and the physical density of the tissue it represents,
(2) the physical density of non-aerated lung parenchyma
(q) (including vessels and extravascular water) can be
approximated as q=1 kg/l, namely the same physical
density of water, corresponding to 0 HU, (3) the physical
density of air corresponds to attenuation values
B-1,000 HU. This means a voxel with such an attenua-
tion value contains only air.
For each lung voxel we calculated its weight in grams
(W
v
) and its gas content in milliliters (G
v
) by considering
three cases:
•A
v
C0: V
v
is entirely represented by lung parenchyma
and therefore W
v
=q9V
v
, whereas G
v
=0
•AvB-1,000: V
v
is entirely represented by gas and
therefore W
v
=0, whereas G
v
=V
v
•-1,000 \A
v
\0: V
v
is partially represented by gas
and the rest by parenchyma. According to assumption
1, we can write G
v
=A
v
9V
v
/-1,000. Consequently,
W
v
=(V
v-
G
v
)9q.
Segmentation algorithm
The first step in our algorithm was to segment the struc-
tures that are easily identifiable (e.g., trachea, main
bronchi, bones) and to remove them from the image. In a
second step, we selected a voxel as the starting point for
the segmented region. Every voxel surrounding the seg-
mented region was then automatically tested against an
attenuation value threshold (0 HU or less) and included in
the segmented region, if the test was passed. The process
was repeated until no new voxels could be added to the
segmented region. As a last step, we filtered the image in
the segmented region to include smaller areas (e.g., con-
solidations, vessels, etc.) with attenuation values greater
than 0 HU that did not pass the threshold test, but that
were clearly inside the lung.
This computer-assisted methodology allowed us to
perform lung segmentation in a very short time. Normal
lungs could be segmented in less than 5 min which is
much less time than is needed for manual segmentation.
When lungs were very edematous, with pleural effusions
and atelectasis, the time required increased and part of the
segmentation task had to be performed manually.
Ultrasound beam simulation
Ultrasound energy is rapidly dissipated as the beam
encounters alveolar air. Therefore, LUS actually provides
information on a peripheral layer of subpleural lung tis-
sue. The thickness of the LUS-explorable layer increases
when the air content of the lung diminishes because the
ultrasound beam is less reflected. We performed a math-
ematical simulation to estimate the distance traveled by
the ultrasound beam as it is dissipated by the subpleural
SLD extrapolated from CT images.
First, we segmented the CT scans considering only the
anterolateral portions of the thorax corresponding to the
lung surface evaluated by LUS. This was achieved by
manually cutting the images at the mid-axillary lines. We
then constructed a series of ten 1-mm thick concentric
subpleural layers for each lung, and calculated their vol-
ume, weight and density (Fig. 1). Subpleural layers were
automatically constructed by passing the segmented
images through a distance filter that selects only voxels
that are at most xmillimeters away from the external
contour of the lung, repeating the process for xin the
range 1–10. We decided to construct ten layers corre-
sponding to 1 cm of subpleural lung tissue because we
estimated it to be the maximum distance reachable by the
ultrasound beam. The upper and lower lobes were seg-
mented using the same algorithm employed for
construction of subpleural layers, simply increasing the
distance filter tolerance.
We then used the resulting data to simulate the energy
loss of the ultrasound beam as it travels through consec-
utive layers of tissue. The simulation is simply performed
by calculating the energy reflection coefficient
(R) between two interfaces with different acoustic
impedance. More precisely R=[(Z
2
-Z
1
)/(Z
2
?Z
1
)]
2
where Z
1
and Z
2
are the acoustic impedances of two
adjacent subpleural layers, respectively. We considered
the point where 99 % of the beam energy is dissipated as
the maximum distance (d
M
) reached by the beam. We use
d
M
to calculate the density of the subpleural layer (SLD)
with the corresponding thickness.
Our simulation of beam energy loss was based on the
assumption that air and tissue are uniformly distributed in
each layer. Since this is not the case in real lung, we
obtained an overestimation of the maximum distance
reached. However, the difference between SLD and
total lung density was generally very small after the
second subpleural layer (SLD -lung density =0.03 ±
0.05 g/ml).
References
1. Sakka SG, Klein M, Reinhart K, Meier-
Hellmann A (2002) Prognostic value of
extravascular lung water in critically ill
patients. Chest 122:2080–2086
2. Martin GS, Eaton S, Mealer M, Moss M
(2005) Extravascular lung water in
patients with severe sepsis: a
prospective cohort study. Crit Care
9:R74–R82
3. Isakow W, Schuster DP (2006)
Extravascular lung water measurements
and hemodynamic monitoring in the
critically ill: bedside alternatives to the
pulmonary artery catheter. Am J
Physiol Lung Cell Mol Physiol
291:L1118–L1131
4. Tatsushi M, Kazumata K, Ajiki M,
Ushikoshi S, Terasaka S (2007) Goal-
directed fluid management by bedside
transpulmonary hemodynamic
monitoring after subarachnoid
hemorrhage. Stroke 38:3218–3224
5. Pistolesi M, Giuntini C (1978)
Assessment of extravascular lung water.
Radiol Clin North Am 15:551–574
6. Rouby JJ, Puybasset L, Cluzel P,
Richecoeur J, Lu Q, Grenier P (2000)
Regional distribution of gas and tissue
in acute respiratory distress syndrome.
II. Physiological correlations and
definition of an ARDS severity score:
CT scan ARDS study group. Intensive
Care Med 26:1046–1056
7. Lichtenstein D, Goldstein I, Mourgeon
E, Cluzel P, Grenier P, Rouby JJ (2004)
Comparative diagnostic performances
of auscultation, chest radiography, and
lung ultrasonography in acute
respiratory distress syndrome.
Anesthesiology 100:9–15
8. Roch A, Michelet P, Lambert D,
Delliaux S, Saby C, Perrin G, Ghez O,
Bregeon F, Thomas P, Carpentier JP,
Papazian L, Auffray JP (2004)
Accuracy of the double indicator
method for measurement of
extravascular lung water depends on the
type of acute lung injury. Crit Care Med
32:811–817
9. Sakka SG, Reinhart K, Meier-Hellmann
A (1999) Comparison of pulmonary
artery and arterial thermodilution
cardiac output in critically ill patients.
Intensive Care Med 25:843–846
10. Rossi P, Wanecek M, Rudehill A,
Konrad D, Weitzberg E, Oldner A
(2006) Comparison of a single indicator
and gravimetric technique for
estimation of extravascular lung water
in endotoxemic pigs. Crit Care Med
34:1437
11. Picano E (2004) Sustainability of
medical imaging. BMJ 328:578–580
12. Lichtenstein D, Meziere G, Biderman
P, Gepner A, Barre O (1997) The
comet-tail artifact. An ultrasound sign
of alveolar-interstitial syndrome. Am J
Respir Crit Care Med 156:1640–1646
13. Picano E, Frassi F, Agricola E,
Gligorova S, Gargani L, Mottola G
(2006) Ultrasound lung comets: a
clinically useful sign of extravascular
lung water. J Am Soc Echocardiogr
19:356–363
14. Copetti R, Soldati G, Copetti P (2008)
Chest sonography: a useful tool to
differentiate acute cardiogenic
pulmonary edema from acute
respiratory distress syndrome.
Cardiovasc Ultrasound 6:16
15. Volpicelli G, Elbarbary M, Blaivas M,
Lichtenstein DA, Mathis G, Kirkpatrick
AW, Melniker L, Gargani L, Noble VE,
Via G, Dean A, Tsung JW, Soldati G,
Copetti R, Bouhemad B, Reissig A,
Agricola E, Rouby JJ, Arbelot C,
Liteplo A, Sargsyan A, Silva F,
Hoppmann R, Breitkreutz R, Seibel A,
Neri L, Storti E, Petrovic T;
International Liaison Committee on
Lung Ultrasound (ILC-LUS) for
International Consensus Conference on
Lung Ultrasound (ICC-LUS) (2012)
International evidence-based
recommendations for point-of-care lung
ultrasound. Intensive Care Med
38:577–591
16. Gargani L (2011) Lung ultrasound: a
new tool for the cardiologist.
Cardiovasc Ultrasound 9:6
17. Lichtenstein D, Meziere G, Biderman
P, Gepner A (2000) The ‘‘lung point’’:
an ultrasound sign specific to
pneumothorax. Intensive Care Med
26:1434–1440
18. Rowan KR, Kirkpatrick AW, Liu D,
Forkheim KE, Mayo JR, Nicolaou S
(2002) Traumatic pneumothorax
detection with thoracic US: correlation
with chest radiography and CT – initial
experience. Radiology 225:210–214
19. Lichtenstein D, Pinsky MR, Jardin F
(2007) General ultrasound in the
critically ill. Springer, Berlin
20. Gryminski J, Krakowka P, Lypacewicz
G (1976) The diagnosis of pleural
effusion by ultrasonic and radiologic
techniques. Chest 70:33
21. Yang P, Luh K, Chang D, Wu H, Yu C,
Kuo S (1992) Value of sonography in
determining the nature of pleural
effusion: analysis of 320 cases. AJR
Am J Roentgenol 159:29
22. Mathis G, Blank W, Reissig A,
Lechleitner P, Reuss J, Schuler A,
Beckh S (2005) Thoracic ultrasound for
diagnosing pulmonary embolism: a
prospective multicenter study of 352
patients. Chest 128:1531–1538
23. Reissig AK (2007) Sonographic
diagnosis and follow-up of pneumonia:
a prospective study. Respiration
74:537–547
24. Gargani L, Frassi F, Soldati G, Tesorio
P, Gheorghiade M, Picano E (2008)
Ultrasound lung comets for the
differential diagnosis of acute
cardiogenic dyspnoea: a comparison
with natriuretic peptides. Eur J Heart
Fail 10:70–77
25. Soldati G, Copetti R, Sher S (2009)
Sonographic interstitial syndrome: the
sound of lung water. J Ultrasound Med
28:163–174
26. Volpicelli G, Caramello V, Cardinale L,
Mussa A, Bar F, Frascisco MF (2008)
Bedside ultrasound of the lung for the
monitoring of acute decompensated
heart failure. Am J Emerg Med
26:585–591
27. Jambrik Z, Monti S, Coppola V,
Agricola E, Mottola G, Miniati M,
Picano E (2004) Usefulness of
ultrasound lung comets as a
nonradiologic sign of extravascular
lung water. Am J Cardiol 15:1265–1270
28. Agricola E, Bove T, Oppizzi M, Marino
G, Zangrillo A, Margonato A, Picano E
(2005) ‘‘Ultrasound comet-tail
images’’: a marker of pulmonary
edema: a comparative study with wedge
pressure and extravascular lung water.
Chest 127:1690–1695
29. Gargani L, Lionetti V, Di Cristofano C,
Bevilacqua G, Recchia FA, Picano E
(2007) Early detection of acute lung
injury uncoupled to hypoxemia in pigs
using ultrasound lung comets. Crit Care
Med 35:2769–2774
30. Lichtenstein D (2012) Fluid
administration limited by lung
sonography: the place of lung
ultrasound in assessment of acute
circulatory failure (the FALLS-
protocol). Expert Rev Respir Med
6:155–162
31. Martin K, Iba
´n
˜ez L, Avila L, Barre
´S,
Kaspersen J (2005) Integrating
segmentation methods from the insight
toolkit into a visualization application.
Med Image Anal 9:579–593
32. Drummond GB (1998) Computed
tomography and pulmonary
measurements. Br J Anaesth
80:665–671
33. Malbouisson LM, Preteux F, Puybasset
L, Grenier P, Coriat P, Rouby JJ (2001)
Validation of a software designed for
computed tomographic (CT)
measurement of lung water. Intensive
Care Med 27:602–608
34. Pelosi P, Goldner M, McKibben A,
Adams A, Eccher G, Caironi P,
Losappio S, Gattinoni L, Marini JJ
(2001) Recruitment and derecruitment
during acute respiratory failure: an
experimental study. Am J Respir Crit
Care Med 164:122–130
35. Patroniti N, Bellani G, Manfio A,
Maggioni E, Giuffrida A, Foti G,
Pesenti A (2004) Lung volume in
mechanically ventilated patients:
measurement by simplified helium
dilution compared to quantitative CT
scan. Intensive Care Med 30:282–289
36. Puybasset L, Cluzel P, Gusman P,
Grenier P, Preteux F, Rouby JJ (2000)
Regional distribution of gas and tissue
in acute respiratory distress syndrome.
I. Consequences for lung morphology.
CT scan ARDS study group. Intensive
Care Med 26:857–869
37. Gargani L, Doveri M, D’Errico L,
Frassi F, Bazzichi ML, Delle Sedie A,
Scali MC, Monti S, Mondillo S,
Bombardieri S, Caramella D, Picano E
(2009) Ultrasound lung comets in
systemic sclerosis: a chest sonography
hallmark of pulmonary interstitial
fibrosis. Rheumatology (Oxford)
48:1382–1387
38. Bouhemad B, Brisson H, Le-Guen M,
Arbelot C, Lu Q, Rouby JJ (2011)
Bedside ultrasound assessment of
positive end-expiratory pressure-
induced lung recruitment. Am J Respir
Crit Care Med 183:341–347
39. Bouhemad B, Liu ZH, Arbelot C,
Zhang M, Ferarri F, Le-Guen M, Girard
M, Lu Q, Rouby JJ (2010) Ultrasound
assessment of antibiotic-induced
pulmonary reaeration in ventilator-
associated pneumonia. Crit Care Med
38:84
40. Bellani G, Messa C, Guerra L,
Spagnolli E, Foti G, Patroniti N,
Fumagalli R, Musch G, Fazio F, Pesenti
A (2009) Lungs of patients with acute
respiratory distress syndrome show
diffuse inflammation in normally
aerated regions: a [18F]-fluoro-2-
deoxy-D-glucose PET/CT study. Crit
Care Med 37:2216
41. Mallamaci F, Benedetto F, Tripepi R,
Rastelli S, Castellino P, Tripepi G,
Picano E, Zoccali C (2010) Detection of
pulmonary congestion by chest
ultrasound in dialysis patients. JACC
Cardiovasc Imaging 3:586–594
42. Bedetti G, Gargani L, Corbisiero A,
Frassi F, Poggianti E, Mottola G (2006)
Evaluation of ultrasound lung comets
by hand-held echocardiography.
Cardiovasc Ultrasound 4:34
43. Shah S, Noble VE, Umulisa I,
Dushimiyimana JM, Bukhman G,
Mukherjee J, Rich M, Epino H (2008)
Development of an ultrasound training
curriculum in a limited resource
international setting: successes and
challenges of ultrasound training in
rural Rwanda. Int J Emerg Med
1(3):193–196
44. Jambrik Z, Gargani L, Adamicza A,
Kaszaki J, Varga A, Forster T, Boros
M, Picano E (2010) B-lines quantify the
lung water content: a lung ultrasound
versus lung gravimetry study in acute
lung injury. Ultrasound Med Biol
36:2004–2010
45. Delle Sedie A, Doveri M, Frassi F,
Gargani L, D’Errico G, Pepe P,
Bazzichi L, Riente L, Caramella D,
Bombardieri S (2010) Ultrasound lung
comets in systemic sclerosis: a useful
tool to detect lung interstitial fibrosis.
Clin Exp Rheumatol 28(5 Suppl
62):S54
