- Email: [email protected]
Spatio-temporal characterization of intercostal activity during breathing in the cat
J. theor. Biol. (1987) 125, 125-140
Spatio-temporal Characterization of Intercostal Activity During Breathing in the Cat C. RUGGIERO,~" B. M c A . SAYERS,~ Z. A. SEARS§ AND P. A. KIRKWOOD§
t Department of Communications, Computer and Systems Science, University of Genoa, Italy; ~. Department of Computing, Imperial College, London, U.K.; and § Sobell Department of Neurophysiology, Institute of Neurology, University of London, U.K. (Received 28 September 1985, and in revised form 7 July 1986) It is envisaged that the motor control of the intercostal musculature--an assembly of mobile structures--can be characterized in terms of a conceptual spatially continuous control function, that underlies the discretely distributed muscular activity and reflects an inferred global dynamic control of the thoracic cage during breathing. The global control function is estimated by the spatio-temporal pattern obtained by averaging in time and space and interpolation of multichannel simultaneous intercostal EMG recording in the anaesthetized cat. Different examples of the experimental preparation in the presence of stimuli of different kinds are analysed. The resultant signal patterns are found to be self-consistent and capable of exhibiting systematically differing features in systematically differing experimental conditions, thus supporting the validity of the analysis and the choice of the estimator. It is concluded that a more detailed analysis of the requirements of this approach is then warranted. Such requirements are discussed, and, specifically, results that bear on the adequacy of spatial sampling rate are presented. It is suggested that such methods offer a promising approach in the study of motor control strategies of the respiratory apparatus.
1. Introduction
A method for characterizing motor activity of the external respiratory apparatus based on spatio-temporai intercostal E M G pattern analysis is proposed, with a view to clarifying motor control strategies involved in the mechanics of breathing. In the thoracic c a g e - - a n assembly of mobile structures--breathing and postural changes take place while structural stability is maintained. The importance of the role of muscular activity associated with respiration in relation to the preservation of mechanical stability was recognized in previous work (da Silva et al., 1977), in which the kinematics of changes in configuration of the structures that participate in respiratory movements was explored. Specifically, it was suggested that spatially integrated nervous control of the intercostal muscles is essential for satisfactory performance by the external respiratory apparatus. The characterization proposed here envisages that the activity of the intercostal muscles can be interpreted in terms of a conceptual spatially continuous control function underlying the discretely distributed muscular activity. The objective of 125 0022-5193/87/060125+ 16 $03.00/0
© 1987 Academic Press Inc. (London) Ltd
126
c. RUGGIERO
ET AL
this study is to identify an estimator of the continous underlying control function, whose features can represent the global muscular effects acting on the thorax, thereby clarifying neurophysiological characteristics of breathing movements. The estimator of the underlying control function is, in this work, based on the activity of the intercostal muscles; it is formed from discrete observation of EMG, processed to produce a complete spatio-temporal pattern throughout the inspiratory phase of the breath. The problem considered here has two aspects: deriving a suitable estimator from muscular activity; then assessing whether the estimation is self-consistent and capable of exhibiting systematically differing features in systematically differing experimental conditions. If this is the case, a more detailed analysis of the requirements of this approach would be warranted. 2. A Spatio-temporal Estimator of Intercostal Activity It is envisaged that the spatially discrete intercostal muscular activity can be interpreted in terms of a conceptual underlying spatially continuous control function; the underlying function is to be estimated from multichannel simultaneous EMG recording. The electrode system needs to suppress ECG interference, to achieve a spatiallylimited field of electrical sensitivity, and to produce EMG activity that is nevertheless an adequate local "sample" and so reasonably representative of local group of muscle units. Accordingly, bipolar EMG recording was used. It was proposed to rectify and filter the EMG in each channel, smoothing sufficiently to produce a signal fluctuating only at a rate appropriate to the envisaged control action, then to estimate a continuous function by suitable spatial interpolation at each instant studied. The resulting spatial signal would then be regarded as characterizing the global muscular effects acting on the thorax. The patterns of global activity could be portrayed by means of contour maps representing the lines of intersection of the signal with equally spaced planes parallel to zero-level surface. The succession of contour patterns in time would then present a global generalization of the time evolution of the EMG distribution. These patterns are proposed as an estimator of the underlying global spatiotemporal activity that reflects an inferred dynamic control of the thorax. The influence of different experimental conditions due to stimuli of different kinds is then to be examined to assess whether this estimator varies in a distinctive and systematic fashion. 3. Experimental Methods Electromyographic recording was made from two cats anaesthetized with nembutal (initial dose 40 mg/kg) and maintained in a reasonably constant state of light anaesthesia during the experiments. The bipolar electrodes consist of pairs of fine wires; they were inserted using a serum needle which was then withdrawn, leaving the electrodes held in the muscle by their hooked ends. Further details of this technique are given by Newsom-Davis & Sears (1970) and Taylor (1960).
SPATIO-TEMPORAL
INTERCOSTAL
ACTIVITY
127
It was decided to record simultaneous E M G signals from eight points on the left part of the chest, creating thereby a fairly regular array covering the inspiratory part of half thorax (no electrodes were placed in the lowest intercostal region, not active in inspiration). Figure 1 shows the positions of the electrodes. The number of channels was limited by the available means of recording. Prior to recording, the signals were amplified in such a way as to achieve about the same peak to peak excursion of the most distinctive motor units in all channels. As a consequence, the spatio-temporal patterns obtained are only comparable within the same experiments; normalizing the data in a different way would be necessary to permit comparison of sets o f data from different experiments.
FIG. 1. Positions of the electrodes on the thorax.
The E M G signal was recorded on a PI 6200 FM tape recorder and digitized at 11 bit precision in an IBM-1800 computer. The analog to digital conversion set-up available allows for a maximum sampling rate of 20 kHz, which, when sampling eight channels simultaneously, corresponds to 2.5 kHz sampling rate per channel. Each signal contained no significant contributions above 1.25 kHz. Prior to analog to digital conversion the analog tape was replayed at 8/10 of recording tape speed and copied on a new tape, to achieve an equivalent bandwidth of 1 kHz, and the sampling rate used was 2.5 kHz, slightly greater than the 2 kHz rate required to avoid aliasing for 1 kHz bandwidth signal (as advisable to achieve good amplitude definition). Referring to the time scale of the signal, the adopted sampling rate corresponds to 3125 Hz/channel. Further analysis was carried out on an IBM 1800 and on a CDC 6500 or, equivalently, on a CII 10070. The experiments reported here refer to the following experimental conditions. For the first animal: different CO2 content in the inspired air; tracheal closure. For the second animal: denervation of one intercostal segment; localized mechanical stimulation of the thorax. Further details are specified later.
128
C. R U G G I E R O
ET
AL
4. Spatio-temporal Pattern Analysis Intercostal EMG activity is characterized by variability in time and in space. Space-temporal aspects of the respiratory EMG patterns were examined as follows. TEMPORAL
VARIABILITY
EMG signal variation in time has been analysed by (software) rectification and low-pass filtering of the signal, aiming to clarify a hypothesized underlying consistent firing pattern for each channel, minimizing the random fluctuations of the data. The natural frequency of the human respiratory system has been found to be about 5 Hz (Campbell et al., 1970). This figure offered an approximate basis for choosing the degree of data smoothing. The rectified signals were low-pass filtered to about twice this figure. The rectified EMG signal was filtered in the time domain using an autoregressive, zero phase shift, integer coefficient digital filter. Autoregressive filtering offers advantages in computational economy and speed; integer coefficient implementation eliminates problems of filter stability and accuracy caused by finite computer word length. A description and practical realization of the filter used is given by Lynn (1971, 1975). Its transfer function is (1--z2k+l)3/z3k(1--Z) 3 and its autoregressive realization is
y(n)=y(n-3)-3y(n-2)+3y(n-
1)-x(n-3k-3)+3x(n-k-2)-3x(n-k-
1)
+x(n+3k) where x, is the present input of the filter, y, is the present output of the filter, k is a constant that selects the cutoff frequency of the filter. The first transmission zero is at 1/(2k+ 1) of the sampling frequency. The data were smoothed by three sequential filtering operations with k = 5 (reducing the sampling rate by a factor of 5 after each of the first two filtering operations), achieving a final signal bandwidth of about 11 Hz. Components under a few Hz can be assumed not to have been altered to any appreciable extent. The signal resulting from these operations shows a reasonably consistent temporal pattern in successive breaths, for each channel. Figure 2 shows the signals corresponding to the electrodes in Fig. 1 for one of the experimental conditions considered and Fig. 3 shows the rectified and low-pass filtered EMG signal for the inspiratory part of the same breath in Fig. 2. SPATIAL DATA DISTRIBUTION
Given the data measured at discrete sample locations and rectified and low-pass filtered according to the criteria discussed above, additional values for intermediate locations between actual points of observations can, in principle, be estimated. Then a two-dimensional display of the spatial charactersitics of the EMG activity on the chest could be achieved. The chest surface has been treated as a rectangular surface in which the data points are equally spaced. The rectangular surface approximation simplified corn-
SPATIO-TEMPORAL
iiil~Jllilllltll
INTERCOSTAL
129
ACTIVITY
I IIIII liliLlltili,lilliilJiililiLllllILii,lJLllII1
i, tl]L111111 ',tliil.IJIJ.Llllllli I l _l111111~lT.lll,iiLilli]llLiili,l
llidi, Wlllll I1 I1 II II li r,"i, I11l Jiili I...]!IN.N.!! 7ltllllv r-rqr ,rlrll~pl, F lll l]l I
I llll
ti[111
Uii~ tl ~ri~ +.......
~lllJ r[llll
, II.
II L illI. i
1
IIII IItI],~LItI, LItL/I.I~I I1111IIH 1
:-
rpll,r,i I p~l,llrpl
~ '1 'I'I'Pl l'l't'rrr
l tl ILl11 ILllJ]
J
I!l
~L IL, LJ,, iIl~,l llllJ ~ L~WJ,J,J~LIJJik~ILLL L, I JlIIl~ "' "'' ~'~7"'"'~"1 '~'~'~ 1 -
. . . . .
I ~~~"~'p~"~'r"1-
.....
FIG. 2. Intercostal EMG signals during respiration with 9% CO2 in the inspired air. The traces shown refer, from top down, to electrodes 1 to 8 of Fig. 1. Full scale is 3.84 seconds and the markers are 150 samples or 0-048 seconds apart.
p u t e r i m p l e m e n t a t i o n a l t h o u g h it is r e c o g n i z e d that a c o n i c surface w o u l d have given a m o r e realistic r e p r e s e n t a t i o n . D a t a i n t e r p o l a t i o n has b e e n carried out by a F o u r i e r b a n d l i m i t e d i n t e r p o l a t i o n t e c h n i q u e . This m e t h o d c o m p u t e s the f o r w a r d discrete F o u r i e r t r a n s f o r m for the s a m p l e d d a t a s e q u e n c e a n d e x p a n d s the t r a n s f o r m e d signal by inserting zeros, p r e s e r v i n g the s y m m e t r y o f the t r a n s f o r m ; this is e q u i v a l e n t to d e c r e a s i n g the
o
8
.=_
-M
m
•~
;:3
0 q
l-J
•="-' '-n
_"Ln =_
o~"
~.-n
0 0
~"
0
0
0
0
o
~
0
0
)
r
~ I
oi
~
I
0
I~ I
o~
~ I
0
0 I
r
o=
EMG
I
0
0
0
0
0
0
0
o
0
>
oO,~
( x l O z)
I
o~
1
~
0~ I
I
~0 I
i
~
z
0
0
.
o,~
(•-.)'
0
0
0
~
I
~
I
~ I
~ I
P0 ~
~ 0
Q
EMG 0
0
o
$
oO
0
8
0
( x l O 2)
~ I
I
0 I
~
o~ I
I
~
oi I
0 I
I
~
0
SPATIO-TEMPORAL
INTERCOSTAL
ACTIVITY
131
inter-sample spacing. The result is then inverse Fourier transformed, reconstructing the original signal. Fourier band-limited interpolation requires adequate spatial sampling rate (if frequencies greater than the Nyquist frequency are present in the data, the pattern of the amplitudes within the range of frequencies specified by the sampling rate would be irrecoverably altered). The error due to such aliasing can only be reduced by taking more samples. As only eight channels were available, errors of this kind should be regarded as inherent inaccuracies due to the limited number of spatial data available in this exploratory study. (If further work can be justified, an assessment would be made of the spatial sampling rate and other requirements of the method). Fourier band-limited interpolation also requires minimal boundary transients, and it is desirable, in order to use algorithms for the fast evaluation of Fourier transforms (by Chirp-z transform where necessary), to have a data array with even or power-of-two numbers of rows and columns. Some redundant values have been assumed outside the range of the sampled area according to the following criteria. The electrodes were placed to form an array of three equally spaced rows and columns. One extra sample at the corner of the array has been synthetized as the average between the two nearby samples. With reference to Fig. 1, in which the electrodes have been assumed equally spaced, the upper row of the array is formed by electrodes 3, 4, 7, the medium row is formed by electrodes 2, 5, 8, the lower row is formed by electrodes 1, 6; an extra third electrode for this row has been assumed, as specified above. The right side of the thorax has been assumed symmetrical in EMG with respect to the left and the signal has been assumed zero along the sternum and the spine. Each row is then periodic and contains eight points. Further redundant values have been assumed along the parallels to the vertebral column, extending the original array towards the edges of the thoracic cage: the added row values are assumed equal to those of the neighbouring row, scaled down by x/2 and the signal was assumed zero towards the edges of the thoracic cage. The result is a configuration with 6 rows of 8 channels. The original electrode configuration (a) and the periodic electrode array (b), completed by adding redundant values (electrodes 9 to 12) as explained above, are shown below E E
3 2 1
4 5 6 (a)
7 8
.~.
0 12 7 8 9 0
0 11 4 5 6 0
0 10 3 2 1 0
0 0 0 0 0 0
0 10 3 2 1 0
0 11 4 5 6 0
0 12 7 8 9 0
0 0 0 0 0 0
(b) Interpolation on the rows was conducted by a fast Fourier algorithm; on the columns a Chirp-transform programme has been used. The algorithms used are that of Monro (1976) and of Monro & Branch (1977).
132
C. R U G G I E R O
ET AL
The two-dimensional signal obtained can be displayed by means of contour maps, whose succession in time represents the time evolution of the EMG distribution.
5. Results TEMPORAL
VARIABILITY OF SPATIAL EMG DISTRIBUTION
Spatio-temporal patterns throughout the inspiratory phase of one breath are shown in the contours of Fig. 4; the sequence of the first column (from left) illustrates the patterns for a breath generating 4.2% end-expiratory CO2, the sequence of the central column refers to an increased inspired CO2 level generating 5.8% endexpiratory CO2, and the sequence of the third column for the 9% case. The first column sequence can be treated as a control for this series of experiments. Figure 5 shows a later experiment at a different level of anaesthesia (also after gain adjustment of the amplifiers); the sequence of the first column (from left) is the control sequence throughout the inspiratory phase of a breath, the sequence of the central column refers to the effect of tracheal closure starting at the end-expiratory point and maintained throughout the breath, while the sequence of the third column shows the effect of tracheal closure with lifting of the abdomen by means of a bandage sling. The signal characteristics can be discussed in terms of the regional quadrants displayed in the contour plots: Q1 is upper right, and the others are numbered anticlockwise. The 4.2% CO2 records exhibit two approximately equiactive foci, in opposite quadrants (Q1 and Q3). Increasing the CO2 content to 5.8% shifts the balance of activity to the lower left quadrant Q3. In the 9% CO2 case, the activity level rises in all quadrants, with the major activity in Q2. This is compatible with the notion that additional activity in Q2 is added to that maintained in Q3. The Q1 maximum shows relatively little movement during the control breath but shifts towards Q4 (compatible with an additive Q4 source) with increasing CO~_. Tracheal closure provokes an increase in intensity in a highly localized two-maximum configuration; there is no appreciable shift of the locations of the maxima. Thus, during respiration with 4.2% end-expiratory CO2, EMG activity was present mainly in the mid lower parasternal region (around electrodes 1 and 2 in Fig. 1) and in the mid-upper midcostal region (around electrode 7), while the lateral costal region (around electrodes 4, 5, 6) was almost entirely inactive. The signal maxima in the two active regions do not move substantially during one breath and in successive breaths. When CO2 level increases, intercostal activity spreads over the whole thorax. For the 9% CO2 case, the increase in the sternal region (around electrodes 1, 2, 3) and in the mid costal region (around electrodes 7, 8, 9) is greater in the segments in which activity was not previously maximum: then, at the peak of inspiration, the highly localized two-maximum configuration of the 4.2% CO2 case is replaced by much more widespread maximum activity. While the abscissae of the signal maxima in the two active regions do not vary considerably, the ordinates (assuming the axis of the ordinates parallel to the sternum and the spine) show recurrent fluctuations
133
SPATIO-TEMPORAL INTERCOSTAL ACTIVITY I-
7
I"-
"7
f-
_/
L
"7
E
L
i
L
~
r 0
7
i
0
_1
l] -I
F
"q
0 L0
_1
FtG. 4. Contour maps (sternum at the left boundary; spine at the right) at equal time throughout the inspiratory phase of a breath, starting approximately at the beginning 4.2% end-expiratory CO: sequence is shown on the left column (first frame at the sequence relates to a breath with 5.8% end-expiratory CO.,, and the sequence at the breath with 9% end-expiratory CO2). Top of frame is rostral.
intervals (0.32 s) of inspiration. A top). The centre right relates to a
134
C. R U G G I E R O
ET AL
F
-I
F
-I
r-
q
L F
J 7
L r
-J 7
L r
I q
b F
_1 "n
L
J
F
n
L
3 F
n
@ @ @
L F
_1 "9
L
_I
F
L
@
b
q
I
7
1
L
J
~-~ L F
3
L
_1
FIG. 5. As in Fig. 4. The sequence at the left shows the control case (4.5% end-expiratory CO:). The centre sequence relates to a breath with tracheal closure. The sequence at the right relates to a breath with tracheal closure and lifting o f the abdomen. A + s y m b o l indicates a positive peak while - i s used to indicate a through, not a negative maximum.
SPATIO-TEMPORAL
INTERCOSTAL
ACTIVITY
135
which occur roughly at corresponding locations, but subject to considerable variations in amplitude and latency. The intersections of the spatial signal, in each of the two regions, with planes parallel to the zero activity level and lower than the plane of the maximum have been considered. The ordinates of the centres of gravity of the parts above the section planes show a lesser variability within each breath and in successive breaths. The variability decreases when the upper regions considered become more extended, and the trajectories become close to the coherent average pattern. The comparison of individual breath trajectories with the coherent average confirms that the breath by breath variability is not so large as to prevent any systematic pattern being identified. Figure 6 shows the trajectories of the signal maxima throughout several (4) successive breaths, and the trajectories of the centre of gravity of activity above 80% of maximum level, for the activity in quadrants 1 and 4 (posterior region) and quadrants 2 and 3 (anterior). The EMG activity in a single channel is also shown, for reference. Figure 7 shows the coherent average pattern of these trajectories (and single channel EMG) over an ensemble of 28 successive breaths (for several choices of the range of activity). The 5.8% CO, case produces similar effects that are not so marked. In the tracheal closure case, a highly localized two-maximum configuration is maintained, although there is a moderate increase of activity throughout the whole region analysed. When the abdomen is lifted, intercostal activity appears in the lower costal lateral region. The observations can be summarized as follows: different loci of intercostal activity are evident, and they alter characteristically and individually throughout the respiratory cycle, according to similar cycle-by-cycle patterns of intensity and location. There is one apparently basic pattern of activation; it is modified, perhaps additively, by a change of the mechanical and chemical conditions away from the control state. The various changes exhibited are compatible with the concept of the global, spatially integrated control of the intercostal muscles.
EFFECTS OF OTHER
INTERVENTIONS
Experiments were conducted in which a single intercostal segment has been partially or completely denervated. The results are illustrated by the selective display in Fig. 8. The position of the denervated segment (segment 6) is in the middle of the frame. The contours are shown at the peak of inspiration in each case, to exhibit the important effect: a decrease of activity in the lower intercostal segments, a moderate increase in the upper intercostal region. In the other intervention, the thorax was subjected to a local applied force by a probe located at the mid-point of the 6th rib. The effect is shown, again at the peak inspiratory point, in Fig. 9 for a control breath, and for the intervention. The position of the probe is in the centre of the frame. There is little change in activity near to the probe; but an appreciable change occurs more distantly; foci already active are intensified, and a new source activated in Q4.
. . . .• .. . . . = ~ . . ~ - ~
120 160 2 0 0 2 4 0
~
280
""
v
120 160 200 240
I
...
280
~
t,. -
o 4 ~ . . - . ~ o2 oo1-, , , , , 8o 12o ,~o 20o 24o 28o
...... ~'..
] 1 l I I 120 160 2 0 0 240 280
06
I
80
i
..o 120 160 200 240 280
I
I
..,.~ 1
0 0 1
[
280
.'1~ "
o.2r -
O.6
120 160 200 240
.'z..t%~.",
120 160 200 2 4 0 280
........
o8~
80
o.8I
80
FIG. 6. The sequence on the upper row shows a single channel E M G (for the 9.0% C O , case) during inspiration in 4 subsequent breaths (inter-sample interval: 0.008 s; arbitrary units on the axis of the ordinates). The sequence on the central row shows the corresponding trajectories of the ordinates of the signal m a x i m a in anterior (solid line) and posterior (dotted line) regions. The range of ordinates represents the "'height" o f the contour plots shown in Fig. 4. The sequence on the lower row shows the correspondent trajectories of the ordinates of the centres of gravity of the activity above 80% of the m a x i m u m for anterior (solid line) and posterior (dotted line) regions appropriately. Range of ordinates: as in the previous sequence. The three sequences refer to corresponding sample number.
Somple number
o 4 ~ _ J ' ....... . . . ~ r . . o2 F v o o , , , , , , 8o 12o 1~o 2oo 24o 280
t
80
o~ r-
oo/
•
80
~ 0o6 ~ . . .~. . . . . ' "% .. . .. ~ ~
t 06
o4~__~r ........ ..~.. o2ir_ "" oo l , , , , , 8o 12o 1~o 200 240 28o
........... • ~_.'...
I I I I I 120 160 2 0 0 240 280
•
280
o6
I
0.01 80
o~ F
.......... " ~ " "
120 160 2 0 0 240
o.4~----~.._" ~
o~.
80
° 61""" . . . . . . . , ~ , , . - - ' - ~ t ~ F
I I 1 I 1 120 160 2 0 0 240 280
~-~.!
0.2 0.01 80
o,.
80
7~ O t'n
I11
C)
(3
O~
SPATIO-TEMPORAL
80
12'0
INTERCOSTAL
160
2 ( ] 0 240
ACTIVITY
137
2'80
0.6 ................
~
.
"',....,.
0 - 0 /
80
I
J
,
i
,
120
160
200
240
280
.
0.6 ................
"%., 0-0/ 80
~
120
,
J
160 200 Sample
J
240
280
FIG. 7. Upper diagram: signal obtained by coherent averaging of a single channel EMG on 28 breaths (inspiratory phase). Centre diagram: correspondent trajectories of the ordinates of the centres of gravity of the activity above 98% of the maximum of the spatio-temporal signal obtained by coherent averaging on 28 breaths. Lower diagram: as in the previous for activity above 92%. Solid line: anterior region; dotted line: posterior region. Abscissae and ordinates: as in Fig. 6.
These observations also are compatible with the concept of a global response to localized interventions affecting the thorax. While in the experiments described in the previous section the electrodes were placed at about the same distances, irrespective of the shape of the ribs, in the experiments described here the electrodes were arranged in three rows along intercostal segment 3, 5, 7. When representing these rows as rows of a rectangular array the position of the electrodes in the posterior region has been lowered with respect to the experimental configuration--a further reason that prevents direct comparison with the experiments of the previous section. 6. Discussion and Conclusions
The approach adopted in this study characterizes intercostal activity by a global control function estimated by EMG spatio-temporal pattern. In all experiments presented the main features have been found to be self-consistent in each experiment,
138
C. R U G G I E R O
ET
AL
7
[--
I
-7
I
t_ FIG. 8. Patterns for a control breath (left frame) and for a breath following partial intercostal segment denervation (upper right frame) or complete denervation (lower right frame). The contours refer only to the peak inspiratory point in each case.
r-
7
L.
_1
7
FIG. 9. Patterns at the peak inspiratory point for a control breath (left frame) and during a local applied force intervention (right frame).
SPAT10-TEMPORAL
INTERCOSTAL
ACTIVITY
139
and it is possible to recognize a distinct pattern of activity for each experimental condition. A more detailed analysis of the estimator on which this approach is based is then warranted. The requirements of this estimator are now examined. Turning to the data acquisition and processing aspects, it is possible both that limitations and uncertainties are imposed by the experimental constraints and also that alternative processing methods may be better. Specifically, the spatial sampling rate may be inadequate due to the limited number of available channels, and the interpolation method may be unsatisfacotry in requiring the addition of supplementary redundant values to permit a Fourier band-limited interpolation. Representing EMG activity in terms of a conceptual continuous signal implies certain sampling rate requirements, specified by the Shannon theorem. In a further experiment 7 electrodes were arranged in a single row and the potentials measured. Then it was presumed that the potentials at four of these were unknown, and their values estimated by band-limited interpolation; a comparison between actual and estimated potential was carried out. Figure 10 shows the signal observed at each of the 7 electrodes (the triangles). If only three electrodes had been used (at 0.25, 0-5, 0.75 of the sternum-spine distance), the interpolation would have generated the intermediate values shown starred. While the broad form of the signal change with position is detected, the detailed description is certainly inadequate. The signal exhibits a greater spatial rate of change than could be accomodated by the original 3-point sampling. Evidently a significantly greater number of electrodes is, not surprisingly, required, and it is hoped that this improvement will be implemented.
t.9 hi
w~ ,A- •
0-0
t
Slernum
I 0-2
J 0.5 Electrode position
I 0.75
I "0
t
Spine
FIG. 10. Interpolation experiment (at the peak of inspiration); triangles are for the EMG signal observed, stars are for the interpolated values if only three electrodes had been used. Arbitrary units for the ordinates. One of the alternatives to Fourier band-limited interpolation is the spline technique. The results of interpolation by this method, using the expanded set of data assembled for the Fourier method, are very closely similar to those of the Fourier method. On the other hand, if it is preferred not to add extra values to the initial 9, the interpolated pattern is affected appreciably by the values chosen for the
140
c. R U G G I E R O E T A L
n e c e s s a r y a l t e r n a t i v e - - t h a t o f the b o u n d a r y values that must be c h o s e n for the s e c o n d d e r i v a t i v e o f the spline f u n c t i o n at the extremes. N e v e r t h e l e s s , d e s p i t e the limitations i m p o s e d by t e c h n i c a l constraints, the results o f all these e x p e r i m e n t s s u p p o r t the basic h y p o t h e s i s - - t h a t a g l o b a l l y - i n t e g r a t e d c o n t r o l o f the intercostal m u s c u l a t u r e o p e r a t e s o v e r the thorax. This suggests that t h e a p p r o a c h t h r o u g h t h e t e c h n i c a l d e v i c e o f a c o n c e p t u a l l y c o n t i n u o u s control f u n c t i o n is useful in c h a r a c t e r i z i n g the system b e h a v i o u r . F u r t h e r studies with i m p r o v e d e l e c t r o d e a r r a n g e m e n t s are p l a n n e d . This work was supported by the MRC and the Italian Research Council.
REFERENCES
CAMPBELL, E. J. M., AGOSTONI, E. & NEWSOM-DAVIS,J. (1970). The respirator), muscles--mechanics and neural control. London: Lloyd-Luke. LYNN, P. A. (1971), Recursive digital filters for biological signals. Med. Biol. Eng. 9, 37. LYNN, P. A. (1975). Frequency sampling filters with integer multipliers. In: Introduction to digitalfihering. (Bogner, R. E. & Constantinides, A. eds London: Wiley. MONRO, D. M. (1977). Real discrete fast Fourier transform algorithm AS 97. AppL Statist. 25, 166. MONRO, D. M. & BRANCH, J. L. (1977). The chirp-discrete Fourier transform of general length-Algorithm AS 117. Appl. Statist. 26, 351. N EWSOM-DAVIS,J. & SEARS,T. A. (1970). The proprioceptive reflex control of the intercostal muscles during their voluntary activation../. Physiol. Lond. 209, 711. DA SILVA, K. M. C., SAYERS, B. McA., SEARS,T. A. & STAGG, D. T. (1977). The changes in configuration of the rib cage and abdomen during breathing in the anaesthetised cat. J. PhysioL (Lond.) 266, 499. TAYLOR, A. (1960). The contribution of the intercostal muscles to the effort of respiration in man. Z Physiol. (Lond.) 151,390.
Spatio-temporal Characterization of Intercostal Activity During Breathing in the Cat C. RUGGIERO,~" B. M c A . SAYERS,~ Z. A. SEARS§ AND P. A. KIRKWOOD§
t Department of Communications, Computer and Systems Science, University of Genoa, Italy; ~. Department of Computing, Imperial College, London, U.K.; and § Sobell Department of Neurophysiology, Institute of Neurology, University of London, U.K. (Received 28 September 1985, and in revised form 7 July 1986) It is envisaged that the motor control of the intercostal musculature--an assembly of mobile structures--can be characterized in terms of a conceptual spatially continuous control function, that underlies the discretely distributed muscular activity and reflects an inferred global dynamic control of the thoracic cage during breathing. The global control function is estimated by the spatio-temporal pattern obtained by averaging in time and space and interpolation of multichannel simultaneous intercostal EMG recording in the anaesthetized cat. Different examples of the experimental preparation in the presence of stimuli of different kinds are analysed. The resultant signal patterns are found to be self-consistent and capable of exhibiting systematically differing features in systematically differing experimental conditions, thus supporting the validity of the analysis and the choice of the estimator. It is concluded that a more detailed analysis of the requirements of this approach is then warranted. Such requirements are discussed, and, specifically, results that bear on the adequacy of spatial sampling rate are presented. It is suggested that such methods offer a promising approach in the study of motor control strategies of the respiratory apparatus.
1. Introduction
A method for characterizing motor activity of the external respiratory apparatus based on spatio-temporai intercostal E M G pattern analysis is proposed, with a view to clarifying motor control strategies involved in the mechanics of breathing. In the thoracic c a g e - - a n assembly of mobile structures--breathing and postural changes take place while structural stability is maintained. The importance of the role of muscular activity associated with respiration in relation to the preservation of mechanical stability was recognized in previous work (da Silva et al., 1977), in which the kinematics of changes in configuration of the structures that participate in respiratory movements was explored. Specifically, it was suggested that spatially integrated nervous control of the intercostal muscles is essential for satisfactory performance by the external respiratory apparatus. The characterization proposed here envisages that the activity of the intercostal muscles can be interpreted in terms of a conceptual spatially continuous control function underlying the discretely distributed muscular activity. The objective of 125 0022-5193/87/060125+ 16 $03.00/0
© 1987 Academic Press Inc. (London) Ltd
126
c. RUGGIERO
ET AL
this study is to identify an estimator of the continous underlying control function, whose features can represent the global muscular effects acting on the thorax, thereby clarifying neurophysiological characteristics of breathing movements. The estimator of the underlying control function is, in this work, based on the activity of the intercostal muscles; it is formed from discrete observation of EMG, processed to produce a complete spatio-temporal pattern throughout the inspiratory phase of the breath. The problem considered here has two aspects: deriving a suitable estimator from muscular activity; then assessing whether the estimation is self-consistent and capable of exhibiting systematically differing features in systematically differing experimental conditions. If this is the case, a more detailed analysis of the requirements of this approach would be warranted. 2. A Spatio-temporal Estimator of Intercostal Activity It is envisaged that the spatially discrete intercostal muscular activity can be interpreted in terms of a conceptual underlying spatially continuous control function; the underlying function is to be estimated from multichannel simultaneous EMG recording. The electrode system needs to suppress ECG interference, to achieve a spatiallylimited field of electrical sensitivity, and to produce EMG activity that is nevertheless an adequate local "sample" and so reasonably representative of local group of muscle units. Accordingly, bipolar EMG recording was used. It was proposed to rectify and filter the EMG in each channel, smoothing sufficiently to produce a signal fluctuating only at a rate appropriate to the envisaged control action, then to estimate a continuous function by suitable spatial interpolation at each instant studied. The resulting spatial signal would then be regarded as characterizing the global muscular effects acting on the thorax. The patterns of global activity could be portrayed by means of contour maps representing the lines of intersection of the signal with equally spaced planes parallel to zero-level surface. The succession of contour patterns in time would then present a global generalization of the time evolution of the EMG distribution. These patterns are proposed as an estimator of the underlying global spatiotemporal activity that reflects an inferred dynamic control of the thorax. The influence of different experimental conditions due to stimuli of different kinds is then to be examined to assess whether this estimator varies in a distinctive and systematic fashion. 3. Experimental Methods Electromyographic recording was made from two cats anaesthetized with nembutal (initial dose 40 mg/kg) and maintained in a reasonably constant state of light anaesthesia during the experiments. The bipolar electrodes consist of pairs of fine wires; they were inserted using a serum needle which was then withdrawn, leaving the electrodes held in the muscle by their hooked ends. Further details of this technique are given by Newsom-Davis & Sears (1970) and Taylor (1960).
SPATIO-TEMPORAL
INTERCOSTAL
ACTIVITY
127
It was decided to record simultaneous E M G signals from eight points on the left part of the chest, creating thereby a fairly regular array covering the inspiratory part of half thorax (no electrodes were placed in the lowest intercostal region, not active in inspiration). Figure 1 shows the positions of the electrodes. The number of channels was limited by the available means of recording. Prior to recording, the signals were amplified in such a way as to achieve about the same peak to peak excursion of the most distinctive motor units in all channels. As a consequence, the spatio-temporal patterns obtained are only comparable within the same experiments; normalizing the data in a different way would be necessary to permit comparison of sets o f data from different experiments.
FIG. 1. Positions of the electrodes on the thorax.
The E M G signal was recorded on a PI 6200 FM tape recorder and digitized at 11 bit precision in an IBM-1800 computer. The analog to digital conversion set-up available allows for a maximum sampling rate of 20 kHz, which, when sampling eight channels simultaneously, corresponds to 2.5 kHz sampling rate per channel. Each signal contained no significant contributions above 1.25 kHz. Prior to analog to digital conversion the analog tape was replayed at 8/10 of recording tape speed and copied on a new tape, to achieve an equivalent bandwidth of 1 kHz, and the sampling rate used was 2.5 kHz, slightly greater than the 2 kHz rate required to avoid aliasing for 1 kHz bandwidth signal (as advisable to achieve good amplitude definition). Referring to the time scale of the signal, the adopted sampling rate corresponds to 3125 Hz/channel. Further analysis was carried out on an IBM 1800 and on a CDC 6500 or, equivalently, on a CII 10070. The experiments reported here refer to the following experimental conditions. For the first animal: different CO2 content in the inspired air; tracheal closure. For the second animal: denervation of one intercostal segment; localized mechanical stimulation of the thorax. Further details are specified later.
128
C. R U G G I E R O
ET
AL
4. Spatio-temporal Pattern Analysis Intercostal EMG activity is characterized by variability in time and in space. Space-temporal aspects of the respiratory EMG patterns were examined as follows. TEMPORAL
VARIABILITY
EMG signal variation in time has been analysed by (software) rectification and low-pass filtering of the signal, aiming to clarify a hypothesized underlying consistent firing pattern for each channel, minimizing the random fluctuations of the data. The natural frequency of the human respiratory system has been found to be about 5 Hz (Campbell et al., 1970). This figure offered an approximate basis for choosing the degree of data smoothing. The rectified signals were low-pass filtered to about twice this figure. The rectified EMG signal was filtered in the time domain using an autoregressive, zero phase shift, integer coefficient digital filter. Autoregressive filtering offers advantages in computational economy and speed; integer coefficient implementation eliminates problems of filter stability and accuracy caused by finite computer word length. A description and practical realization of the filter used is given by Lynn (1971, 1975). Its transfer function is (1--z2k+l)3/z3k(1--Z) 3 and its autoregressive realization is
y(n)=y(n-3)-3y(n-2)+3y(n-
1)-x(n-3k-3)+3x(n-k-2)-3x(n-k-
1)
+x(n+3k) where x, is the present input of the filter, y, is the present output of the filter, k is a constant that selects the cutoff frequency of the filter. The first transmission zero is at 1/(2k+ 1) of the sampling frequency. The data were smoothed by three sequential filtering operations with k = 5 (reducing the sampling rate by a factor of 5 after each of the first two filtering operations), achieving a final signal bandwidth of about 11 Hz. Components under a few Hz can be assumed not to have been altered to any appreciable extent. The signal resulting from these operations shows a reasonably consistent temporal pattern in successive breaths, for each channel. Figure 2 shows the signals corresponding to the electrodes in Fig. 1 for one of the experimental conditions considered and Fig. 3 shows the rectified and low-pass filtered EMG signal for the inspiratory part of the same breath in Fig. 2. SPATIAL DATA DISTRIBUTION
Given the data measured at discrete sample locations and rectified and low-pass filtered according to the criteria discussed above, additional values for intermediate locations between actual points of observations can, in principle, be estimated. Then a two-dimensional display of the spatial charactersitics of the EMG activity on the chest could be achieved. The chest surface has been treated as a rectangular surface in which the data points are equally spaced. The rectangular surface approximation simplified corn-
SPATIO-TEMPORAL
iiil~Jllilllltll
INTERCOSTAL
129
ACTIVITY
I IIIII liliLlltili,lilliilJiililiLllllILii,lJLllII1
i, tl]L111111 ',tliil.IJIJ.Llllllli I l _l111111~lT.lll,iiLilli]llLiili,l
llidi, Wlllll I1 I1 II II li r,"i, I11l Jiili I...]!IN.N.!! 7ltllllv r-rqr ,rlrll~pl, F lll l]l I
I llll
ti[111
Uii~ tl ~ri~ +.......
~lllJ r[llll
, II.
II L illI. i
1
IIII IItI],~LItI, LItL/I.I~I I1111IIH 1
:-
rpll,r,i I p~l,llrpl
~ '1 'I'I'Pl l'l't'rrr
l tl ILl11 ILllJ]
J
I!l
~L IL, LJ,, iIl~,l llllJ ~ L~WJ,J,J~LIJJik~ILLL L, I JlIIl~ "' "'' ~'~7"'"'~"1 '~'~'~ 1 -
. . . . .
I ~~~"~'p~"~'r"1-
.....
FIG. 2. Intercostal EMG signals during respiration with 9% CO2 in the inspired air. The traces shown refer, from top down, to electrodes 1 to 8 of Fig. 1. Full scale is 3.84 seconds and the markers are 150 samples or 0-048 seconds apart.
p u t e r i m p l e m e n t a t i o n a l t h o u g h it is r e c o g n i z e d that a c o n i c surface w o u l d have given a m o r e realistic r e p r e s e n t a t i o n . D a t a i n t e r p o l a t i o n has b e e n carried out by a F o u r i e r b a n d l i m i t e d i n t e r p o l a t i o n t e c h n i q u e . This m e t h o d c o m p u t e s the f o r w a r d discrete F o u r i e r t r a n s f o r m for the s a m p l e d d a t a s e q u e n c e a n d e x p a n d s the t r a n s f o r m e d signal by inserting zeros, p r e s e r v i n g the s y m m e t r y o f the t r a n s f o r m ; this is e q u i v a l e n t to d e c r e a s i n g the
o
8
.=_
-M
m
•~
;:3
0 q
l-J
•="-' '-n
_"Ln =_
o~"
~.-n
0 0
~"
0
0
0
0
o
~
0
0
)
r
~ I
oi
~
I
0
I~ I
o~
~ I
0
0 I
r
o=
EMG
I
0
0
0
0
0
0
0
o
0
>
oO,~
( x l O z)
I
o~
1
~
0~ I
I
~0 I
i
~
z
0
0
.
o,~
(•-.)'
0
0
0
~
I
~
I
~ I
~ I
P0 ~
~ 0
Q
EMG 0
0
o
$
oO
0
8
0
( x l O 2)
~ I
I
0 I
~
o~ I
I
~
oi I
0 I
I
~
0
SPATIO-TEMPORAL
INTERCOSTAL
ACTIVITY
131
inter-sample spacing. The result is then inverse Fourier transformed, reconstructing the original signal. Fourier band-limited interpolation requires adequate spatial sampling rate (if frequencies greater than the Nyquist frequency are present in the data, the pattern of the amplitudes within the range of frequencies specified by the sampling rate would be irrecoverably altered). The error due to such aliasing can only be reduced by taking more samples. As only eight channels were available, errors of this kind should be regarded as inherent inaccuracies due to the limited number of spatial data available in this exploratory study. (If further work can be justified, an assessment would be made of the spatial sampling rate and other requirements of the method). Fourier band-limited interpolation also requires minimal boundary transients, and it is desirable, in order to use algorithms for the fast evaluation of Fourier transforms (by Chirp-z transform where necessary), to have a data array with even or power-of-two numbers of rows and columns. Some redundant values have been assumed outside the range of the sampled area according to the following criteria. The electrodes were placed to form an array of three equally spaced rows and columns. One extra sample at the corner of the array has been synthetized as the average between the two nearby samples. With reference to Fig. 1, in which the electrodes have been assumed equally spaced, the upper row of the array is formed by electrodes 3, 4, 7, the medium row is formed by electrodes 2, 5, 8, the lower row is formed by electrodes 1, 6; an extra third electrode for this row has been assumed, as specified above. The right side of the thorax has been assumed symmetrical in EMG with respect to the left and the signal has been assumed zero along the sternum and the spine. Each row is then periodic and contains eight points. Further redundant values have been assumed along the parallels to the vertebral column, extending the original array towards the edges of the thoracic cage: the added row values are assumed equal to those of the neighbouring row, scaled down by x/2 and the signal was assumed zero towards the edges of the thoracic cage. The result is a configuration with 6 rows of 8 channels. The original electrode configuration (a) and the periodic electrode array (b), completed by adding redundant values (electrodes 9 to 12) as explained above, are shown below E E
3 2 1
4 5 6 (a)
7 8
.~.
0 12 7 8 9 0
0 11 4 5 6 0
0 10 3 2 1 0
0 0 0 0 0 0
0 10 3 2 1 0
0 11 4 5 6 0
0 12 7 8 9 0
0 0 0 0 0 0
(b) Interpolation on the rows was conducted by a fast Fourier algorithm; on the columns a Chirp-transform programme has been used. The algorithms used are that of Monro (1976) and of Monro & Branch (1977).
132
C. R U G G I E R O
ET AL
The two-dimensional signal obtained can be displayed by means of contour maps, whose succession in time represents the time evolution of the EMG distribution.
5. Results TEMPORAL
VARIABILITY OF SPATIAL EMG DISTRIBUTION
Spatio-temporal patterns throughout the inspiratory phase of one breath are shown in the contours of Fig. 4; the sequence of the first column (from left) illustrates the patterns for a breath generating 4.2% end-expiratory CO2, the sequence of the central column refers to an increased inspired CO2 level generating 5.8% endexpiratory CO2, and the sequence of the third column for the 9% case. The first column sequence can be treated as a control for this series of experiments. Figure 5 shows a later experiment at a different level of anaesthesia (also after gain adjustment of the amplifiers); the sequence of the first column (from left) is the control sequence throughout the inspiratory phase of a breath, the sequence of the central column refers to the effect of tracheal closure starting at the end-expiratory point and maintained throughout the breath, while the sequence of the third column shows the effect of tracheal closure with lifting of the abdomen by means of a bandage sling. The signal characteristics can be discussed in terms of the regional quadrants displayed in the contour plots: Q1 is upper right, and the others are numbered anticlockwise. The 4.2% CO2 records exhibit two approximately equiactive foci, in opposite quadrants (Q1 and Q3). Increasing the CO2 content to 5.8% shifts the balance of activity to the lower left quadrant Q3. In the 9% CO2 case, the activity level rises in all quadrants, with the major activity in Q2. This is compatible with the notion that additional activity in Q2 is added to that maintained in Q3. The Q1 maximum shows relatively little movement during the control breath but shifts towards Q4 (compatible with an additive Q4 source) with increasing CO~_. Tracheal closure provokes an increase in intensity in a highly localized two-maximum configuration; there is no appreciable shift of the locations of the maxima. Thus, during respiration with 4.2% end-expiratory CO2, EMG activity was present mainly in the mid lower parasternal region (around electrodes 1 and 2 in Fig. 1) and in the mid-upper midcostal region (around electrode 7), while the lateral costal region (around electrodes 4, 5, 6) was almost entirely inactive. The signal maxima in the two active regions do not move substantially during one breath and in successive breaths. When CO2 level increases, intercostal activity spreads over the whole thorax. For the 9% CO2 case, the increase in the sternal region (around electrodes 1, 2, 3) and in the mid costal region (around electrodes 7, 8, 9) is greater in the segments in which activity was not previously maximum: then, at the peak of inspiration, the highly localized two-maximum configuration of the 4.2% CO2 case is replaced by much more widespread maximum activity. While the abscissae of the signal maxima in the two active regions do not vary considerably, the ordinates (assuming the axis of the ordinates parallel to the sternum and the spine) show recurrent fluctuations
133
SPATIO-TEMPORAL INTERCOSTAL ACTIVITY I-
7
I"-
"7
f-
_/
L
"7
E
L
i
L
~
r 0
7
i
0
_1
l] -I
F
"q
0 L0
_1
FtG. 4. Contour maps (sternum at the left boundary; spine at the right) at equal time throughout the inspiratory phase of a breath, starting approximately at the beginning 4.2% end-expiratory CO: sequence is shown on the left column (first frame at the sequence relates to a breath with 5.8% end-expiratory CO.,, and the sequence at the breath with 9% end-expiratory CO2). Top of frame is rostral.
intervals (0.32 s) of inspiration. A top). The centre right relates to a
134
C. R U G G I E R O
ET AL
F
-I
F
-I
r-
q
L F
J 7
L r
-J 7
L r
I q
b F
_1 "n
L
J
F
n
L
3 F
n
@ @ @
L F
_1 "9
L
_I
F
L
@
b
q
I
7
1
L
J
~-~ L F
3
L
_1
FIG. 5. As in Fig. 4. The sequence at the left shows the control case (4.5% end-expiratory CO:). The centre sequence relates to a breath with tracheal closure. The sequence at the right relates to a breath with tracheal closure and lifting o f the abdomen. A + s y m b o l indicates a positive peak while - i s used to indicate a through, not a negative maximum.
SPATIO-TEMPORAL
INTERCOSTAL
ACTIVITY
135
which occur roughly at corresponding locations, but subject to considerable variations in amplitude and latency. The intersections of the spatial signal, in each of the two regions, with planes parallel to the zero activity level and lower than the plane of the maximum have been considered. The ordinates of the centres of gravity of the parts above the section planes show a lesser variability within each breath and in successive breaths. The variability decreases when the upper regions considered become more extended, and the trajectories become close to the coherent average pattern. The comparison of individual breath trajectories with the coherent average confirms that the breath by breath variability is not so large as to prevent any systematic pattern being identified. Figure 6 shows the trajectories of the signal maxima throughout several (4) successive breaths, and the trajectories of the centre of gravity of activity above 80% of maximum level, for the activity in quadrants 1 and 4 (posterior region) and quadrants 2 and 3 (anterior). The EMG activity in a single channel is also shown, for reference. Figure 7 shows the coherent average pattern of these trajectories (and single channel EMG) over an ensemble of 28 successive breaths (for several choices of the range of activity). The 5.8% CO, case produces similar effects that are not so marked. In the tracheal closure case, a highly localized two-maximum configuration is maintained, although there is a moderate increase of activity throughout the whole region analysed. When the abdomen is lifted, intercostal activity appears in the lower costal lateral region. The observations can be summarized as follows: different loci of intercostal activity are evident, and they alter characteristically and individually throughout the respiratory cycle, according to similar cycle-by-cycle patterns of intensity and location. There is one apparently basic pattern of activation; it is modified, perhaps additively, by a change of the mechanical and chemical conditions away from the control state. The various changes exhibited are compatible with the concept of the global, spatially integrated control of the intercostal muscles.
EFFECTS OF OTHER
INTERVENTIONS
Experiments were conducted in which a single intercostal segment has been partially or completely denervated. The results are illustrated by the selective display in Fig. 8. The position of the denervated segment (segment 6) is in the middle of the frame. The contours are shown at the peak of inspiration in each case, to exhibit the important effect: a decrease of activity in the lower intercostal segments, a moderate increase in the upper intercostal region. In the other intervention, the thorax was subjected to a local applied force by a probe located at the mid-point of the 6th rib. The effect is shown, again at the peak inspiratory point, in Fig. 9 for a control breath, and for the intervention. The position of the probe is in the centre of the frame. There is little change in activity near to the probe; but an appreciable change occurs more distantly; foci already active are intensified, and a new source activated in Q4.
. . . .• .. . . . = ~ . . ~ - ~
120 160 2 0 0 2 4 0
~
280
""
v
120 160 200 240
I
...
280
~
t,. -
o 4 ~ . . - . ~ o2 oo1-, , , , , 8o 12o ,~o 20o 24o 28o
...... ~'..
] 1 l I I 120 160 2 0 0 240 280
06
I
80
i
..o 120 160 200 240 280
I
I
..,.~ 1
0 0 1
[
280
.'1~ "
o.2r -
O.6
120 160 200 240
.'z..t%~.",
120 160 200 2 4 0 280
........
o8~
80
o.8I
80
FIG. 6. The sequence on the upper row shows a single channel E M G (for the 9.0% C O , case) during inspiration in 4 subsequent breaths (inter-sample interval: 0.008 s; arbitrary units on the axis of the ordinates). The sequence on the central row shows the corresponding trajectories of the ordinates of the signal m a x i m a in anterior (solid line) and posterior (dotted line) regions. The range of ordinates represents the "'height" o f the contour plots shown in Fig. 4. The sequence on the lower row shows the correspondent trajectories of the ordinates of the centres of gravity of the activity above 80% of the m a x i m u m for anterior (solid line) and posterior (dotted line) regions appropriately. Range of ordinates: as in the previous sequence. The three sequences refer to corresponding sample number.
Somple number
o 4 ~ _ J ' ....... . . . ~ r . . o2 F v o o , , , , , , 8o 12o 1~o 2oo 24o 280
t
80
o~ r-
oo/
•
80
~ 0o6 ~ . . .~. . . . . ' "% .. . .. ~ ~
t 06
o4~__~r ........ ..~.. o2ir_ "" oo l , , , , , 8o 12o 1~o 200 240 28o
........... • ~_.'...
I I I I I 120 160 2 0 0 240 280
•
280
o6
I
0.01 80
o~ F
.......... " ~ " "
120 160 2 0 0 240
o.4~----~.._" ~
o~.
80
° 61""" . . . . . . . , ~ , , . - - ' - ~ t ~ F
I I 1 I 1 120 160 2 0 0 240 280
~-~.!
0.2 0.01 80
o,.
80
7~ O t'n
I11
C)
(3
O~
SPATIO-TEMPORAL
80
12'0
INTERCOSTAL
160
2 ( ] 0 240
ACTIVITY
137
2'80
0.6 ................
~
.
"',....,.
0 - 0 /
80
I
J
,
i
,
120
160
200
240
280
.
0.6 ................
"%., 0-0/ 80
~
120
,
J
160 200 Sample
J
240
280
FIG. 7. Upper diagram: signal obtained by coherent averaging of a single channel EMG on 28 breaths (inspiratory phase). Centre diagram: correspondent trajectories of the ordinates of the centres of gravity of the activity above 98% of the maximum of the spatio-temporal signal obtained by coherent averaging on 28 breaths. Lower diagram: as in the previous for activity above 92%. Solid line: anterior region; dotted line: posterior region. Abscissae and ordinates: as in Fig. 6.
These observations also are compatible with the concept of a global response to localized interventions affecting the thorax. While in the experiments described in the previous section the electrodes were placed at about the same distances, irrespective of the shape of the ribs, in the experiments described here the electrodes were arranged in three rows along intercostal segment 3, 5, 7. When representing these rows as rows of a rectangular array the position of the electrodes in the posterior region has been lowered with respect to the experimental configuration--a further reason that prevents direct comparison with the experiments of the previous section. 6. Discussion and Conclusions
The approach adopted in this study characterizes intercostal activity by a global control function estimated by EMG spatio-temporal pattern. In all experiments presented the main features have been found to be self-consistent in each experiment,
138
C. R U G G I E R O
ET
AL
7
[--
I
-7
I
t_ FIG. 8. Patterns for a control breath (left frame) and for a breath following partial intercostal segment denervation (upper right frame) or complete denervation (lower right frame). The contours refer only to the peak inspiratory point in each case.
r-
7
L.
_1
7
FIG. 9. Patterns at the peak inspiratory point for a control breath (left frame) and during a local applied force intervention (right frame).
SPAT10-TEMPORAL
INTERCOSTAL
ACTIVITY
139
and it is possible to recognize a distinct pattern of activity for each experimental condition. A more detailed analysis of the estimator on which this approach is based is then warranted. The requirements of this estimator are now examined. Turning to the data acquisition and processing aspects, it is possible both that limitations and uncertainties are imposed by the experimental constraints and also that alternative processing methods may be better. Specifically, the spatial sampling rate may be inadequate due to the limited number of available channels, and the interpolation method may be unsatisfacotry in requiring the addition of supplementary redundant values to permit a Fourier band-limited interpolation. Representing EMG activity in terms of a conceptual continuous signal implies certain sampling rate requirements, specified by the Shannon theorem. In a further experiment 7 electrodes were arranged in a single row and the potentials measured. Then it was presumed that the potentials at four of these were unknown, and their values estimated by band-limited interpolation; a comparison between actual and estimated potential was carried out. Figure 10 shows the signal observed at each of the 7 electrodes (the triangles). If only three electrodes had been used (at 0.25, 0-5, 0.75 of the sternum-spine distance), the interpolation would have generated the intermediate values shown starred. While the broad form of the signal change with position is detected, the detailed description is certainly inadequate. The signal exhibits a greater spatial rate of change than could be accomodated by the original 3-point sampling. Evidently a significantly greater number of electrodes is, not surprisingly, required, and it is hoped that this improvement will be implemented.
t.9 hi
w~ ,A- •
0-0
t
Slernum
I 0-2
J 0.5 Electrode position
I 0.75
I "0
t
Spine
FIG. 10. Interpolation experiment (at the peak of inspiration); triangles are for the EMG signal observed, stars are for the interpolated values if only three electrodes had been used. Arbitrary units for the ordinates. One of the alternatives to Fourier band-limited interpolation is the spline technique. The results of interpolation by this method, using the expanded set of data assembled for the Fourier method, are very closely similar to those of the Fourier method. On the other hand, if it is preferred not to add extra values to the initial 9, the interpolated pattern is affected appreciably by the values chosen for the
140
c. R U G G I E R O E T A L
n e c e s s a r y a l t e r n a t i v e - - t h a t o f the b o u n d a r y values that must be c h o s e n for the s e c o n d d e r i v a t i v e o f the spline f u n c t i o n at the extremes. N e v e r t h e l e s s , d e s p i t e the limitations i m p o s e d by t e c h n i c a l constraints, the results o f all these e x p e r i m e n t s s u p p o r t the basic h y p o t h e s i s - - t h a t a g l o b a l l y - i n t e g r a t e d c o n t r o l o f the intercostal m u s c u l a t u r e o p e r a t e s o v e r the thorax. This suggests that t h e a p p r o a c h t h r o u g h t h e t e c h n i c a l d e v i c e o f a c o n c e p t u a l l y c o n t i n u o u s control f u n c t i o n is useful in c h a r a c t e r i z i n g the system b e h a v i o u r . F u r t h e r studies with i m p r o v e d e l e c t r o d e a r r a n g e m e n t s are p l a n n e d . This work was supported by the MRC and the Italian Research Council.
REFERENCES
CAMPBELL, E. J. M., AGOSTONI, E. & NEWSOM-DAVIS,J. (1970). The respirator), muscles--mechanics and neural control. London: Lloyd-Luke. LYNN, P. A. (1971), Recursive digital filters for biological signals. Med. Biol. Eng. 9, 37. LYNN, P. A. (1975). Frequency sampling filters with integer multipliers. In: Introduction to digitalfihering. (Bogner, R. E. & Constantinides, A. eds London: Wiley. MONRO, D. M. (1977). Real discrete fast Fourier transform algorithm AS 97. AppL Statist. 25, 166. MONRO, D. M. & BRANCH, J. L. (1977). The chirp-discrete Fourier transform of general length-Algorithm AS 117. Appl. Statist. 26, 351. N EWSOM-DAVIS,J. & SEARS,T. A. (1970). The proprioceptive reflex control of the intercostal muscles during their voluntary activation../. Physiol. Lond. 209, 711. DA SILVA, K. M. C., SAYERS, B. McA., SEARS,T. A. & STAGG, D. T. (1977). The changes in configuration of the rib cage and abdomen during breathing in the anaesthetised cat. J. PhysioL (Lond.) 266, 499. TAYLOR, A. (1960). The contribution of the intercostal muscles to the effort of respiration in man. Z Physiol. (Lond.) 151,390.