Ingen bildrubrik

1
Matched Filtering and Multiple Hypothesis
Tracking Applied to C-Fiber Action Potentials
Recorded in Human Nerves
Björn Hansson,a c Clemens Forster,b and Erik
Torebjörk c
aSignals and Systems, Uppsala University,
Sweden bInstitute of Physiology and Experimental
Pathophysiology, University of Erlangen,
Germany cDepartment of Clinical Neurophysiology,
Uppsala University Hospital, Sweden
Introduction

A human skin nerve contains many fibers of
different types. One type of fibers, the
unmyelinated (C-) fibers, were studied in this
work as they play a key role in pain. Within the
C-fiber class, there are further subclasses which
can be identified by their characteristic
responses to mechanical, thermal, or chemical
stimuli. The response characteristics of a
C-fiber can be studied by observing the changes
in conduction velocity due to different
provocations. This utilizes the peculiar quality
of the C-fibers that the conduction velocity is
decreased temporarily after an action potential
(AP) has been conducted. By evoking APs
repetitively, changes in conduction velocity can
be observed through changes in the latency of the
APs. Here we describe a tracking technique which
facilitates the analysis of the latency shifts
induced by various stimuli in different classes
of human C-fibers.
Cross section
READ ME! Dear downloader, Thank you for your
interest in my research! This poster was
presented on the SPIE conference Signal and Data
Processing of Small Targets 1998 in Orlando Apr
13-17, 1998. To facilitate the design of the
poster, I chose the A2 page size. It is however
printable in A4 if you choose the alternative Fit
on paper or Passa in på papper (swedish). If you
have any questions or comments, please feel free
to contact me. Best whishes, Björn ------------
--------------------------------------------------
------- Bjorn Hansson, PhD student in
Neurophysiological Signal Processing Dep. of
Clinical Neurophysiology Signals and
Systems University Hospital
Uppsala University http//www.neurofys.uu.se
http//www.signal.uu.se E-mail
Bjorn.Hansson_at_neurofys.uu.se, or
Bjorn.Hansson_at_signal.uu.se
Overview
The fibers in the nerve are bundled in fascicles.
The needle electrode is inserted into the
fascicle of the C-fiber under study and the APs
are recorded.

• The action potentials were detected with a
  matched filter (MF).
• The set of subsequent APs originating from the
  same C-fiber was found with the multiple
  hypothesis tracking (MHT) method.
• The time course of the latency change after
  activation was modeled as an exponential decay
  and estimated by a Kalman-filter.
• The amplitude of the APs was varying slowly over
  time for each C-fiber and was in general
  different for different fibers. The amplitude was
  therefore incorporated into the tracking
  algorithm to improve its performance.

Electrical stimuli are applied repetitively with
a period of 4 seconds to obtain an estimate of
the conduction velocity. Additional stimuli
(mechanical, chemical and/or electrical) are
applied to study the resulting latency shift and
the recovery time course.
Recording
Latency ms
The action potentials of one C-fiber are shown.
The responses are elicited by the periodic
electrical stimuli. By displaying successive
traces from top to bottom, changes in conduction
velocity are easily detected. At trace 13, the
fiber is activated by a mechanical stimulus
leading to a decrease in conduction velocity,
i.e. an increase in latency. Following this, the
conduction velocity recovers gradually as
indicated by the APs returning to the latency
prior to the activation.
Results
Trace 1
Latency ms
Latency ms
Latency ms
Bottom The latencies in each trace where the
matched filter detected APs are marked with dots.
The responses of the five fibers can now be seen
more easily. The mechano-sensitive fiber is,
however, difficult to track as its trajectory
crosses the two sympathetic trajectories. Top
To improve the tracking performance, the MF peak
output at the detection time instant was used as
a measure of the AP amplitude. As can be seen in
the figure, it was in general different for
different fibers.
Bottom The action potentials from (at least)
five C-fibers were recorded. There are two
insensitive fibers (441 and 456 ms) and one
mechano-sensitive fiber (467 ms). At trace 11,
the latter was activated by a mechanical stimulus
causing an increase in latency. Furthermore,
there are two sympathetic fibers (gt 490 ms).
Characteristic for this type of C-fibers is that
they are continuously active resulting in their
irregular trajectories. Top Each trace is a
mixture of noise and APs of different amplitude
due to the relatively large recording electrode.
Bottom and Top The two insensitive fibers (cyan)
and the mechano-sensitive fiber (blue) were
tracked successfully. Both corresponded well with
the model used in the Kalman-filter. The need for
the amplitude information can be seen in the
upper plot. Without this measure, the tracking
performance of the mechano-sensitive fiber would
have been degraded. Although not modeled, the
two sympathetic fibers (red) were partly tracked.
In this particular case, the tracking was even
more difficult as their amplitudes were about the
same.
Conclusions

• The method simplifies the analysis and diminishes
  the required manual intervention to a minimum.
• The tracking is somewhat sensitive to model
  errors.

• The matched filter is a good detector of the
  action potentials.
• The tracking results of the MHT method correspond
  well with what an experienced analyst would
  consider to be correct.