This web only provides the extract of this article. If you want to read the figures and tables, please reference the PDF full text on Blackwell Synergy. Thank you.
- Original Article -
Application of correlation techniques in the analysis of corpus
cavernosum electromyographic signals
Xiao-Gang Jiang1, Jan
Holsheimer2, Ljubomir
Manola2, Gorm Wagner3, Hessel
Wijkstra4, Ben Knipscheer1, Eric J. H. Meuleman5
1Department of Urology, Radboud University Nijmegen Medical Centre, 6500 HB Nijmegen, the Netherlands
2Institute for Biomedical Technology, University of Twente, 7500 AE Enschede, the Netherlands
3Institute of Preventive Medicine, Copenhagen University Hospital, DK-1375 Copenhagen, Denmark
4Department of Urology, Academic Medical Centre, 1105 AZ Amsterdam, the Netherlands
5Department of Urology, Free University Medical Centre,1007 MB Amsterdam, the Netherlands
Abstract
Aim: To establish an objective, easy-to-use and comprehensive method to analyze corpus cavernosum
electromyographic signals (CC-potentials).
Methods: CC-potentials were recorded during flaccidity in 23 young healthy volunteers,
with surface electrodes placed on the penile shaft bilaterally. Based on the correlation function of Matlab software, an
application program for the analysis of CC-potentials was developed. Individual CC-potentials and their autocorrelation
function were evaluated, yielding parameters amplitude
(A), duration (D), and dominant frequency
(DF). The cross-correlation function of both longitudinal and bilateral pairs of adjacent electrodes was calculated to assess the
similarity and mutual delay of CC-potentials recorded simultaneously from different parts of the CC. The parameters derived
were squared maximum cross-correlation coefficient
(Rmax) and delay (τ). Based on the absolute value of
τ and the corresponding inter-electrode distance, propagation velocity
(PV) was calculated. Results: The values of the
parameters were determined automatically. No significant difference related to the locations of the electrodes for
parameters A, D, and DF was detected. The cross-correlation showed that both longitudinal and
bilateral CC-potential pairs had highly similar waveforms (the absolute values of
Rmax were 0.80 ± 0.05 and 0.87 ± 0.06, respectively).
PV of longitudinal pairs was estimated as 6.15
± 3.98 cm/s. Conclusion: The application program for correlation analysis
of CC-potentials is a comprehensive and versatile method to analyze corpus cavernosum electromyographic recordings.
Its objectiveness makes multi-center application possible.
(Asian J Androl 2007 May; 9: 369_376)
Keywords: corpus cavernosum; corpus cavernosum electromyography; electrophysiology; erectile dysfunction; smooth muscle
Correspondence to: Dr Eric J. H. Meuleman, Department of Urology, Free University Medical Centre, PO Box 7057, 1007 MB Amsterdam,
the Netherlands.
Tel: +31-20-444-0272 Fax: +31-20-642-5085
E-mail: E.Meuleman@vumc.nl
Received 2006-05-18 Accepted 2006-11-06
DOI: 10.1111/j.1745-7262.2007.00233.x
1 Introduction
Corpus cavernosum electromyography (CC-EMG) has failed to mature into a useful clinical tool for
diagnosing erectile dysfunction (ED), as a result of
insufficient understanding of the electrophysiology of the
corpus cavernosum (CC) and the recorded signals
(CC-potentials), and a lack of standardization of the recording
technique as well as signal processing and signal analysis
methods [1]. Recently, significant progress has been made
to overcome these shortcomings. The methodology of
CC-EMG recording was revisited. Monopolar recording has been shown to be superior to the traditional
bipolar recording [2]. With this setup, further evidence has
been obtained to support the notion that CC-potentials
reflect sympathetically mediated electrical activity of
ca-vernous smooth muscle (CSM) [2, 3]. However, a valid,
objective, and easy-to-use method to analyze
CC-potentials has not been established yet. Most clinical
investigators measured the values of parameters manually
[4_6], which is time-consuming, imprecise and not objective.
Stief et al. [7] and Kellner et al. [8] addressed this issue
by introducing Fourier analysis and computerized
classification of CC-potential components by fuzzy logic and
neural networks. However, these methods have not been
applied by other centers, probably because the required
basic knowledge of linear systems analysis is generally
beyond the expertise of clinical physicians. Later on the
same group explored the application of cross-correlation
function to estimate time delay of CC-potentials recorded
at different sites of the penis, which has only been
published in a monograph (in German) [9] and did not result
in the introduction of this method for the evaluation of
CC-EMG recordings.
This study was aimed at establishing an analysis
method that is easy-to-use for physicians,
comprehensive and objective. Because CC-potentials can be
considered spindle-like wave complexes [2], correlation
techniques were used in a CC-EMG application program. This
program was applied to a set of recordings in a group of
young volunteers.
2 Materials and methods
2.1 Data collection
The methodology to record CC-potentials has been
described in detail in a previous study [2]. Briefly, with
six or four surface electrodes (depending on the size of
the penis) placed on the penile shaft bilaterally,
CC-potentials were recorded simultaneously and monopolarly
for 20_30 min during flaccidity. To allow monitoring
the signals during the measurements, the recorded
signals were digitized and filtered with a bandpass filter
(cut-off frequencies 0.1 Hz and 20.0 Hz). Both unfiltered and
filtered digitized signals were stored on a computer. The
longitudinal inter-electrode distances (center to center)
were measured after the completion of the measurements.
Using Matlab software (MathWorks, Natick, MA, USA), the unfiltered signals of each channel were extracted
separately and filtered with a digital second-order bandpass
filter (cut-off frequencies 0.1 Hz and 5.0 Hz). The
purpose of reducing the higher cut-off frequency to 5.0 Hz
was to suppress high frequency, common signal
components (in particular electrocardiographic signals) as much
as possible without attenuating the CC-potential amplitude.
It has been demonstrated that CC-potential power in
normal subjects is actually below 5.0 Hz [2, 7].
2.2 CC-potential analysis
An application program for the analysis of
CC-potentials was developed based on the correlation function
of Matlab. Basically, the auto- and cross-correlation
functions reduce the influence of noise, thus allowing
the parameters of the CC-potential to be estimated more
accurately than from the signal itself [10]. Because the
onset and end of a CC-potential were not always definite,
an objective method to decide them seemed necessary.
With this application program the onset and end of a
CC-potential could be set to correspond with oscillations
exceeding a user-defined percentage of its maximum
amplitude. After testing with different percentages, 20%
turned out to be proper, because 20% of the maximum
peak-to-peak amplitude was just above the baseline
fluctuation (noise) level (approximately 75 µV) [6].
Individual CC-potentials were characterized by
analyzing the original signals and their autocorrelation
functions, and the corresponding values of the parameters
amplitude (A), duration (D), and dominant frequency
(DF) were determined. In Figure 1A and Figure 1B a
CC-potential and its autocorrelation function, respectively,
are shown. A was defined as the voltage difference
between the highest negative peak and the higher of the
two adjacent positive peaks. This strategy was chosen
because measuring the voltage difference between the
highest negative peak and the highest positive peak not
adjacent to the negative peak may be markedly affected
by baseline fluctuations when the baseline is instable
(Figure 2). D is the time window in seconds where the
A of the CC-potential exceeds a user-defined percentage
(20% in this study) of its maximum amplitude. DF
(Hz), the frequency of a CC-potential where most signal power
is present, was calculated from the time intervals
between zero crossings of the autocorrelation function.
To assess the similarity and mutual delay of
CC-potentials recorded simultaneously from different sites of
the corpora cavernosa, the cross-correlation function of
both longitudinal and bilateral pairs of adjacent electrodes
was calculated. Seven combinations were made for
recordings with six electrodes (Figure 3), and four
combinations in recordings with four electrodes. In Figure 4A
and Figure 4B, respectively, two simultaneously recorded
CC-potentials and their cross-correlation function are
shown. The parameters derived from the
cross-correlation function are the squared maximum
cross-correlation coefficient (Rmax) and the delay
(τ) between two CC-potentials. If the two signals are identical, then
Rmax is 1; if they have no components in common,
Rmax is 0; if they are identical but their phases are shifted by
exactly 180° (i.e. mirrored), then
Rmax is _1. All visually identified CC-potentials with a typical spindle-like
polyphasic waveform and a definite start and end were
included. However, in order to allow a comparison of
recordings with different numbers of CC-potentials, only
data of five CC-potential pairs with the highest absolute
values of Rmax were included. Based on the absolute
value of τ and the corresponding inter-electrode distance,
propagation velocity (PV) was estimated. To improve
the accuracy, PV was calculated only from pairs of
recordings with all five, or four out of five selected
CC-potential pairs having either a positive or a negative t and
thus the same "propagation" direction. When four out of
five selected CC-potential pairs had the same sign
of τ, only these values were included and the CC-potential pair
with opposite sign of τ was omitted. As a pilot
study, electrode combinations not adjacent to each other were
analyzed in several subjects. The results showed that
the sum of τ between, for example, distal-middle
electrodes and middle-proximal electrodes were
approximately equal to τ between distal and proximal electrodes
(data not shown). Therefore, it seems that including
more combinations may not provide extra information,
while it would be much more time-consuming.
In practice, the cross-correlation analysis was
performed first, in order to select five longitudinal
CC-potential pairs with the highest absolute Rmax
for the autocorrelation analysis.
2.3 Study population and recording equipment
Twenty-three healthy Caucasian men with a mean age of 24.7 years (range 19_32 years) were included in
this study. The volunteers were asked to refrain from
alcohol, coffee, smoking and sexual activity within a period
of 12-h prior to the measurements. Informed consent
was obtained from each subject.
Screener or a Porti system (TMS International, Enschede, the Netherlands) connected to a portable
computer (Satellitepro 6100, Toshiba, Tokyo, Japan) was used
to record CC-potentials. The amplifier parts of the two
systems were identical. The two systems had a slightly
different, fixed sampling rate (128 Hz and 100 Hz,
respectively). This difference does not affect the
digitized CC-potentials, since the same filter was used to
process the signals recorded with the two devices, and
the low-pass cutoff frequency was only approximately
4_5% of the sampling rates. The electrodes were
pre-gelled surface electrodes (type 9021S0231; Medtronic,
Copenhagen, Denmark).
2.4 Statistical analysis
SPSS software (SPSS Inc., Chicago, USA) was used
for the statistical analysis. One-Sample K-S Test was
applied to test the normality of distribution of parameter
values. Two Factorial Analysis of Variance was used to
test the effect of electrode location (left and right side,
proximal, middle and distal site) on the values of the
parameters.
3 Results
Among all 23 subjects, 10 had six electrodes and the
other 13 had four electrodes. Five electrode pairs showed
nearly identical baseline fluctuations and CC-potentials
(Rmax 1 and τ 0), indicating a short circuit between
the corresponding adjacent electrodes. The
CC-potentials recorded with these electrodes were excluded.
3.1 Single CC-potential analysis
The values of the parameters A, D, and
DF were determined automatically by the program. Normal
distribution was found in the values of all the three
parameters. The three parameters of CC-potentials
recorded at the same level (proximal-middle-distal) but on
opposite sides of the penis did not differ significantly
(Table 1). Therefore, the values on both sides were taken
together. The values of these parameters are shown in
Table 2. No significant difference related to the
locations of the electrodes was found for any parameter
(Tables 1 and 2).
3.2 Analysis of CC-potential pairs
Data of 60 longitudinal and 47 bilateral electrode pairs
were analyzed. The values of Rmax and
PV showed normal distribution. The mean ± SD of
Rmax of longitudinal and bilateral pairs were 0.80 ± 0.05 and 0.87 ±
0.06, respectively. In Tables 3 and 4 the "propagation"
direction of longitudinal and bilateral CC-potential pairs are
shown. In 71.7% of longitudinal pairs all five or four
out of five CC-potential pairs propagated distally, and in
8.3% of them all five or four out of five CC-potential pairs
propagated proximally. PV of longitudinal pairs of
CC-potentials was estimated at 6.15 ± 3.98 cm/s (mean ± SD).
The bilateral pairs had similar counts for a delay from left to
right (27.7%) and in the opposite direction (29.8%).
4 Discussion
In this study the correlation techniques are introduced
as an easy-to-use, comprehensive, and objective method
to analyze CC-potentials. In neurophysiology,
correlation analysis is a well-established methodology to
quantify the properties of striated muscle EMG and other
bioelectrical signals [10_12]. By analyzing original signals
and their autocorrelation functions, individual bioelectric
signals, such as CC-potentials, can be characterized
accurately. The cross-correlation function allows
quantifying the similarity of simultaneously recorded signals
(quantified as Rmax) and estimating their mutual time
delay (τ). If a signal, such as a CC-potential, is
propagating between two electrodes, PV can be simply
calculated [10, 11].
CC-potential amplitude (A) is the most commonly
used parameter [4_6]. In this study the value of
A in healthy men (Table 2) was in the same range as those in
the published literature [6]. Because CC-potentials
recorded with surface electrodes are supposed to reflect
the superimposed oscillatory membrane currents of a
group of adjoining CSM cells [1], presumably, a decrease
of A is expected when: (i) the CSM content is decreased
(CSM degeneration), (ii) the thickness of the tissue
enveloping CSM is increased, (iii) the intercellular
communication via gap-junctions is impaired, (iiii) the
sympathetic input is affected. The latter two situations may
result in a diminished synchronization of the
depolarization and repolarization of the CSM cell membranes. This
presumption has been supported by the observation that
the CC-potential A was decreased in patients with CSM
degeneration [13], penile edema [14], diabetes mellitus
[4], and in patients undergoing radical pelvic surgery,
and so on [5, 15].
Correlation analysis and spectral (Fourier) analysis
of CC-potentials provide the same information in the time
domain and the frequency domain, respectively. Accordingly,
DF calculated from the autocorrelation function corresponds with the frequency with the
highest power in the power density spectrum
[7]. As CC-potentials are supposed to reflect the superimposed
membrane currents caused by Ca2+ influx through L-type
voltage gated Ca2+ channels of
CSM cells [1], the value of DF as calculated in this study (approximately 0.25
Hz) is likely to correspond with the kinetics of these
Ca2+ channels [16]. Theoretically, myogenic pathologies
changing the membrane properties of CSM cells may alter the value of
DF.
The high Rmax of longitudinal pairs (0.80
± 0.05) indicates that the waveforms of CC-potentials recorded
simultaneously at different sites along the penile shaft
are highly similar, although not identical. Therefore,
Rmax, a parameter quantifying the similarity of
CC-potentials recorded simultaneously in different parts of the
CC, may be useful to reflect coordination of electrical
activity in different parts of the CC. According to the
existing knowledge, the sympathetic input and
communication via gap junctions between CSM cells are
responsible for the initiation, propagation and coordination of
electrical activity within the CC [17, 18]. Based on this
notion, sympathetic neuropathy or conditions affecting
the communication between CSM cells may result in a
decreased Rmax of CC-potential pairs by, for example,
having irregular waveforms, different DF
and D, or even a failure of propagating electrical activity from one site
to another.
The majority of CC-potentials propagated in a distal
direction, towards the tip of the CC (Table 3) indicating
that CC-potentials are mostly initiated in the proximal part
of the CC. However, in 38.3% of the longitudinal pairs
the CC-potentials propagated both distally and proximally,
suggesting that CC-potentials may be initiated at more
than one site in the CC. In 5% of the pairs all five
CC-potentials propagated in a proximal direction, suggesting
that the initiation sites can be present in the distal part of
the CC.
Although bilateral CC-potential pairs had highly
similar waveforms (high Rmax), the fact that bilateral pairs
had no preferential direction of τ (Table 4), together with
the existing knowledge that the left and right corpora
cavernosa are innervated separately [17], indicate that
the coordination of electrical activity in the two
cavernous bodies possibly occurs in the pre-CC level (e.g.
spinal cord), rather than by a mechanism in which
electrical activity propagates among CSM cells via gap
junctions as happens within one CC.
In this study, PV was estimated as 6.15 ± 3.98
cm/s, whereas Gorek et al. [9] calculated a similar mean value
of 5 cm/s. This value indicates that CC-potentials travel
via gap junctions among CSM cells rather than via nerve
fibers. In the latter case, PV would have been much
higher (approximately 100 cm/s) [19]. Although a change
in PV is an important indicator of muscle disease in
striated muscle EMG [20], the value of
PV to detect myogenic pathology in CC-EMG is questionable. Because
the inter-electrode distance cannot be kept constant
during a 20_30 min recording session (the penile length
changes depending on a subject's sympathetic tone,
room temperature, etc.), PV is only useful if its change
caused by pathology is larger than the variation caused
by the variable inter-electrode distance. Furthermore,
PV can only be estimated correctly when the electrodes
are aligned with the propagation direction. Because CSM
is a three-dimensional structure, CC-potentials may
propagate in various directions, even though the overall
direction is longitudinal.
No difference for any parameters in relation with the
electrodes locations was detected, indicating that only
using one or two electrodes would be enough to obtain
representative CC-potentials. In fact, in our later study
we only included CC-potentials from left and right
proximal sites for calculating the values of
A, D and DF, although four electrodes were used. The reason we
included CC-potentials from both sides of the penis is that,
under certain circumstances (e.g. unilateral cavernous
nerve damage following pelvic surgery), the change in
CC-potential patterns may be only unilateral and, therefore,
recording CC-potentials only from one CC may result in
a false-negative finding. In order to assess the
coordination of the electrical activity in different parts of the
penis (by calculating Rmax), four electrodes were still used,
although the CC-potentials from the distal electrodes were
not included for calculating other parameters.
In conclusion, the application program for
correlation analysis of CC-potentials introduced in this study is
a comprehensive and easily applicable method to analyze
CC-EMG recordings. Its objectiveness makes
multi-center application possible. By calculating the parameters
A, D, DF, Rmax, and
PV, CC-potentials can be characterized adequately. The next steps of our study are to
test the reproducibility (stability) of the parameters, and
to investigate their sensitivity and specificity to detect
myo- and neurogenic pathology of the CC.
Acknowledgment
This study was supported by an unrestricted research
grant of Pfizer (NL), a grant from Amsterdam 1998 Foundation, and is part of the EU-supported program
COST Action B18. The authors thank Dr Jos Frantzen
for his valuable comments, and Mr. John Philippi and
Mr. Hilco van Moerkerk for their technical assistance.
References
1 Jiang XG, Speel TG, Wagner G, Meuleman EJ, Wijkstra
H. COST Action B18 project. The value of corpus cavernosum
electromyography in erectile dysfunction: current status and
future prospect. Eur Urol 2003; 43: 211_8.
2 Jiang XG, Wijkstra H, Meuleman EJH, Wagner G. The
methodology of corpus cavernosum electromyography revisited. Eur
Urol 2004; 46: 370_6.
3 Jiang X, Meuleman EJ, Wijkstra H, Wagner G. Corpus
cavernosum electromyography during morning naps in healthy
volunteers: further evidence that CC-potentials reflect
sympathetically mediated activity. J Urol 2005; 174: 1917_20.
4 Merckx L, Schmedding E, De Bruyne R, Stief C, Keuppens F.
Penile electromyography in the diagnosis of impotence. Eur
Urol 1994; 25: 124_30.
5 Sasso F, Gulino G, Alcini A, Alcini E. Early experience of
corpus cavernosum electromyography in impotent patients
after radical cystoprostatectomy. Eur Urol 1996; 29: 466_9.
6 Merckx L, Gerstenberg TC, Da Silva JP, Portner M, Stief CC.
A consensus on the normal characteristics of corpus cavernosum
EMG. Int J Impot Res 1996; 8: 75_9.
7 Stief CG, Kellner B, Hartung C, Hauck E, Schlote N, Truss M,
et al. Computer-assisted evaluation of the smooth-muscle
electromyogram of the corpora cavernosa by fast Fourier
transformation. Eur Urol 1997; 31: 329_34.
8 Kellner B, Stief CG, Hinrichs H, Hartung C. Computerized
classification of corpus cavernosum electromyogram signals
by the use of discriminant analysis and artificial neural
networks to support diagnosis of erectile dysfunction. Urol Res
2000; 28: 6_13.
9 Gorek M, Hartung C, Stief CG.
Bioelektrische Signalausbreitung im humanen glattmuskulären Corpus cavernosum: modellierung
sowie experimentelle Untersuchungen in vivo und
in vitro. Düsseldorf: VDI Verlag; 2003. p98.
10 van der Vliet GH, Holsheimer J, Bingmann D. Calculation of
the conduction velocity of short nerve fibers.
Med Biol Eng Comput 1980;18: 749_57.
11 Naeije M, Zorn H. Estimation of the action potential
conduction velocity in human skeletal muscle using the surface EMG
cross-correlation technique. Electromyogr Clin Neurophysiol
1983; 23: 73_80.
12 Houk JC, Dessem DA, Miller LE, Sybirska EH. Correlation
and spectral analysis of relations between single unit
discharge and muscle activities. J Neurosci Methods 1987; 21:
201_24.
13 Sattar AA, Merckx LA, Wespes E. Penile electromyography
and its smooth muscle content: interpretation of 25 impotent
patients. J Urol 1996;155: 909_12.
14 Stief CG, Thon WF, Djamilian M, Allhoff EP, Jonas U.
Transcutaneous registration of cavernous smooth muscle electrical
activity: noninvasive diagnosis of neurogenic autonomic
impotence. J Urol 1992; 147: 47_50.
15 Truss MC, Djamilian MH, Tan HK, Hinrichs H, Feistner H,
Stief CG, et al. Single potential analysis of cavernous electrical
activity. Eur Urol 1993; 24: 358_65.
16 Hashitani H, Fukuta H, Dickens EJ, Suzuki H. Cellular
mechanisms of nitric oxide-induced relaxation of corporeal smooth
muscle in the guinea-pig. J Physiol 2002; 538(Pt 2): 573_81.
17 Andersson KE, Wagner G. Physiology of penile erection.
Physiol Rev 1995; 75: 191_236.
18 Christ GJ. The "syncytial tissue triad": a model for
understanding how gap junctions participate in the local control of
penile erection. World J Urol 1997; 15: 36_44.
19 Fagius J, Wallin BG. Sympathetic reflex latencies and
conduction velocities in normal man. J Neurol Sci 1980; 47: 433_48.
20 Van der Hoeven JH, Zwarts MJ, Van Weerden TW. Muscle
fiber conduction velocity in amyotrophic lateral sclerosis and
traumatic lesions of the plexus brachialis. Electroencephalogr
Clin Neurophysiol 1993; 89: 304_10. |