MATERIALS AND METHODS
6. DATA ANALYSIS
Detailed analysis of rhythmic sympathetic discharges recorded from
sympathetic fibres or whole nerves is performed using time domain or
frequency domain analytical techniques. Time domain analysis involves the
detection of a discharge and plots the occurrence of an event (e.g. sympathetic
discharge) following an event (e.g. sympathetic discharge, phrenic nerve burst
discharge, R-wave on ECG) on a correlogram. This approach is ideal for
revealing the rhythmicity of discriminated action potentials (e.g. Johnson &
Gilbey, 1996; Yusof & Coote, 1988a). However, it is not suitable for whole nerve
recordings in which the sympathetic discharges cannot be discriminated using a
spike processor or interface (e.g. sympathetic activity recorded VCNs, DCNs
and saphenous nerves; see Figures 3.2, 4.1, 4.4). Thus, sympathetic activity
recorded from whole nerves is often subjected to frequency domain analysis
(e.g. Kenney e t at, 1991; Kocsis e t a/., 1990; Taylor & Schramm, 1987). When nerve activity is analysed in the frequency domain, the frequency components
in the nerve recording are separated by fast Fourier transformation (FFT); the
amount of nerve activity occurring at a particular frequency is then quantified
(voltage squared) and plotted on an autospectrum (Chatfield, 1996). Cross-
spectral calculations can also be performed to produce coherence spectra that
indicate constant phase and amplitude ratios of frequency components between
paired nerve recordings (Chatfield, 1996; Christakos 1986 & 1994; Rosenburg
e t at, 1989).
6.1 Time domain analysis of ongoing discharges recorded from single PGNs
300 8 data sets of TTL pulses generated from discriminated PGN action
potentials and rhythmic phrenic nerve activity were sampled by the computer
using Spike2 computer software (see above, Figure 2.3). Autocorrelograms (50
ms bin width) were used to reveal the periodicity of PGN and phrenic nerve
burst discharges. Autocorrelograms are self triggered histograms that plot the
occurrence of TTL pulses following a TTL pulse. The dominant frequency of
rhythm revealed on autocorrelograms was calculated from the reciprocal of the
modal time of the first peak, as shown in Figure 2.5.
6.2 Frequency domain analysis of the sympathetic discharges recorded from whole nerves
300s of integrated whole nerve activity were sampled at 100 Hz by the
computer, using Spike2 computer software (see above). The Spike2 data file
was then converted to a text file and analysed using Matlab computer software
(Maths Works). The Matlab scripts used to analyse data in the present study
were purpose written by Dr. H.-S. Chang (Physiology Department, Royal Free
and University College Medical School, London, UK).
(i) Autospectra
FFT (size 2048, 50% overlap) was performed on 286 72 s of integrated
whole nerve activity (sampled at 100 Hz). This separated the frequency
components in the nerve activity and divided the data set into 28 half
the spectrum a frequency resolution of 0.049 Hz. Do components and linear
trend were then removed from the data set. An autospectrum was averaged
from these subsections according to the Welch Method (see Chatfield, 1996).
(ii) Definition of a peak in an autospectrum
The term “peak” was allocated to frequencies with discrete spectral
density (> 50% of background power density) where the “rising phase"
occurred within 0.3 Hz. During central apnoea renal nerve autospectra were
considered to lack “peak” frequencies, since remaining power density on such
autospectra were similar to the broad power under the peaks in autospectra
seen during periods of rhythmic phrenic nerve activity, see Figure 2.6.
(iii) Allocation of dominant nerve discharge frequency
Dominant nerve discharge frequency (or T-peak frequency for VCN,
DCN and saphenous nerve recordings) was allocated to the frequency of the
highest peak. However, when the frequency of the highest peak coincided with
the 1®* harmonic frequency of phrenic nerve burst discharge frequency and
there was another peak which was more than 50% of its size (9/102
“sympathetic nerve” autospectra), dominant nerve discharge frequency was
allocated to the 2"^ highest peak. This criterion was applied to insure that
dominant nerve discharge / T-peak frequency was not allocated inappropriately
to a harmonic frequency of rhythmic phrenic nerve activity (Application of
frequency domain analysis to sympathetic nerve activity (which is not a
sinusoidal waveform) generates spectral density at harmonic frequencies of
rhythmic nerve discharges (Chatfield, 1996)).
(iv) Coherence spectra
A coherence spectrum was used to reveal linear correlations between
rhythmic sympathetic discharges at T-rhythm frequency recorded from the VCN
and this activity recorded simultaneously from another nerve (e.g. contralateral
VCN, ipsilateral saphenous nerve, DCN or renal nerve). Assuming that
synchrony between discharges on two nerve recordings is regular and stable,
then a coherence value significantly different from 0 would indicate that
discharges at that particular freauencv were linearly related.
Coherence spectra were averaged from the same 28 half-overlapped
subsections used to generate autospectra (see above). The squared coherence
coefficient (referred to as coherence value) at each frequency was estimated by
normalising the cross spectrum between two nerve activities (see Chatfield
1996; Christakos, 1986 & 1994).
The upper 95% confidence limit for coherence at a particular frequency
to be significantly different from 0 was calculated using the following equation:
---riTi:^---
Upper 95% Confidence Limit = 1-0-05
where: L = number of subsections in the spectrum (Rosenburg e t a/., 1989)
Figure 2.7 shows the upper 95% confidence limits for coherence spectra
constructed from different numbers of subsections. The coherence spectra
generated in the present study were constructed using 28 subsections. Thus,
the upper 95% confidence limit used in the present study was 0.1.
6.3 Calculation of mean arterial blood pressure
equation:
MAP = diastole pressure + / (systolic - diastolic pressure)
Diastolic and systolic blood pressure values were averaged from 6
values taken from the data set at 60 s intervals.
6.4 Linear regression analysis
Linear regression analysis was performed to test if dominant nerve
discharge, T-peak or T-rhythm frequency displayed a linear 1:1 relationship
with phrenic nerve burst discharge frequency. Least square linear regression
analysis was performed using Microcal Origin 3.54 (Microcal Software Inc). The
slopes of regression lines were compared to 1 using a Student t-test (Glantz,
1996).