- Email: [email protected]
An EMG index: Its values in health and disease and its changes with fatigue
Journal of Electromyography and Kinesiology Vol. 3, No. 3, pp 174-182 9 1993 Butterworth-tteinemann Ltd 1050-6411/93/030174-09
An EMG Index: Its Values in Health and Disease and its Changes with Fatigue Israel Yaar Neurology Section, VA Medical Center, Providence, RI, U.S.A.
Summary: This study was designed to measure an EMG index which relates to the width of the motor unit innervation zone under the assumptions made in Lindstrom's model. This index was measured in healthy and in diseased muscles, intramuscularly, in fatigue inducing maximal voluntary contractions. The bicipital EMG interference patterns of 188 subjects in five diagnostic groups (32 healthy, 83 neuropathic, 28 myasthenic, 13 myotonic and 32 myopathic) were recorded with coaxial needle electrodes. From them, our index was computed repeatedly, in periods of 5.84 s each, to complete fatigue. The data display showed that our index decreased with fatigue. This decrement was found to be significant at P = 0.00001 when all cases' data was combined. Separately, however, only the normals and the neuropathic groups reached significance at P < 0.05. Measured by our index, the five groups divided into two significantly different clusters: (1) The myopathic/myasthenic cluster with large index values, and (2)the normal/neuropathic/myotonic cluster with smaller ones. Several explanations to these findings are entertained. Also, as significant group differences were found, it is safe to anticipate that this index may prove helpful, complementing other EMG indices, in the diagnosis of neuromuscular disorders. The suggested index and its decrement with fatigue deserve additional study. Key Words: Innervation zone--EMG signal-Fatigue--Neuromuscular disorders.
INTRODUCTION
methods 1. U n d e r certain assumptions, the innervation zone width can be estimated from the E M G signal power spectral m o m e n t s (M,) and the muscle fibre conduction velocity (MFCV). This 'zone width is expressed as standard deviation of the scattering of innervation points '1~ This model assumes that the main cause for the variability of the arrival times of the single fibre action potentials under the recording electrodes is the dispersion of the M U I Z - that the individual muscle fibres conduct at very similar M F C V , and that the variance of conduction along the terminal nerve branches is negligible. In this study, the I Z is inferred from electrodiagnostic measurements using Lindstrom's model with an additional assumption that the E M G
The innervation zone (IZ) is that part of the muscle where neuro-muscular joining takes place. These regions are usually located within short stretches of the muscle bulk 13-15. The width of the innervation zone is an important factor in the configuration of the m o t o r unit action potential ( M U A P ) 4-7,16. The anatomical location, structure and width of the I Z was determined and reported in healthy and in diseased muscles using histological Accepted April 2, 1993. Address correspondence and reprint requests to Israel Yaar, Neurology Section, VA Medical Center, Davis Park (127A), Providence, RI 02908, U.S.A.
i74
INNERVA TION Z O N E A N D FATIGUE mean frequency (MF) is a linear function of MFCV that starts close to the origin. As in fact we measured EMG from several MU at a time, as the MFCV of different muscle fibres is known not to be identical, and as the more MUs are involved in the recording the greater the effect of the terminal branches will be - we actually measured a transmission index of sorts, that only in part relates to the IZ dispersion. Therefore, it will be referred to in this paper as Dispersion Index or DI 1~ IZ differences were previously reported between healthy and diseased muscle 7. The present study was designed to find differences between normals and neuromuscular disorders (NMDs) and also in-between NMDs, using the DI. This study was conducted with intramuscular recordings in maximal voluntary contractions (MVC) and during fatigue. The changes in DI may help explain the changes observed in the EMG signal during fatigue, with special reference to differences between various muscle conditions. MATERIALS AND METHODS Subjects EMG interference patterns (EMGIPs) were recorded from biceps muscles of 188 subjects in five diagnostic groups. There were 106 men and 82 women, 8-80 years old (average age 43 -+ 15 years). Thirty-two of them were healthy subjects, 83 with peripheral polyneuropathy, 28 with myasthenia gravis, 13 with myotonia and 32 with myopathy. Recording The recordings were done according to our previously published technique 21 simultaneously from two locations in the biceps muscle, alongside the muscle fibres, and separated exactly 1 cm, during prolonged MVC. The recording system was made of two standard coaxial needle electrodes (DISA 13L49, platinum surface of 0.02 mm 2) that were mounted one alongside the other, one needle protruding 1 cm ahead of the other. Two differential amplifiers were utilized, one per electrode. A large surface electrode was attached to the skin overlying the biceps muscle and connected to each of the amplifiers ground input; the electrically common sheaths of the two electrodes were connected to the two inverting inputs, while the central wires were connected separately to the non-inverting inputs of these amplifiers. The tip of the needle
175
electrode was inserted into the lower third nontendinous part of the relaxed and extended biceps muscle, and slid upwards to dip the second recording surface into the muscle bulk, while taking care to stay as parallel as possible to the long axis of the muscle and to stay short of its mid section where the innervation zone is expected to be. Then the elbow was positioned to 90~ and maintained that way throughout the study. The subjects were instructed to pull competitively against the examiner for as long as they could and then were encouraged to 'go the extra mile'. Two EMGIPs signals, named s,:(t) and Sy(t), each from a different electrode, were amplified and digitized at 5000 samples per second. The frequency response of the system was set to band-pass 9-2200 Hz (3 dB points), with attenuation of 30 dB octave -1 on both ends. Data Reduction and VariabLe Generation Each of the two EMGIPs time series, sx(t) and sy(t), was segmented into three long overlapping segments, 16 384 data points (3.28 s) long each, and into short overlapping segments 256 data points (0.0512 s) each from within the long ones. The short segments were overlapped 50%, while the long segments were overlapped 60% of their lengths (done this way to conform with previous unrelated analysis). Sy(t) was subtracted from sx(t) for each epoch resulting in the differential signal sa(t). Each set of 256 data points of sa(t) was Hanning windowed and fast Fourier transformed. The emergent coefficients were squared, summated for each frequency, and averaged (to a total of) 381 times, generating the difference-power-spectrum {Gad(f)}. The nominal frequency resolution was 19.53 Hz. Our index was computed along the same lines the innervation zone width was calculated by Lindstroml~
D1
=
K 1"
V/B
The V in the equation above was replaced by the relation below,
V = K3.F,,, where
Fm= MJMo The equivalent bandwidth was computed from the power spectral moments,
B = K4.M~/EG~d(f) where Journal of Electromyography & Kinesiology Vol. 3, No. 3, 1993
176
I. Y A A R M . = K2"Ef"Gdd(f) 13.0
Therefore DI = K.[2f.add(f).EGSd(f)l/[Ea..(f)p
(1)
uJ
12.5
tm
D I = dispersion index (arbitrary units
~q
+I
12.0
of length) V = muscle fibre conduction velocity
11.5 >. L
(MFCV) B = equivalent noise bandwidth
e
11.o
r-
10.5
c-
10.0
Fm = mean power frequency (MF) M, = spectral moment of order n
9.5 -10
K1-K4 & K = constants.
The DI was calculated for each case, using Eqn (1), repeatedly for each 5.84 s period, as the subjects were exerting in MVC all the way to complete fatigue. As stated above, the MFCV was not measured directly, but rather substituted by its supposedly linear correlate. Constant 'K' is the product of several constants, some are known, and 'K3' which was not computed. Therefore, DI will be given in arbitrary units of length. Each period was normalized by the total duration of the MVC as per cent of the time it took to complete fatigue (PCF and also % fatigue). The complete fatigue time was the subjective determination of each subject when he or she could not pull any longer even though, as stated above, he or she was encouraged to continue. These values were displayed in Figure 1, averaged separately for each diagnostic group after the best polynomial fit (up to order five) was computed by SPSS regression procedure 19. DI and its statistics were evaluated in multivariate analysis of variance analyses (SPSS, M A N O V A procedure) and presented in Tables 1 and 2. The strength of our index as a diagnostic tool was evaluated in a Jackknife discriminant analysis (SPSS, discriminant procedure). The classification of each case was performed by a discriminant function that was built on the basis of all-cases' data and grouping information less the data and the grouping information of the subject to be classified - so that each case's data and grouping information could not bias its discriminant analysis classification. RESULTS
Figure 1 shows that DI decreased with % fatigue in each diagnostic group. Table 1 shows the polyJournal of Electromyography & Kinesiology Vol. 3, No. 3, 1993
I
I
I
I
I
I
I
1
I
0
10
20
30
40
50
60
70
80
I
I
I
90 100 110
% Fatigue FIG. 1. Changes in the dispersion index with fatigue, groups. - - r q - - normals; - - I - neuropathics; + myasthenics; myotonics; - - 7 1 - - myopathics.
T A B L E 1. Polynomial-fit ~ to and multivariate analysis of the dispersion index's fatigue changes Group
Normals Neuropathics Myasthenics Myotonics Myopathics Total
No. of cases
Cg
C1
C2
32 83 28 13 32
10.6 10.5 12.0 11.0 12.0
- 0 . 0 2 5 0.018 -0.006 --0.007 --0.011 --0.005 --
0.006 0.01 NS NS NS
188
11.1
-0.02
0.00001
0.01
Fatigue changes # m=
@ I = Co + C1.PCF + C2.PCF 2 § Ca'PCFa + + DI = Dispersion index (arbitrary units) PCF = % f a t i g u e - s e e Figure 1. Co = First coefficient = innervation-zone wi dt h e x t r a p o l a t e d to before MVC. * = All Co values are greater than 0 at P < 0.00001. C1-C5 = Coefficients of the p o l y n o m i a l fit; Ca-Cs w e r e never significantly < > 0. # = Significance of the a p p r o p r i a t e slopes in Figure 1. The null hypothesis is slope = 0.0. NS = Not significant: P > 0.05.
nomial-fit coefficients, best suited for each group, and shows that the overall groups fatigue effect on DI was decremental at a significant level of P = 0.00001. However, calculated separately for each group, the fatigue trend reached significance levels of P < 0.006 and P < 0.01 only for the normal and the neuropathic groups respectively (Table 1). There was a significant group effect (P = 0.0001) in the all-groups M A N O V A analysis,
INNERVATION ZONE AND FATIGUE Pairwise repeated-measures MANOVA comparisons for coefficients Co to C3; P values table
TABLE 2.
Group
Normals
Neuropathics n=
Neuropathics
Myasthenics
Myotonics
40 0.03 59 0.9
n = 45 P = 0.15
115
P = 0.6 Myasthenics
Myotonics Myopathics
n = 59 P = 0.0001 n = 45 P = 0.2 n = 64 P=0.005
n = 110 P = 0.0001 n = 96 P = 0.3 n = 115 P=0.0001
n = P = n = P=
suggesting that some or all the groups were significantly different in group-average DI values. This finding made it permissible to perform separate, repeated pairwise comparisons of the groups for differences in DI values. Figure 1 and Table 2 show that the normal, the neuropathic and the myotonic groups kept significantly smaller index average values than the myopathic and the myasthenic ones throughout fatigue. Analysis of the DI values of the different groups at no fatigue (Figure 1, % fatigue = 0) could not separate the myotonic group from the others. Practically, as judged by the DI, the data reflect two distinct clusters: (1) the myopathic/myasthenic group with large index values, and (2) the normal/ neuropathic/myotonic group with smaller, assumed normal index values. The MANOVA procedure showed these two clusters to be different at P = 0.0001 significance level. Also, Jackknife discriminant analysis showed that if the question to answer is whether a subject belongs to one of the two clusters above, using our index alone, a 71% correct classification can be reached. The MANOVA procedure showed no significant effect of sex or age, and there was no significant differential effect of fatigue on the various groups' D I - they all showed decrement in the DI average values, and no group's decrement was (statistically) significantly different than any other's (a non-significant group* fatigue interaction factor). DISCUSSION
Correlation Between Findings and Known Pathology This study emphasizes relations and changes - the relative DI size in the various NMDs and healthy
177
groups, and its changes with fatigue. Intentionally our index values are computed in arbitrary units of length, in order to preclude exact computations of MFCV and several constants. It is interesting that the results seem to aggregate the five groups into two clinically viable and sound clusters: the normal/ neuropathic/myotonic group with 'normal' index values, and the myopathic/myasthenic group with significantly larger index values. These findings support the notion that the DI as defined in this paper is related to the IZ of the muscle as suggested by Lindstrom a~ and as follows below. When neuropathic denervation/reinnervation processes are involved it is known7,8 that the reinnervating sprout has preference to grow back into the old neuromuscular junction (NMJ). These processes are likely to increase the size of the MU but are unlikely to increase the size of the I Z - o v e r l a p p i n g and close-by MUs are known to have similar sizes and locations for their IZs 6. This agrees well with the above clustering of the normals with the neuropathics. However, when there is random damage to muscle fibres they are expected to be reinnervated far away from their original NMJ, resulting in the unavoidable expansion of the IZ width. This process is common in myopathies7,9 and to a lesser extent affects myasthenic muscles too 8. Another possible reason for the large IZ of myasthenic muscle is the complex and ongoing modulation of terminal innervating branches and the NMJs themselves8. This agrees well with the above clustering of myasthenics with myopathics by the DI.
DI Decrements with Fatigue One of the reasons that this study was originated was to measure synchronization of MUs during fatigue. The reasoning was that for a 'function detector' like the DI several synchronized MUs should register as one rather larger MU, and one expects that larger MUs will have wider IZs. Synchronization of MUs increases in fatigue T M , therefore the latter was expected to translate into the detection of increasing index values with fatigue, assuming of course that the DI reflects the IZ size. Obviously the inverse resulted - DI decreased significantly with fatigue. At this time and with this data we have no definitive explanation for this phenomenon. It is possible that during fatigue, mainly in MVC induced fatigue (that in some muscles was accompanied by ischaemia), there Journal of Electromyography & Kinesiology VoL 3, No. 3, 1993
178
L YAAR
might have been a change in the muscle volume conductor resulting in the E M G electrodes failure to record from far-away muscle fibres, causing (functional) restriction of the innervation zone widths. Consequently, relatively more muscle fibres will be recorded closer to the needle electrodes and near-by to one another. As close-by muscle fibres are expected to have closely positioned NMJs 6,7 this may explain the decrement in the DI with fatigue. This process, if it exists, may actually be a stronger process than is evident from Figure 1 - as what one sees is the result of this process being attenuated by an opposing effect of synchronization. The key factor, the changes in muscle volume conductor with fatigue and its effect on our index, seems to be an important issue and in need of additional research. Another explanation for our index decrement with fatigue may be evident from Eqn (1) and the assumptions in its conception: There are several publications documenting a linear relationship between MFCV and the E M G mean power frequency (MF) at rest and during fatigue 2,a7,22 supporting the notion that/s in Eqn (1) is a constant. It is assumed that if there is a dependency between MFCV and MF and there is no noise at each recording electrode, i.e. the signal in one electrode is completely conducted to the other one, then, this dependency if graphed should go through the origin of the graph (i.e. when MF = 0 MFCV = 0, and vice versa) as presented in the mathematical relations preceding Eqn (1). This assumption comes naturally as an extension of the extreme condition that if there is no conduction (from the innervation zone and down-stream) there will be no signal, no power spectrum and no MF. However, there might be local conduction of E M G signals that do not extend, or extend only in part from one electrode to the other - this will result in a differential E M G signal that will be dependent only in part on the computed MFCV. It is also quite possible that /s is not constant but rather changes with fatigue either in a linear or, even worse, in a non-linear way. Yet another possibility is that the relation between MFCV does go through the origin, is not linear if all possible values are included, but is linear in that range of values seen in normal muscles, diseased muscles and during fatigue. A relation of this nature may be of the type V = a'F,n + b. This will result in an additional segment to Eqn (1) which may change our findings. In an analysis, now in press TM, we found that muscles with high E M G I P MF cause Journal of Electromyography & Kinesiology Vol. 3, No. 3, 1993
a curvilinear relation between the MF and MFCV. We also found that at the onset of continuous MVC the MFCV and the MF co-vary in a non-linear way. Other investigators found that spectral parameters (e.g. MF) are more sensitive to fatigue and change dissimilarly to MFCV during fatigue 3'12,2~ During fatigue and probably in various muscular disorders drop out of certain MUs or more likely in our paradigm the addition of larger MUs may occur, changing the MF out of proportion to the change, if any, in the MFCV. Also, changes in the MFCV distribution may occur that will change the MF far beyond its effect on the DI. All these non-linear changes indicate that substituting K3"Fm for MFCV to a lesser or greater extent violates the assumptions in the original model. These claims indicate that different value corrections may be needed for each stage of fatigue. Therefore, as Figure 1 was computed without these corrections the slopes may actually represent fatigue changes of the 'constant' K (% fatigue) rather than showing fatigue changes in the innervation zone width itself. All these possibilities are yet to be investigated.
Diagnostic Value The size of the anatomical I Z is a parameter known to differentiate NMDs from one another and from healthy muscles 7. The IZ size is one of the important factors in determining the shape of the M U A P in health and in disease 5'6. It seems logical therefore that the DI, assuming it is somewhat related to the size of the IZ, will serve well as a diagnostic tool. When calculated during fatigue the DI showed some promise in discriminating among those two c l u s t e r s - up to 71% correctly classified subjects. In this respect, as a diagnostic measure, it is not important if the parameter that we were measuring was really a genuine measure of the innervation zone width and its change with fatigue or just an index of muscle activity that changed with fatigue. As it is likely that some of the biceps muscles that were recorded from subjects with diagnosed NMDs were actually normal, the above mentioned discriminating power is quite good. This should encourage future studies using the DI in combinations with other E M G indices. Acknowledgements: I thank Mrs. Sheryl Chicoine for her help in typing this manuscript; and also VA Merit Review for their support.
INNERVATION ZONE AND FATIGUE REFERENCES 1. Aquilonius SM, Askmark H, Gillberg PG, Nandedkar S, Olsson Y, Stalberg E: Topographical localization of motor endplates in cryosections of whole human muscles. Muscle & Nerve 7:287-293, 1984. 2. Arendt-Nielsen L, Mills KR: The relationship between mean power frequency of the EMG spectrum and muscle fiber conduction velocity. Electroenceph Clin Neurophysiol 60:130-134, 1985. 3. Broman H, Biloto G, DeLuca CJ: Myoelectric signal conduction velocity and spectral parameters: influence of force and time. J Appl Physiol 58:1428-1437, 1985. 4. Broman H, Lindstrom L: A model describing the power spectrum of myoelectric signals. Part II: Motor unit signal. Technical Report 8:74, Res Lab Med Electr, Goteborg, Sweden, 1974. 5. Buchthal F, Guld C, Rosenfalck P: Innervation zone and propagation velocity in human muscle. Acta Physiol Scand 35:175-190, 1955. 6. Buchthal F, Guld C, Rosenfalck P: Multielectrode study of the territory of a motor unit. Acta Physiol Scand 39:83-104 , 1957. 7. Desmedt JE: Plasticity of motor unit organization studied by coherent electromyography in patients with nerve lesions or with myopathic neuropathic diseases. In: Prog Clin Neurophysiol, Vol 9, Ed by Desmedt JE, Karger, Basel, pp 250-304, 1981.. 8. Engel AG, Banker BQ, eds: Myology basic and clinical, McGraw-Hill, NY, pp 245, 380-384, 1986. 9. Hilton-Brown P, Stalberg E: The motor unit in muscular dystrophy, a single fibre EMG and scanning EMG study. J Neurol Neurosurg Psychiatry 46:981-995, 1983. 10. Lindstrom L: Spectral analysis of EMG. In: Electromyog and Evoked Potentials. Theories and Applications, Ed by Struppler A, Weindle A, Springer-Verlag, NY, pp 103-107, 1985. 11. Lloyd AJ: Surface electromyography during sustained isometric contractions. J Appl Physiol 30:713-719, 1971. 12. Merletti R, Knaflitz M, DeLuca CJ: Myoelectric manifes-
13.
14. 15. 16.
17.
18.
19. 20. 21.
22.
179
tations of fatigue in voluntary and electrically elicited contractions. J Appl Physiol 69:1810-1820, 1990. Masuda T, Miyano H, Sadoyama T: The distribution of myoneural junctions in the biceps brachii investigated by surface electromyography. Electroenceph and Clin Neurophysiol 56:597-603, 1983. Masuda T, Sadoyama T: Processing of myoelectric signals for estimating the location of innervation zones in the skeletal muscles. Frontiers Med Biol Engng 1:299-314, 1989. Masuda T, Sadoyama T: Distribution of innervation zones in the human biceps brachii. J Electrornyogr Kinesiol 1:107-115, 1991. Roy SH, De Luca CJ: Evolving characteristics of the median frequency of the EMG signal. In: Computer-aided electromyography and expert systems, Clinical neurophysiology update, Vol 2, Ed by Desmedt JE, Elsevier, Amsterdam, pp 205-221, 1989. Sadoyama T, Masuda T, Miyano H: Relationship between muscle fiber conduction velocity and frequency parameters of surface EMG during sustained contraction. Eur J Appl Physiol 51:247-256, 1983. Shochina M, Gonen B, Wolf E, Mahler Y, Magora A: A study of the application of the Fourier series in the analysis of the electrophysiological pattern of development of muscle fatigue. Electromyogr Clin Neurophysiol 26:513-520, 1986. SPSS/PC+: SPSS Inc., Chicago, IL, 1988. Yaar I, Niles L: Muscle fiber conduction velocity and mean power spectrum frequency in neuro-muscular disorders and in fatigue. Muscle & Nerve (in press). Yaar I, Shapiro MB, Mitz AR, Pottala EW: A new technique for measuring muscle fiber conduction velocities in full interference patterns. Electroenceph Clin Neurophysiot 57:427--434, 1984. Zwarts MJ, Van Weerden TW, Haenen HT: Relationship between average muscle fiber conduction velocity and EMG power spectra during isometric contraction, recovery and applied ischemia. Eur J Appl Physiol 56:212-216, 1987.
Journal of Electromyography & Kinesiology Vol. 3, No. 3, 1993
An EMG Index: Its Values in Health and Disease and its Changes with Fatigue Israel Yaar Neurology Section, VA Medical Center, Providence, RI, U.S.A.
Summary: This study was designed to measure an EMG index which relates to the width of the motor unit innervation zone under the assumptions made in Lindstrom's model. This index was measured in healthy and in diseased muscles, intramuscularly, in fatigue inducing maximal voluntary contractions. The bicipital EMG interference patterns of 188 subjects in five diagnostic groups (32 healthy, 83 neuropathic, 28 myasthenic, 13 myotonic and 32 myopathic) were recorded with coaxial needle electrodes. From them, our index was computed repeatedly, in periods of 5.84 s each, to complete fatigue. The data display showed that our index decreased with fatigue. This decrement was found to be significant at P = 0.00001 when all cases' data was combined. Separately, however, only the normals and the neuropathic groups reached significance at P < 0.05. Measured by our index, the five groups divided into two significantly different clusters: (1) The myopathic/myasthenic cluster with large index values, and (2)the normal/neuropathic/myotonic cluster with smaller ones. Several explanations to these findings are entertained. Also, as significant group differences were found, it is safe to anticipate that this index may prove helpful, complementing other EMG indices, in the diagnosis of neuromuscular disorders. The suggested index and its decrement with fatigue deserve additional study. Key Words: Innervation zone--EMG signal-Fatigue--Neuromuscular disorders.
INTRODUCTION
methods 1. U n d e r certain assumptions, the innervation zone width can be estimated from the E M G signal power spectral m o m e n t s (M,) and the muscle fibre conduction velocity (MFCV). This 'zone width is expressed as standard deviation of the scattering of innervation points '1~ This model assumes that the main cause for the variability of the arrival times of the single fibre action potentials under the recording electrodes is the dispersion of the M U I Z - that the individual muscle fibres conduct at very similar M F C V , and that the variance of conduction along the terminal nerve branches is negligible. In this study, the I Z is inferred from electrodiagnostic measurements using Lindstrom's model with an additional assumption that the E M G
The innervation zone (IZ) is that part of the muscle where neuro-muscular joining takes place. These regions are usually located within short stretches of the muscle bulk 13-15. The width of the innervation zone is an important factor in the configuration of the m o t o r unit action potential ( M U A P ) 4-7,16. The anatomical location, structure and width of the I Z was determined and reported in healthy and in diseased muscles using histological Accepted April 2, 1993. Address correspondence and reprint requests to Israel Yaar, Neurology Section, VA Medical Center, Davis Park (127A), Providence, RI 02908, U.S.A.
i74
INNERVA TION Z O N E A N D FATIGUE mean frequency (MF) is a linear function of MFCV that starts close to the origin. As in fact we measured EMG from several MU at a time, as the MFCV of different muscle fibres is known not to be identical, and as the more MUs are involved in the recording the greater the effect of the terminal branches will be - we actually measured a transmission index of sorts, that only in part relates to the IZ dispersion. Therefore, it will be referred to in this paper as Dispersion Index or DI 1~ IZ differences were previously reported between healthy and diseased muscle 7. The present study was designed to find differences between normals and neuromuscular disorders (NMDs) and also in-between NMDs, using the DI. This study was conducted with intramuscular recordings in maximal voluntary contractions (MVC) and during fatigue. The changes in DI may help explain the changes observed in the EMG signal during fatigue, with special reference to differences between various muscle conditions. MATERIALS AND METHODS Subjects EMG interference patterns (EMGIPs) were recorded from biceps muscles of 188 subjects in five diagnostic groups. There were 106 men and 82 women, 8-80 years old (average age 43 -+ 15 years). Thirty-two of them were healthy subjects, 83 with peripheral polyneuropathy, 28 with myasthenia gravis, 13 with myotonia and 32 with myopathy. Recording The recordings were done according to our previously published technique 21 simultaneously from two locations in the biceps muscle, alongside the muscle fibres, and separated exactly 1 cm, during prolonged MVC. The recording system was made of two standard coaxial needle electrodes (DISA 13L49, platinum surface of 0.02 mm 2) that were mounted one alongside the other, one needle protruding 1 cm ahead of the other. Two differential amplifiers were utilized, one per electrode. A large surface electrode was attached to the skin overlying the biceps muscle and connected to each of the amplifiers ground input; the electrically common sheaths of the two electrodes were connected to the two inverting inputs, while the central wires were connected separately to the non-inverting inputs of these amplifiers. The tip of the needle
175
electrode was inserted into the lower third nontendinous part of the relaxed and extended biceps muscle, and slid upwards to dip the second recording surface into the muscle bulk, while taking care to stay as parallel as possible to the long axis of the muscle and to stay short of its mid section where the innervation zone is expected to be. Then the elbow was positioned to 90~ and maintained that way throughout the study. The subjects were instructed to pull competitively against the examiner for as long as they could and then were encouraged to 'go the extra mile'. Two EMGIPs signals, named s,:(t) and Sy(t), each from a different electrode, were amplified and digitized at 5000 samples per second. The frequency response of the system was set to band-pass 9-2200 Hz (3 dB points), with attenuation of 30 dB octave -1 on both ends. Data Reduction and VariabLe Generation Each of the two EMGIPs time series, sx(t) and sy(t), was segmented into three long overlapping segments, 16 384 data points (3.28 s) long each, and into short overlapping segments 256 data points (0.0512 s) each from within the long ones. The short segments were overlapped 50%, while the long segments were overlapped 60% of their lengths (done this way to conform with previous unrelated analysis). Sy(t) was subtracted from sx(t) for each epoch resulting in the differential signal sa(t). Each set of 256 data points of sa(t) was Hanning windowed and fast Fourier transformed. The emergent coefficients were squared, summated for each frequency, and averaged (to a total of) 381 times, generating the difference-power-spectrum {Gad(f)}. The nominal frequency resolution was 19.53 Hz. Our index was computed along the same lines the innervation zone width was calculated by Lindstroml~
D1
=
K 1"
V/B
The V in the equation above was replaced by the relation below,
V = K3.F,,, where
Fm= MJMo The equivalent bandwidth was computed from the power spectral moments,
B = K4.M~/EG~d(f) where Journal of Electromyography & Kinesiology Vol. 3, No. 3, 1993
176
I. Y A A R M . = K2"Ef"Gdd(f) 13.0
Therefore DI = K.[2f.add(f).EGSd(f)l/[Ea..(f)p
(1)
uJ
12.5
tm
D I = dispersion index (arbitrary units
~q
+I
12.0
of length) V = muscle fibre conduction velocity
11.5 >. L
(MFCV) B = equivalent noise bandwidth
e
11.o
r-
10.5
c-
10.0
Fm = mean power frequency (MF) M, = spectral moment of order n
9.5 -10
K1-K4 & K = constants.
The DI was calculated for each case, using Eqn (1), repeatedly for each 5.84 s period, as the subjects were exerting in MVC all the way to complete fatigue. As stated above, the MFCV was not measured directly, but rather substituted by its supposedly linear correlate. Constant 'K' is the product of several constants, some are known, and 'K3' which was not computed. Therefore, DI will be given in arbitrary units of length. Each period was normalized by the total duration of the MVC as per cent of the time it took to complete fatigue (PCF and also % fatigue). The complete fatigue time was the subjective determination of each subject when he or she could not pull any longer even though, as stated above, he or she was encouraged to continue. These values were displayed in Figure 1, averaged separately for each diagnostic group after the best polynomial fit (up to order five) was computed by SPSS regression procedure 19. DI and its statistics were evaluated in multivariate analysis of variance analyses (SPSS, M A N O V A procedure) and presented in Tables 1 and 2. The strength of our index as a diagnostic tool was evaluated in a Jackknife discriminant analysis (SPSS, discriminant procedure). The classification of each case was performed by a discriminant function that was built on the basis of all-cases' data and grouping information less the data and the grouping information of the subject to be classified - so that each case's data and grouping information could not bias its discriminant analysis classification. RESULTS
Figure 1 shows that DI decreased with % fatigue in each diagnostic group. Table 1 shows the polyJournal of Electromyography & Kinesiology Vol. 3, No. 3, 1993
I
I
I
I
I
I
I
1
I
0
10
20
30
40
50
60
70
80
I
I
I
90 100 110
% Fatigue FIG. 1. Changes in the dispersion index with fatigue, groups. - - r q - - normals; - - I - neuropathics; + myasthenics; myotonics; - - 7 1 - - myopathics.
T A B L E 1. Polynomial-fit ~ to and multivariate analysis of the dispersion index's fatigue changes Group
Normals Neuropathics Myasthenics Myotonics Myopathics Total
No. of cases
Cg
C1
C2
32 83 28 13 32
10.6 10.5 12.0 11.0 12.0
- 0 . 0 2 5 0.018 -0.006 --0.007 --0.011 --0.005 --
0.006 0.01 NS NS NS
188
11.1
-0.02
0.00001
0.01
Fatigue changes # m=
@ I = Co + C1.PCF + C2.PCF 2 § Ca'PCFa + + DI = Dispersion index (arbitrary units) PCF = % f a t i g u e - s e e Figure 1. Co = First coefficient = innervation-zone wi dt h e x t r a p o l a t e d to before MVC. * = All Co values are greater than 0 at P < 0.00001. C1-C5 = Coefficients of the p o l y n o m i a l fit; Ca-Cs w e r e never significantly < > 0. # = Significance of the a p p r o p r i a t e slopes in Figure 1. The null hypothesis is slope = 0.0. NS = Not significant: P > 0.05.
nomial-fit coefficients, best suited for each group, and shows that the overall groups fatigue effect on DI was decremental at a significant level of P = 0.00001. However, calculated separately for each group, the fatigue trend reached significance levels of P < 0.006 and P < 0.01 only for the normal and the neuropathic groups respectively (Table 1). There was a significant group effect (P = 0.0001) in the all-groups M A N O V A analysis,
INNERVATION ZONE AND FATIGUE Pairwise repeated-measures MANOVA comparisons for coefficients Co to C3; P values table
TABLE 2.
Group
Normals
Neuropathics n=
Neuropathics
Myasthenics
Myotonics
40 0.03 59 0.9
n = 45 P = 0.15
115
P = 0.6 Myasthenics
Myotonics Myopathics
n = 59 P = 0.0001 n = 45 P = 0.2 n = 64 P=0.005
n = 110 P = 0.0001 n = 96 P = 0.3 n = 115 P=0.0001
n = P = n = P=
suggesting that some or all the groups were significantly different in group-average DI values. This finding made it permissible to perform separate, repeated pairwise comparisons of the groups for differences in DI values. Figure 1 and Table 2 show that the normal, the neuropathic and the myotonic groups kept significantly smaller index average values than the myopathic and the myasthenic ones throughout fatigue. Analysis of the DI values of the different groups at no fatigue (Figure 1, % fatigue = 0) could not separate the myotonic group from the others. Practically, as judged by the DI, the data reflect two distinct clusters: (1) the myopathic/myasthenic group with large index values, and (2) the normal/ neuropathic/myotonic group with smaller, assumed normal index values. The MANOVA procedure showed these two clusters to be different at P = 0.0001 significance level. Also, Jackknife discriminant analysis showed that if the question to answer is whether a subject belongs to one of the two clusters above, using our index alone, a 71% correct classification can be reached. The MANOVA procedure showed no significant effect of sex or age, and there was no significant differential effect of fatigue on the various groups' D I - they all showed decrement in the DI average values, and no group's decrement was (statistically) significantly different than any other's (a non-significant group* fatigue interaction factor). DISCUSSION
Correlation Between Findings and Known Pathology This study emphasizes relations and changes - the relative DI size in the various NMDs and healthy
177
groups, and its changes with fatigue. Intentionally our index values are computed in arbitrary units of length, in order to preclude exact computations of MFCV and several constants. It is interesting that the results seem to aggregate the five groups into two clinically viable and sound clusters: the normal/ neuropathic/myotonic group with 'normal' index values, and the myopathic/myasthenic group with significantly larger index values. These findings support the notion that the DI as defined in this paper is related to the IZ of the muscle as suggested by Lindstrom a~ and as follows below. When neuropathic denervation/reinnervation processes are involved it is known7,8 that the reinnervating sprout has preference to grow back into the old neuromuscular junction (NMJ). These processes are likely to increase the size of the MU but are unlikely to increase the size of the I Z - o v e r l a p p i n g and close-by MUs are known to have similar sizes and locations for their IZs 6. This agrees well with the above clustering of the normals with the neuropathics. However, when there is random damage to muscle fibres they are expected to be reinnervated far away from their original NMJ, resulting in the unavoidable expansion of the IZ width. This process is common in myopathies7,9 and to a lesser extent affects myasthenic muscles too 8. Another possible reason for the large IZ of myasthenic muscle is the complex and ongoing modulation of terminal innervating branches and the NMJs themselves8. This agrees well with the above clustering of myasthenics with myopathics by the DI.
DI Decrements with Fatigue One of the reasons that this study was originated was to measure synchronization of MUs during fatigue. The reasoning was that for a 'function detector' like the DI several synchronized MUs should register as one rather larger MU, and one expects that larger MUs will have wider IZs. Synchronization of MUs increases in fatigue T M , therefore the latter was expected to translate into the detection of increasing index values with fatigue, assuming of course that the DI reflects the IZ size. Obviously the inverse resulted - DI decreased significantly with fatigue. At this time and with this data we have no definitive explanation for this phenomenon. It is possible that during fatigue, mainly in MVC induced fatigue (that in some muscles was accompanied by ischaemia), there Journal of Electromyography & Kinesiology VoL 3, No. 3, 1993
178
L YAAR
might have been a change in the muscle volume conductor resulting in the E M G electrodes failure to record from far-away muscle fibres, causing (functional) restriction of the innervation zone widths. Consequently, relatively more muscle fibres will be recorded closer to the needle electrodes and near-by to one another. As close-by muscle fibres are expected to have closely positioned NMJs 6,7 this may explain the decrement in the DI with fatigue. This process, if it exists, may actually be a stronger process than is evident from Figure 1 - as what one sees is the result of this process being attenuated by an opposing effect of synchronization. The key factor, the changes in muscle volume conductor with fatigue and its effect on our index, seems to be an important issue and in need of additional research. Another explanation for our index decrement with fatigue may be evident from Eqn (1) and the assumptions in its conception: There are several publications documenting a linear relationship between MFCV and the E M G mean power frequency (MF) at rest and during fatigue 2,a7,22 supporting the notion that/s in Eqn (1) is a constant. It is assumed that if there is a dependency between MFCV and MF and there is no noise at each recording electrode, i.e. the signal in one electrode is completely conducted to the other one, then, this dependency if graphed should go through the origin of the graph (i.e. when MF = 0 MFCV = 0, and vice versa) as presented in the mathematical relations preceding Eqn (1). This assumption comes naturally as an extension of the extreme condition that if there is no conduction (from the innervation zone and down-stream) there will be no signal, no power spectrum and no MF. However, there might be local conduction of E M G signals that do not extend, or extend only in part from one electrode to the other - this will result in a differential E M G signal that will be dependent only in part on the computed MFCV. It is also quite possible that /s is not constant but rather changes with fatigue either in a linear or, even worse, in a non-linear way. Yet another possibility is that the relation between MFCV does go through the origin, is not linear if all possible values are included, but is linear in that range of values seen in normal muscles, diseased muscles and during fatigue. A relation of this nature may be of the type V = a'F,n + b. This will result in an additional segment to Eqn (1) which may change our findings. In an analysis, now in press TM, we found that muscles with high E M G I P MF cause Journal of Electromyography & Kinesiology Vol. 3, No. 3, 1993
a curvilinear relation between the MF and MFCV. We also found that at the onset of continuous MVC the MFCV and the MF co-vary in a non-linear way. Other investigators found that spectral parameters (e.g. MF) are more sensitive to fatigue and change dissimilarly to MFCV during fatigue 3'12,2~ During fatigue and probably in various muscular disorders drop out of certain MUs or more likely in our paradigm the addition of larger MUs may occur, changing the MF out of proportion to the change, if any, in the MFCV. Also, changes in the MFCV distribution may occur that will change the MF far beyond its effect on the DI. All these non-linear changes indicate that substituting K3"Fm for MFCV to a lesser or greater extent violates the assumptions in the original model. These claims indicate that different value corrections may be needed for each stage of fatigue. Therefore, as Figure 1 was computed without these corrections the slopes may actually represent fatigue changes of the 'constant' K (% fatigue) rather than showing fatigue changes in the innervation zone width itself. All these possibilities are yet to be investigated.
Diagnostic Value The size of the anatomical I Z is a parameter known to differentiate NMDs from one another and from healthy muscles 7. The IZ size is one of the important factors in determining the shape of the M U A P in health and in disease 5'6. It seems logical therefore that the DI, assuming it is somewhat related to the size of the IZ, will serve well as a diagnostic tool. When calculated during fatigue the DI showed some promise in discriminating among those two c l u s t e r s - up to 71% correctly classified subjects. In this respect, as a diagnostic measure, it is not important if the parameter that we were measuring was really a genuine measure of the innervation zone width and its change with fatigue or just an index of muscle activity that changed with fatigue. As it is likely that some of the biceps muscles that were recorded from subjects with diagnosed NMDs were actually normal, the above mentioned discriminating power is quite good. This should encourage future studies using the DI in combinations with other E M G indices. Acknowledgements: I thank Mrs. Sheryl Chicoine for her help in typing this manuscript; and also VA Merit Review for their support.
INNERVATION ZONE AND FATIGUE REFERENCES 1. Aquilonius SM, Askmark H, Gillberg PG, Nandedkar S, Olsson Y, Stalberg E: Topographical localization of motor endplates in cryosections of whole human muscles. Muscle & Nerve 7:287-293, 1984. 2. Arendt-Nielsen L, Mills KR: The relationship between mean power frequency of the EMG spectrum and muscle fiber conduction velocity. Electroenceph Clin Neurophysiol 60:130-134, 1985. 3. Broman H, Biloto G, DeLuca CJ: Myoelectric signal conduction velocity and spectral parameters: influence of force and time. J Appl Physiol 58:1428-1437, 1985. 4. Broman H, Lindstrom L: A model describing the power spectrum of myoelectric signals. Part II: Motor unit signal. Technical Report 8:74, Res Lab Med Electr, Goteborg, Sweden, 1974. 5. Buchthal F, Guld C, Rosenfalck P: Innervation zone and propagation velocity in human muscle. Acta Physiol Scand 35:175-190, 1955. 6. Buchthal F, Guld C, Rosenfalck P: Multielectrode study of the territory of a motor unit. Acta Physiol Scand 39:83-104 , 1957. 7. Desmedt JE: Plasticity of motor unit organization studied by coherent electromyography in patients with nerve lesions or with myopathic neuropathic diseases. In: Prog Clin Neurophysiol, Vol 9, Ed by Desmedt JE, Karger, Basel, pp 250-304, 1981.. 8. Engel AG, Banker BQ, eds: Myology basic and clinical, McGraw-Hill, NY, pp 245, 380-384, 1986. 9. Hilton-Brown P, Stalberg E: The motor unit in muscular dystrophy, a single fibre EMG and scanning EMG study. J Neurol Neurosurg Psychiatry 46:981-995, 1983. 10. Lindstrom L: Spectral analysis of EMG. In: Electromyog and Evoked Potentials. Theories and Applications, Ed by Struppler A, Weindle A, Springer-Verlag, NY, pp 103-107, 1985. 11. Lloyd AJ: Surface electromyography during sustained isometric contractions. J Appl Physiol 30:713-719, 1971. 12. Merletti R, Knaflitz M, DeLuca CJ: Myoelectric manifes-
13.
14. 15. 16.
17.
18.
19. 20. 21.
22.
179
tations of fatigue in voluntary and electrically elicited contractions. J Appl Physiol 69:1810-1820, 1990. Masuda T, Miyano H, Sadoyama T: The distribution of myoneural junctions in the biceps brachii investigated by surface electromyography. Electroenceph and Clin Neurophysiol 56:597-603, 1983. Masuda T, Sadoyama T: Processing of myoelectric signals for estimating the location of innervation zones in the skeletal muscles. Frontiers Med Biol Engng 1:299-314, 1989. Masuda T, Sadoyama T: Distribution of innervation zones in the human biceps brachii. J Electrornyogr Kinesiol 1:107-115, 1991. Roy SH, De Luca CJ: Evolving characteristics of the median frequency of the EMG signal. In: Computer-aided electromyography and expert systems, Clinical neurophysiology update, Vol 2, Ed by Desmedt JE, Elsevier, Amsterdam, pp 205-221, 1989. Sadoyama T, Masuda T, Miyano H: Relationship between muscle fiber conduction velocity and frequency parameters of surface EMG during sustained contraction. Eur J Appl Physiol 51:247-256, 1983. Shochina M, Gonen B, Wolf E, Mahler Y, Magora A: A study of the application of the Fourier series in the analysis of the electrophysiological pattern of development of muscle fatigue. Electromyogr Clin Neurophysiol 26:513-520, 1986. SPSS/PC+: SPSS Inc., Chicago, IL, 1988. Yaar I, Niles L: Muscle fiber conduction velocity and mean power spectrum frequency in neuro-muscular disorders and in fatigue. Muscle & Nerve (in press). Yaar I, Shapiro MB, Mitz AR, Pottala EW: A new technique for measuring muscle fiber conduction velocities in full interference patterns. Electroenceph Clin Neurophysiot 57:427--434, 1984. Zwarts MJ, Van Weerden TW, Haenen HT: Relationship between average muscle fiber conduction velocity and EMG power spectra during isometric contraction, recovery and applied ischemia. Eur J Appl Physiol 56:212-216, 1987.
Journal of Electromyography & Kinesiology Vol. 3, No. 3, 1993