| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Meakins-Christie Laboratories, Montréal, Québec, Canada H2X 2P2; and Centro di Bioingegneria-Politecnico di Milano, Fondazione Don Gnocchi, 1-20148 Milano, Italy
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
ACKNOWLEDGEMENTS
FOOTNOTES
REFERENCES
ABSTRACT 
Cala, S. J., C. M. Kenyon, G. Ferrigno, P. Carnevali, A. Aliverti, A. Pedotti, P. T. Macklem, and D. F. Rochester. Chest wall and lung volume estimation by optical reflectance motion analysis.
J. Appl. Physiol. 81(6):
2680-2689, 1996.
Estimation of chest wall motion by surface
measurements only allows one-dimensional measurements of the chest
wall. We have assessed an optical reflectance system (OR), which tracks
reflective markers in three dimensions (3-D) for respiratory use. We
used 86 (6-mm-diameter) hemispherical reflective markers arranged
circumferentially on the chest wall in seven rows between the sternal
notch and the anterior superior iliac crest in two normal standing
subjects. We calculated the volume of the entire chest wall and
compared inspired and expired volumes with volumes
obtained by spirometry. Marker positions were recorded by four TV
cameras; two were 4 m in front of and two were 4 m behind the subject.
The TV signals were sampled at 100 Hz and combined with grid
calibration parameters on a personal computer to obtain the 3-D
coordinates of the markers. Chest wall surfaces were reconstructed by
triangulation through the point data, and chest wall volume was
calculated. During tidal breathing and vital capacity maneuvers and
during CO2-stimulated hyperpnea, there was a very close correlation of the lung volumes
(VL) estimated by spirometry
[VL(SP)] and OR
[VL(OR)]. Regression
equations of VL(OR)
(y) vs.
VL(SP)
(x,
BTPS in liters) for the two subjects were given by y = 1.01x 
0.01 (r = 0.996) and
y = 0.96x + 0.03 (r = 0.997), and by
y = 1.04x + 0.25 (r = 0.97) and
y = 0.98x + 0.14 (r = 0.95) for the two maneuvers,
respectively. We conclude spirometric volumes can be estimated very
accurately and directly from chest wall surface markers, and we
speculate that OR may be usefully applied to calculations of chest wall
shape, regional volumes, and motion analysis.
chest wall motion; imaging; respiratory mechanics
INTRODUCTION EXISTING METHODS for noninvasive measurement of tidal
volume (VT) have considered
VT as the sum of changes in
volume swept by the rib cage and abdomen between end expiration and end
inspiration. Within limits, each of these two compartments has only a
single degree of freedom, so that when appropriately calibrated,
measurement of the motion of the two parts can be converted to the
volumes swept by them, which, when summed, give
VT. Current techniques, utilizing linearized magnetometry to measure distance or inductance plethysmography to measure cross-sectional area and estimate
VT, ignore small but systematic
distortions of the rib cage. Recently, the rib cage has been considered
as a two-compartment system, the lung-apposed or pulmonary rib cage and
the diaphragm-apposed or abdominal rib cage (1, 5, 19). The validity of
the calibration coefficients obtained experimentally to convert one or
two dimensions to volume is limited to the estimation of
VT under conditions matched to
those during which the calibration was performed. Other investigators
have reported the errors introduced by their application to other
conditions (4, 13, 18, 20). In addition, particularly in a clinical
setting, adequate calibration data may be difficult to obtain because
of poor subject cooperation.
Alternatively, as discussed by Konno and Mead (11), a geometric
technique could be used to describe the relationship between the volume
contained by the parts of the chest wall and to calculate their
respective volume changes. This approach has the potential to dispense
with calibration factors and could therefore be free of any associated
errors. Noninvasive three-dimensional (3-D) techniques have already
been employed to obtain measurements of VT directly, although each has
particular limitations. The 3-D X-ray-computed tomography, using the
Dynamic Spatial Reconstructor (12), is capable of measuring a known
volume to within 2% and has adequate temporal resolution to provide
dynamic images of the chest wall during respiratory movements. However,
it requires high doses of ionizing radiation, is limited to the supine
posture, and is unsuitable for long analyses. Optical systems based on sculptured light (16) have low temporal resolution, require time-consuming data processing, and are unavailable for routine use.
Technical developments in image processing and parallel computing have
enabled the analysis of the movements of multiple points on the body
surface to be performed using an optical reflectance motion-analysis
system (OR; 2, 3, 6, 10). The television image processor, known as
ELITE (ELaboratore di Immagini TElevisive; Milan Polytechnic, Milan,
Italy), OR is a digitized vision system designed to
identify objects of a predetermined shape and to monitor their
trajectories in 3-D and in real time. It has been used extensively as
an automatic movement analyzer in biomechanics and in orthopedic and
neurological medicine (6, 9, 15). Recently, OR was adapted to estimate
directly the volume change of the chest wall during respiration by
computing the 3-D coordinates of 32 markers placed on the rib cage and
abdomen (7). A geometric model based on 54 tetrahedrons was used for
computation of volume in which the chest wall between the second rib
and umbilicus was measured as upper rib cage, lower rib cage, and
abdomen. Although the correlation coefficient of the relationship
between spirometrically (SP) obtained lung volume
(VL)
[VL(SP)] and that
computed by the model was high (0.98),
VL calculated by OR
[VL(OR)] was
substantially less than VL(SP)
(13.3% error ATPS; 21.3% error
BTPS). This difference clearly limited the ability of OR to predict
VL directly, because a
calibration factor was still required.
In this study, we report the use of OR to estimate the change in
VL
( METHODS OR
VL), using more extended
boundaries of the chest wall than those reported by Ferrigno et al.
(7). We hypothesized that a direct estimate of
VL based on changes in chest
wall volume required circumferential measurements of chest wall
displacement between the thoracic inlet and the pelvis. In the study by
Ferrigno et al. (7), the portion of the abdominal compartment caudal to
the umbilicus, the rib cage cranial to the second rib, and the chest
wall in the axillary region were not included in the imaging field.
Hence the underestimate of VL in
the study by Ferrigno et al. (7) was likely to be because a substantial
fraction of VL resulted in
displacement of the chest wall beyond the boundaries used in their
analysis. A prediction arising from our hypothesis was that the
accuracy of VL(OR) could be
improved if the number of markers were increased to include these
regions. Therefore, the main difference between this study and that of
Ferrigno et al. (7) is that we have utilized almost three times the
number of markers to create a much larger marker array, arranged in a circumferential rather than cubic configuration. We further anticipated that by calculating the relationship between the error in
VL(OR) and the number of
markers utilized in the analysis we would be able to quantify the
minimum number of markers required to achieve a given level of
accuracy. Finally, we aimed to test the hypothesis that chest wall
shape could be estimated by a simple geometrical model.
Sensor Placement
We used 86 IR-reflective markers in a configuration shown in Fig. 1. The grid system consisted of seven horizontal rows arranged circumferentially between the level of the clavicles and the anterior superior iliac spine. Along the horizontal rows, the markers were arranged anteriorly and posteriorly in five vertical columns, and there was an additional bilateral column in the midaxillary line. The anatomical landmarks for the horizontal rows were, craniocaudally, 1) the clavicular line; 2) the manubrio-sternal joint (adjacent to the 2nd rib); 3) the nipples (~5th rib); 4) the xiphoid process [approximately the top of the area of apposition of the diaphragm to the rib cage at functional residual capacity (FRC) in the upright posture, confirmed by percussion]; 5) the lower costal margin (10th rib in midaxillary line); 6) umbilicus; and 7) anterior superior iliac spine. Along the horizontal rows, the markers were placed in the following positions: 1) the midlines (anteriorly, along the sternum and continuing caudally below the xiphoid through the umbilicus, and posteriorly, along the spinous processes of the vertebral column); 2) both anterior and posterior axillary lines; 3) the midpoint of the interval between the midline and the anterior axillary line (just medial to the nipple line) and the midpoint of the interval between the midline and the posterior axillary line; and 4) the midaxillary lines.
To provide better detail of the costal margin, we included an extra marker bilaterally at the midpoint between the xiphoid and the most lateral portion of the 10th rib; to increase marker density in regions where the markers were well separated, we added two markers in the region overlying the lung-apposed rib cage and two markers in the corresponding posterior positions.
Consequently, there were 42 anterior markers, 34 posterior markers, and 10 lateral markers. With the exception of the midaxillary positions, all markers were attached to the skin with double-sided adhesive tape. Because the midaxillary markers were not visible to the cameras side on, they were mounted on rear-facing right-angle brackets, before being attached to the skin.
Experimental Setup
The experimental setup is shown schematically in Fig. 2. Two IR CCD cameras were positioned 4 m in front of, and two were 4 m behind, the subject. Each pair of cameras was arranged vertically. The bottommost was at waist height and parallel to the floor; the camera ~1.5 m above was inclined downward. The subject stood with hands on the waist in the center of a space where the calibration routine for the image processor had previously been carried out. This allowed direct visualization of the anterior and posterior markers by their respective cameras. The calibration involved calculating the 3-D position of an array of sensors whose alignment on the grid was already known.
The subject breathed on a mouthpiece connected to a 9-liter water-displacement spirometer (Warren Collins) containing room air. For rebreathing experiments, the soda lime absorber was removed. The spirometer had been checked for leaks and calibrated before the experiment, using a set of syringes ranging from 1 to 5 liters, and was fitted with a potentiometer, the analog output of which was transmitted via an analog-to-digital (A-to-D) board to the computer.
Experimental Protocol
We studied two normal male subjects (S1 and S2, ages 27 and 36 yr). One (S1) was experienced in respiratory maneuvers. Subjects had vital capacities (VC) of 5.2 and 5.1 liters, respectively. During the experiment, the subjects carried out a number of ventilatory maneuvers. Initially, after a period of quiet breathing, they performed three slow VC maneuvers. Next, after closing the glottis, they performed belly-in, belly-out isovolume maneuvers at residual volume (RV), ~50% expiratory reserve volume (ERV), FRC, end of tidal breath, ~50% inspiratory capacity (IC), and total lung capacity (TLC). Then, by removing the CO2 absorber and allowing the CO2 to accumulate within the spirometer, they performed a CO2-induced hyperpnea lasting ~3 min and terminated when the VT was ~100% greater than baseline. Baseline drift of the spirometer due to O2 consumption (
O2) in the first maneuver
was corrected for by comparing FRC levels at the beginning and end of
the recording and adjusting for the slope. Spirometer drift in the
second experiment was more complicated because the subjects were not in
a steady state and the respiratory exchange ratio was not unity. Thus
we took the difference between the spirometer and the OR measurements at end expiration at the end of the run (only) and corrected for an
(assumed) constant slope. Finally, to assess the effect of change in
truncal posture at isovolume of lung
(VLiso)
on the estimation of chest wall volume by OR, the subjects performed two additional maneuvers after closing the glottis at relaxed FRC with
both hands placed on the waist. First, they flexed the trunk laterally
from the waist to ~15° from the vertical, held the posture
briefly, and then returned to the original position. Second, they
elevated the shoulder girdle by 4-5 cm, held the position
constant, and returned to baseline again.
Volume Calculation
Volume displacement of the chest wall was calculated by triangulating the surface and integrating the subtended volume. The actual triangulation used is shown in Fig. 3. Note that additional "virtual" points were constructed at various margins to enable easier triangulation. For example, a point was constructed at the mean position of all the points on the anterior superior iliac spine level. Similar virtual point construction and triangulation allow completely arbitrary subdivision of the chest wall. In particular, the detail points on the lower costal margin are useful for division of the chest wall into anatomic compartments.
The steps involved in the calculation are as follows. Having defined an
arbitrary 3-D (x,
y,
z) coordinate system, we found it
possible to define the x,
y, and
z coordinates of each marker, the area
(Ai), the
x, y,
z coordinates of the centroid, and the distance (di) of each triangle
i to an arbitrary reference plane.
From the direction of the normal to each triangle
i, the angle of the plane defined by
that triangle with respect to the reference plane
(qi) can be
determined. Each triangle is then used to define a volume element
(Vi) with respect to an
arbitrary reference plane where
Vi = Ai · di · cosqi.
The term cosqi
will be positive for the triangles, the outer faces of which are away
from the reference plane and negative for those pointing toward it.
Total volume (V) of the thorax enclosed within the surface defined by
these triangles is given by the summation of all volume elements, i.e.,
V = 
iVi. This procedure is
equivalent to the Gauss theorem. It is important to note that it was
not possible to calculate absolute
VL by using OR. All estimates of
VL(OR) were therefore calculated
as the difference between estimates of chest wall volume, which could
then be compared with the spirometric equivalent. For statistical
comparisons of the change in
VL(SP) vs.
VL(OR), we calculated the mean
difference and divided by the SD to obtain the coefficient of
variation. We defined accuracy as the coefficient of variation for the
maneuvers studied where this was applicable and unless otherwise
stated.
Sensitivity Analysis
To determine the dependence of the regression analysis on the number of points, we systematically removed horizontal and vertical rows of points in the volume calculation and recalculated the regression analysis. Horizontal rows were removed both singly and in adjacent pairs, and vertical rows were removed singly. Because we calculated chest wall volume by triangulation between adjacent points, combinations of nonadjacent row removals were simply additive in terms of effect. We also assessed the effect of removing the extra "detail" points.To investigate the theoretical connection between the number of points and expected accuracy, we modeled the cross section of the chest wall using an "athletic track" shape (two semicircles separated by a rectangle). By varying the number of points on the curved regions we could obtain a theoretical limit on the possible accuracy of any number of points on a particular space along its arc (n >2); then the area enclosed by the polygon P that they form is
![P = (<IT>n</IT> − 1)*<IT>r</IT><SUP>2</SUP>* sin [&Pgr;/(<IT>n</IT> − 1)]](/content/vol81/issue6/fulltext/2680/img001.gif)
|
(1) |
![Error (%) = (&Pgr;*<IT>r</IT><SUP>2</SUP> − 2&Pgr;)/ [(&Pgr;*<IT>r</IT><SUP>2</SUP>) + (2<IT>r</IT>*s)]](/content/vol81/issue6/fulltext/2680/img002.gif)
|
(2) |
Finally, an important issue is the time required by the operator(s) to calibrate the equipment, attach markers to the subject, and analyze the data. To minimize operator time, a number of automated routines have been incorporated into the calibration, data processing, and data reduction phases. First, the space occupied by the subject is calibrated before the experiment, using a square pegboard array consisting of a grid of markers having the same dimensions as those to be used in the study at a known constant distance from one another. This is imaged by the system, and the calibration is performed automatically by the software in seconds without operator intervention beyond specifying which calibration grid is to be used. The algorithm contained in the software has been previously described (2). Calibration involves two persons and takes ~15 min to set up and perform. The 86 markers are attached manually, in 30-45 min, depending on the speed and accuracy of the operator in locating the anatomical landmarks, but as noted, it is not necessary to recalibrate afterwards. Finally, the data-reduction stage has been considerably streamlined to minimize operator time, using a two-level architecture incorporating a number of automated calculations. The final level of calculations, namely volume estimation, can easily be performed by a PC rather than workstation. The most time-consuming step in volume estimation (2-3 min) is manual labeling of each of the 86 markers in the first frame of each sequence. Henceforth, remaining steps are highly automated. The hardware processor analyzes the images in real time. Then, using a technique based on a convolution operator, the processor recognizes within each frame objects having a specific shape. The coordinates of the markers are fed into the computer, and marker trajectories are then tracked automatically.
RESULTS
Isovolume Maneuvers
VL(OR) during isovolume maneuvers over the VC is shown for the two subjects in Fig. 4. In S2, the amplitude of oscillations during the isovolume maneuver was consistently <250 ml, irrespective of VL. At TLC, there was a ~600 ml decrease in VL(OR) over the first ~5 s before a stable value was attained. Conversely, the VL(OR) baseline increased by ~300 ml for the first ~5 s at RV before reaching stability. In S1, the fluctuations in VL(OR) during the isovolume maneuver were larger, especially at VL above FRC, with a comparable downward drift in the VL(OR) baseline at TLC.
Comparison of OR with Spirometer
Quiet breathing and VC maneuvers. Figure 5 shows a representative example of a VC breath bracketed by tidal breathing in S1 and S2, comparing estimates of VL(OR) and VL(SP) and regressions of VL(OR) vs. VL(SP) corresponding to the same data. The regression equations relating VL(OR) to VL(SP) for S1 and S2 are VL(OR) = 1.04 VL(SP) + 0.02 (r = 0.995) and VL(OR) = 0.99 VL(SP) 0.01 (r = 0.995), respectively. Analysis of
the residual error from the regression of
VL(OR) vs.
VL(SP) revealed that 96% of the
points were within 250 ml of the regression line in both subjects and
that the coefficient of variation was 1.9 and 2.3% for S1 and S2,
respectively.
VL) by spirometry (SP) and OR
(left) and regressions of
VL by SP vs. OR corresponding
to same data (right).
Involuntary CO2-Stimulated Hyperpnea
Figure 6 shows representative data from the first and last 30 s of CO2-induced hyperpnea, comparing VL(OR) and VL(SP) results for S1 and S2 and regressions of VL(OR) vs. VL(SP) corresponding to the same data. The regression equations for S1 and S2 were VL(OR) = 1.00 VL(SP) 0.05 (r = 0.97) and
VL(OR) = 0.99 VL(SP) 0.04 (r = 0.97), respectively. The
coefficient of variation of the residual error was 3.6 and 4.4% for S1
and S2, respectively, and 96% of the points were within 350 ml of the
regression line.
VL by SP vs. OR corresponding
to same data (right).
Sensitivity Analysis
Marker number. The effect of marker removal on the slope of the relationship between VL(OR) and VL(SP) was dependent on both the maneuver and the subject (Fig. 7). Chest wall volume estimation was more sensitive to marker removal during hyperpnea maneuvers than during simple quiet breathing and VC. For the removal of vertical lines of markers, there was always an increase in the slope of between 3 and 10% for the midlines on the back (S2) and the side markers (S1). A combination of side and midline markers resulted in increases of slope from 5% (S2) to 20% (S1). The effect of horizontal line removal was most important at the clavicular and anterior superior iliac spine levels in both subjects and maneuvers, and at the lower costal margin level for S1, leading to an increase in slope by ~10%. The intercepts of the regression lines were fairly insensitive to the removal of vertical or horizontal lines of marker with a range of changes of only 0.2 liters.
VL. Inverted triangles, S1;
circles, S2; open symbols, quiet breathing and VC maneuvers; filled
symbols, CO2-induced hyperpnea;
r values >0.93 and
P < 0.001 in all cases.
Top: vertical lines of markers removed
were N, none; D, detail markers; BM, back midlines; M, midlines; S, sides; S+C, sides and centerlines; S+M, sides and midlines.
Bottom: horizontal lines of markers
removed: 1, jugular notch; 2, angle of Louis; 3, nipple; 4, xiphoid; 5, lower costal margin; 6, umbilicus; 7, anterior superior iliac crest.
Posture. The effect of posture on the accuracy of VL(OR) is shown in Fig. 8. When the glottis was closed at FRC and the subject performed a slight lateral flexion maneuver of the trunk to ± 10°, there was no systematic effect on VL(OR). In contrast, there was a substantial (~700 ml) apparent increase in VL(OR) when the subject elevated the shoulders by ~4 cm at closed-glottis FRC.
Cross-sectional area. The error in cross-sectional area was calculated from the athletic track model of two semicircles separated by a rectangle (Fig. 9). The semicircle describes the lateral portion of the chest wall bounded medially by a sagittal plane located halfway between the nipple line and the midline. The percent error is inversely related to the number of markers on a semicircle going from >30% underestimate with three markers to ~10% with five markers to >5% with seven markers. We used 12 markers on each horizontal cross section, with five markers on each semicircle. Given the anteroposterior and transverse diameters of the two subjects, the area of the central rectangle ranged from 22% at RV to 26% at TLC. Although the relative dimensions of these central rectangular sections are not equal, we have neglected the small differences in their relative dimensions because they have the same shape. Thus we expect a cross-sectional error of
10% × (100 22)/100 to
10% × (100 26)/100, i.e., 7 to 8%,
or gradients of 0.92 to 0.93, because this model is realistic and error
in cross-sectional area translates directly into volume error.
DISCUSSION
With the use of an optical system to image 86 markers arranged
circumferentially between the clavicles and the anterior superior iliac
crest, it was possible to estimate
VL directly with a coefficient of variation of <2% for VL,
<3.5% for VL, and <1%
for chest wall volume. This was a substantial improvement
compared with the previous method (~21% error,
BTPS), was obtained
without a calibration correction factor, and applied to the entire VC
range under static and dynamic conditions. When marker density was
decreased, the error in estimating
VL increased markedly on removal
of the cranial and caudal horizontals and, to a lesser extent, for the
markers in the axillae. Assuming an athletic track model for chest wall cross-sectional shape, there was a hyperbolic relationship between error in VL and marker number.
When the error prediction for a given cross section was
extrapolated to the entire chest wall, the predicted error (7-8%)
was somewhat larger than the range observed in this study.
Critique of Methods
Measurement errors of SP. Potential sources of error in VL(SP) during prolonged breathing include those due to gas leakage and temperature elevation within the system, thus causing spurious increases and decreases in measured VL, respectively. In non-steady-state breathing, especially as in our second experiment when the soda lime absorber was removed, the volume ratio between CO2 elimination andO2 becomes important and
would presumably have been diminished, because the mixture breathed
would decrease CO2 elimination.
When a weight was added to the sealed system, the deflection in the
volume baseline was <25 ml/min. Although gas temperature within the
drum was not measured, the length of tubing interposed between the
subject and spirometer was ~2 m, so that heat loss from the tube
should have dissipated most of the heat of the expired water vapor,
thus minimizing temperature change within the system. Offline, to
facilitate comparison of the two estimates of
VL, the
O2 slope was calculated and
assumed as a constant to correct the end expiratory baseline values. In the rebreathing runs, the difference between the two estimates of
VL at the end of the run was
used to correct the spirometric drift. There was a systematic increase
in spirometric volumes of about 0.4 l/min, probably due to a decrease
in CO2 consumption and an increase
in O2 by ~150-200
ml/min above a hypothetical baseline value for
O2 of ~250-300 ml/min.
Because the subjects were standing rather than seated,
O2 at rest and during
hyperpnea may have increased because of energy consumption by postural
muscles.
Changes in chest wall volume not due to gas movement.
We assumed that the only factor causing chest wall volume changes was
gas movement. However, lung gas volume would be affected by changes in
alveolar pressure from FRC to TLC relaxed with glottis closed of
probably ~30 cmH2O (0 to ~30
cmH2O), which, at normal atmospheric pressure (~1,000
cmH2O), would give a 3% decrease in gas volume from active TLC with glottis open to fully relaxed with
glottis closed. Alternatively, RV to FRC alveolar pressure changes
during relaxation by ~20 cmH2O
(20 to 0 cmH2O), suggesting increases of ~2%. We observed a consistent decrease in
VL during isovolume maneuvers at
TLC of ~0.5 liter, which for the two subjects was ~6% (Fig. 4).
Below FRC, there were consistent increases in volume only in S1 of
~5%. However, relaxation at RV is difficult, and only S1 was highly
trained in respiratory maneuvers. Ventilation is accompanied by flux of
blood into and out of the thorax from the head, abdomen, pelvis, and
legs. The magnitude of the respiratory-related blood volume shifts is
determined by two factors. The first is the effect of pleural pressure
swings on beat-to-beat variation in cardiac output. During inspiration,
venous return is increased slightly, and cardiac output decreased
slightly. The reverse process occurs during expiration. Flux of blood
volume across the upper and lower chest wall boundaries may lead to a
slight overestimate at end inspiration and a slight underestimate at
end expiration of change in chest wall volume estimated by OR compared
with VL(SP) and therefore to a
spuriously high VL(OR). During
isovolume maneuvers, the shifts may be somewhat larger. This may
account for the rest of the error in the changes in
VL(OR) during isovolume
maneuvers above lung gas pressure changes. The second process,
occurring over the course of the study, is progressive blood pooling in the lower limbs, which may be seen as a gradually increasing
underestimate of VL(OR) compared
with VL(SP). However, blood
pooling is not appreciable for several hours, whereas each of the
observations in these studies was concluded within 10 min.
Measurement errors of OR volume of chest wall.
The error in the calculation of the marker centroid coordinates by OR
in 3-D has been reported previously for static and dynamic conditions
to be <0.2 mm for the size of field of view used for thoracic
imaging, a 1.5 m cube (17).
The volume calculation model is that of an irregular polyhedron,
containing the chest wall and having flat triangular faces with
vertices at the reflective markers. The volume here is directly calculated from a geometric model rather than by measuring movements of
individual parts which have an assumed fixed relation to total volume
changes calculated via a calibration process. Theoretically, this is
superior in that the range of validity of the calculated volume changes
may be much greater and also no calibration factor is required.
Although we observed marker movements of 1-2 cm relative to lower
rib positions during large breaths, marker movement at the clavicular
and anterior superior iliac crest was almost absent (<2 mm),
translating into only 100-300 ml of volume change.
Clearly, a polyhedron model is only an approximation to the curved
surface of the chest wall, but our experiments indicate a very high
accuracy, with an acceptable error of <3.5% for changes in
VL. However, this accuracy is
greater than the theoretical limit, based on an athletic track model of
chest wall cross section (7-8%, Fig. 9). We hypothesize this is
because the cross section of the chest wall is less curved than the
athletic track model. In particular, the back is flatter, as are the
sides of the rib cage and abdomen. This suggests markers should be
positioned on chest wall "corners," because these are well
defined. This also implies that refining the volume calculation model
by using a curved rather than a flat approximation for the surface is
inappropriate.
Sensitivity of VL calculation to
number of markers.
The slope of the relationship between
VL by the optical and
spirometric methods was markedly, but nonsystematically, influenced by
the progressive deletion of markers, in particular horizontal and
vertical lines (Fig. 7). The greatest changes, of ~10%, in both
subjects resulted from removing horizontal rows of markers at the
clavicles or the anterior superior iliac crest. Furthermore, in one
subject (S1), the removal of the vertical lines of side markers also
gave a large error (10%). This was probably because of a more rounded
chest wall shape than in S2. The intercept of the regression
relationships was relatively insensitive to any changes, probably
because all of the points were either within a moving "part" or
very close to its anatomical boundary, so that no significant volume
could be included or removed by removing markers.
In light of these results, an important contribution to the error in
the study by Ferrigno et al. (7) was the model used for
VL estimation. The results
suggest that the explanation for the ~20% error obtained previously
in estimating VL was due to an
approximately equal contribution of neglect of abdominal volume below
the umbilicus (10%) and lack of side markers (~10%). Therefore, when calculating the minimum number of markers required to obtain reasonable accuracy of estimating
VL (<10%), it is evident that the detail markers and those at the xiphoid and umbilical horizontal levels are clearly expendable, saving 28 markers and leaving 58.
In summary, using an OR to calculate the volume of the chest wall
during respiration over the VC in two normal standing subjects, we
obtained a very high level of agreement between the calculated change
in VL and SP. The results of
VL estimation by the optical method were obtained directly, without requiring a calibration factor.
Systematic elimination of markers from the volume algorithm showed that
the accuracy of the calculations was critically dependent on inclusion
of data from the cranio-caudal and lateral extremities of the chest
wall. In addition, with the assumption of an athletic track model of
the chest wall in cross section, a minimum of ten markers is required
for an estimate of cross-sectional area within 5%.
The system has the advantage of suitability for use in unconstrained
subjects breathing without a mouthpiece and has particular advantages
in patient and/or pediatric studies where cooperation may be
difficult to obtain. Once the sensors have been applied, no additional
time costs are incurred during the experiment, and the on-line data
processing makes it possible to analyze the very large data sets
quickly and with relatively modest data storage requirements. The chest
wall reconstruction computations are simple and easily modified to
accommodate a greater or lesser number of markers. Because the chest
wall can be imaged directly in 3-D, the data is amenable to a wide
range of analyses, such as statistical analysis of chest wall
coordination (10) and estimation of chest wall compartment volumes (7).
ACKNOWLEDGEMENTS
This work was supported by the Allen and Hanburys Thoracic Society of Australia and New Zealand Fellowship, the J. T. Costello Memorial Research Fund, the Medical Research Council of Canada, Respiratory Health Network of Centres of Excellence (Canada), Pro Juventute Don Carlo Gnocchi Foundation, and TELETHON Italia.
FOOTNOTES
Address for reprint requests: P. T. Macklem, Montreal Chest Hospital Centre, 3650 St. Urbain, Montreal, Quebec, Canada H2X 2P4.
Received 3 October 1995; accepted in final form 7 August 1996.
REFERENCES
| 1. | Agostoni, E., and E. D'Angelo. Statics of the chest wall. In: The Thorax, edited by C. Roussos, and P. T. Macklem. New York: Dekker, 1988, vol. 29. pt. A, p. 259-295. (Lung Biol. Health Dis. Ser.) |
| 2. | Borghese, N. A., and G. Ferrigno. An algorithm for 3-D automatic movement detection by means of standard TV cameras. IEEE Trans. Biomed. Eng. 32: 1221-1225, 1990. |
| 3. | Carnevali, P., G. Ferrigno, A. Aliverti, and A. Pedotti. A new method for 3-D optical analysis of chest wall motion. Technol. Health Care 4: 43-65, 1996. |
| 4. | Cramer, D., A. Peacock, and D. Denison. Temperature corrections in routine spirometry. Thorax 39: 771-774, 1984. |
| 5. | Decramer, M., A. De Troyer, S. Kelly, L. Zocchi, and P. T. Macklem. Regional differences in abdominal pressure swings in dogs. J. Appl. Physiol. 56: 1682-1687, 1984. |
| 6. | Ferrigno, G., N. A. Borghese, and A. Pedotti. Pattern recognition in 3-D motion analysis. ISPRS J. Photogramm. Remote Sens. 45: 227-246, 1990. |
| 7. | Ferrigno, G., P. Carnevali, A. Aliverti, F. Molteni, G. Beulke, and A. Pedotti. Three-dimensional optical analysis of chest wall motion. J. Appl. Physiol. 77: 1224-1231, 1994. |
| 8. | Ferrigno, G., and A. Pedotti. ELITE: a digital dedicated hardware system for movement analysis via real-time TV signal processing. IEEE Trans. Biomed. Eng. 32: 943-950, 1985. |
| 9. | Frigo, C. Three-dimensional model for studying the dynamic loads on the spine during lifting. Clin. Biomech. 5: 143-152, 1990. |
| 10. | Kenyon, C. M., R. H. Ghezzo, S. J. Cala, G. Ferrigno, A. Pedotti, P. T. Macklem, and D. F. Rochester. Statistical methods for analysis of coordination of chest wall motion using optical reflectance imaging of multiple markers. In: Mathematical Methods in Medical Imaging III. Proceedings of the Society for Optical Engineering, edited by F. L. Bookstein, J. S. Duncan, N. Lange, and D. C. Wilson. San Diego, CA: 25-26 July 1994, p. 169-179. |
| 11. | Konno, K., and J. Mead. Measurement of the separate volume changes of rib cage and abdomen during breathing. J. Appl. Physiol. 22: 407-422, 1967. |
| 12. | Krayer, S., K. Rehder, K. C. Beck, P. D. Cameron, E. P. Didier, and E. A. Hoffman. Quantification of thoracic volumes by three-dimensional imaging. J. Appl. Physiol. 62: 591-598, 1987. |
| 13. | Levine, S., D. Silage, D. Henson, J. Wang, J. Krieg, J. LaManca, and S. Levy. Use of a triaxial magnetometer for respiratory measurements. J. Appl. Physiol. 70: 2311-2321, 1991. |
| 14. | Loring, S. H., and E. N. Bruce. Methods for study of the chest wall. In: Handbook of Physiology. The Respiratory System. Mechanics of Breathing. Bethesda, MD: Am. Physiol. Soc., 1986. sect. 3, vol. III, pt. 2, chapt. 24, p. 415-428. |
| 15. | Mouchnino, L., R. Aurenty, J. Massion, and A. Pedotti. Coordination between equilibrium and head-trunk orientation during leg movement: a new strategy built up by training. J. Neurophysiol. 67: 1587-1598, 1992. |
| 16. | Peacock, A. J., M. D. L. Morgan, S. Gourlay, C. Tourton, and D. M. Denison. Optical mapping of the thoraco-abdominal wall. Thorax 39: 93-100, 1984. |
| 17. | Pedotti, A., and G. Ferrigno. Opto-electronics based systems. In: Three-Dimensional Analysis of Human Movement, edited by P. Allard, I. A. F. Stokes, and J. P. Bianchi. Champaign, IL: Human Kinetics, 1995. chapt. 4, p. 57-78. |
| 18. | Sackner, M. A., H. Watson, A. S. Belsito, D. Feinerman, M. Suarez, G. Gonzalez, F. Bizousky, and B. Krieger. Calibration of respiratory inductance plethysmography during natural breathing. J. Appl. Physiol. 66: 410-420, 1989. |
| 19. | Ward, M. E., J. W. Ward, and P. T. Macklem. Analysis of human chest wall motion using a two-compartment rib cage model. J. Appl. Physiol. 72: 1338-1347, 1992. |
| 20. | Zimmerman, P. V., S. J. Connellan, H. C. Middleton, M. V. Tabona, M. D. Goldman, and N. Price. Postural changes in rib cage and abdominal volume-motion coefficients and their effect on the calibration of a respiratory-inductive plethysmograph. Am. Rev. Respir. Dis. 127: 209-214, 1983. |
This article has been cited by other articles:
|
|
 |
|
  A. Aliverti, M. Quaranta, B. Chakrabarti, A. L. P. Albuquerque, and P. M. Calverley Paradoxical movement of the lower ribcage at rest and during exercise in COPD patients Eur. Respir. J., January 1, 2009; 33(1): 49 - 60. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. M. A. Calverley Exercise and dyspnoea in COPD Eur. Respir. Rev., December 1, 2006; 15(100): 72 - 79. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. T. Macklem Circulatory effects of expiratory flow-limited exercise, dynamic hyperinflation and expiratory muscle pressure Eur. Respir. Rev., December 1, 2006; 15(100): 80 - 84. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A Aliverti, K Rodger, R L Dellaca, N Stevenson, A Lo Mauro, A Pedotti, and P M A Calverley Effect of salbutamol on lung function and chest wall volumes at rest and during exercise in COPD Thorax, November 1, 2005; 60(11): 916 - 924. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. E. Finucane, J. A. Panizza, and B. Singh Efficiency of the normal human diaphragm with hyperinflation J Appl Physiol, October 1, 2005; 99(4): 1402 - 1411. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. De Groote, A. Van Muylem, P. Scillia, G. Cheron, G. Verleden, M. Paiva, and M. Estenne Ventilation Asymmetry after Transplantation for Emphysema: Role of Chest Wall and Mediastinum Am. J. Respir. Crit. Care Med., December 1, 2004; 170(11): 1233 - 1238. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I. Romagnoli, F. Gigliotti, A. Galarducci, B. Lanini, R. Bianchi, D. Cammelli, and G. Scano Chest wall kinematics and respiratory muscle action in ankylosing spondylitis patients Eur. Respir. J., September 1, 2004; 24(3): 453 - 460. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. L. Dellaca, L. D. Black, H. Atileh, A. Pedotti, and K. R. Lutchen Effects of posture and bronchoconstriction on low-frequency input and transfer impedances in humans J Appl Physiol, July 1, 2004; 97(1): 109 - 118. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Aliverti, B. Kayser, and P. T. Macklem Breath-by-breath assessment of alveolar gas stores and exchange J Appl Physiol, April 1, 2004; 96(4): 1464 - 1469. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A Aliverti, N Stevenson, R L Dellaca, A Lo Mauro, A Pedotti, and P M A Calverley Regional chest wall volumes during exercise in chronic obstructive pulmonary disease Thorax, March 1, 2004; 59(3): 210 - 216. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Bianchi, F. Gigliotti, I. Romagnoli, B. Lanini, C. Castellani, M. Grazzini, and G. Scano Chest Wall Kinematics and Breathlessness During Pursed-Lip Breathing in Patients With COPD Chest, February 1, 2004; 125(2): 459 - 465. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Filippelli, R. Duranti, F. Gigliotti, R. Bianchi, M. Grazzini, L. Stendardi, and G. Scano Overall Contribution of Chest Wall Hyperinflation to Breathlessness in Asthma Chest, December 1, 2003; 124(6): 2164 - 2170. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Lanini, R. Bianchi, I. Romagnoli, C. Coli, B. Binazzi, F. Gigliotti, A. Pizzi, A. Grippo, and G. Scano Chest Wall Kinematics in Patients with Hemiplegia Am. J. Respir. Crit. Care Med., July 1, 2003; 168(1): 109 - 113. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Aliverti, G. Ghidoli, R. L. Dellaca, A. Pedotti, and P. T. Macklem Chest wall kinematic determinants of diaphragm length by optoelectronic plethysmography and ultrasonography J Appl Physiol, February 1, 2003; 94(2): 621 - 630. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Duranti, M. Filippelli, R. Bianchi, I. Romagnoli, R. Pellegrino, V. Brusasco, and G. Scano Inspiratory Capacity and Decrease in Lung Hyperinflation With Albuterol in COPD Chest, December 1, 2002; 122(6): 2009 - 2014. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C.-L. Que, C. Kolmaga, L.-G. Durand, S. M. Kelly, and P. T. Macklem Phonospirometry for noninvasive measurement of ventilation: methodology and preliminary results J Appl Physiol, October 1, 2002; 93(4): 1515 - 1526. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
ATS/ERS Statement on Respiratory Muscle Testing Am. J. Respir. Crit. Care Med., August 15, 2002; 166(4): 518 - 624. [Full Text] [PDF]
|
|
|
|
|
|
M. Filippelli, R. Pellegrino, I. Iandelli, G. Misuri, J. R. Rodarte, R. Duranti, V. Brusasco, and G. Scano Respiratory dynamics during laughter J Appl Physiol, April 1, 2001; 90(4): 1441 - 1446. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. ALIVERTI, R. DELLACÁ, P. PELOSI, D. CHIUMELLO, A. PEDOTTI, and L. GATTINONI Optoelectronic Plethysmography in Intensive Care Patients Am. J. Respir. Crit. Care Med., May 1, 2000; 161(5): 1546 - 1552. [Abstract] [Full Text]
|
|
|
|
|
|
A. de Groote, Y. Verbandt, M. Paiva, and P. Mathys Measurement of thoracoabdominal asynchrony: importance of sensor sensitivity to cross section deformations J Appl Physiol, April 1, 2000; 88(4): 1295 - 1302. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Sanna, F. Bertoli, G. Misuri, F. Gigliotti, I. Iandelli, M. Mancini, R. Duranti, N. Ambrosino, and G. Scano Chest wall kinematics and respiratory muscle action in walking healthy humans J Appl Physiol, September 1, 1999; 87(3): 938 - 946. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. GORINI, I. IANDELLI, G. MISURI, F. BERTOLI, M. FILIPPELLI, M. MANCINI, R. DURANTI, F. GIGLIOTTI, and G. SCANO Chest Wall Hyperinflation during Acute Bronchoconstriction in Asthma Am. J. Respir. Crit. Care Med., September 1, 1999; 160(3): 808 - 816. [Abstract] [Full Text]
|
|
|
|
|
|
C. M. Kenyon, S. J. Cala, S. Yan, A. Aliverti, G. Scano, R. Duranti, A. Pedotti, and P. T. Macklem Rib cage mechanics during quiet breathing and exercise in humans J Appl Physiol, October 1, 1997; 83(4): 1242 - 1255. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Aliverti, S. J. Cala, R. Duranti, G. Ferrigno, C. M. Kenyon, A. Pedotti, G. Scano, P. Sliwinski, P. T. Macklem, and S. Yan Human respiratory muscle actions and control during exercise J Appl Physiol, October 1, 1997; 83(4): 1256 - 1269. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I. Iandelli, A. Aliverti, B. Kayser, R. Dellaca, S. J. Cala, R. Duranti, S. Kelly, G. Scano, P. Sliwinski, S. Yan, et al. Determinants of exercise performance in normal men with externally imposed expiratory flow limitation J Appl Physiol, May 1, 2002; 92(5): 1943 - 1952. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Aliverti, I. Iandelli, R. Duranti, S. J. Cala, B. Kayser, S. Kelly, G. Misuri, A. Pedotti, G. Scano, P. Sliwinski, et al. Respiratory muscle dynamics and control during exercise with externally imposed expiratory flow limitation J Appl Physiol, May 1, 2002; 92(5): 1953 - 1963. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
