Conference PaperPDF Available
INKA: CHAIR CATHETER TECHNOLOGIES + IMAGE GUIDED THERAPIES
Otto-von-Guericke-University of Magdeburg
P.O.Box 4120, 39016 Magdeburg, Germany
www.inka-md.de || inka@ovgu.de || +49 391 67 57024
PROXIMAL GUIDE WIRE AUDIO SENSING THE SOUND OF VESSEL PERFORATION
Alfredo Illanes, Iván Maldonado, Naghmeh Mammodian, Anna Schaufler, Axel Boese and Michael Friebe
Otto-von-Guericke University, Department of Medical Engineering, Magdeburg, Germany
Supported by BMBF 03IPT7100X INKA Kathetertechnologien, Contact: alfredo.illanes@ovgu.de
References
Introduction
Materials and Methods
Results
Conclusion
Background and Motivation
Artery guidewire-induced perforation during coronary
interventions is an uncommon but potentially serious
complication with significant morbidity and mortality rates [1].
For minimizing its impact it is crucial for the surgeon to early
detect that a perforation has occurred.
However, this is not always easy since perforation sometimes is
not characterized by any symptom or sign.
Proposal
Time-varying (TV) characterization of coronary artery perforation using
advanced signal processing of an audio signal acquired with a
stethoscope connected to the proximal part of a guide wire.
MAIN IDEA BEHIND
THE APPROACH How to classify an occurring guide wire event as being a vessel
perforation or artifact, by analyzing the time-varying dynamics of
the acquired audio signal during guide wire insertion.
Experimental setup and database implementation
A stethoscope equipped with a microphone
connected to a computer was directly and firmly
attached to the proximal end of a 0.014-inch guide
wire via a 3D printed adapter.
Coronary arteries belonging to 5pork hearts were
perforated using the tip of the guide wire.
Audio database implementation including:
560 coronary artery perforations.
315 induced guidewire audio artifacts: friction
between guidewire and artery wall, tiny
guidewire bumps.
Two methods for time-varying (TV) analysis
TV Autoregressive modelling [2, 3]
Perforation
Classification
Algorithm
Frequency [Hz]
Time [s]
Bispectral analysis [4]
Maximal energy TV pole for heart perforation
TV Autoregressive Spectrum for heart perforation
Audio
transducer
Audio signal
acquisition
Signal pre-
processing
Time-variant
signal
dynamics
extraction
Feature
extraction
Artifact
Before perforation During perforation After perforation
In this work a new approach for characterizing a puncture using non-invasive sensoring has been presented. Audio signals acquired with a stethoscope connected to
the proximal part of a guide wire has shown significant signal dynamic characteristics that can allow a correct classification with a simple indicator and without the use
of any complex intelligent technique. Preliminary results show that the time-varying trace that a perforation leaves in the audio signal is significantly different than
possible artifacts that may occur during an intervention.
[1] Teis-Soley, Albert, et al. "Coronary artery perforation by intracoronary guidewires: Risk Factors and Clinical Outcomes." Revista Española de Cardiología (English Edition) 63.6 (2010): 730-734.
[2] Mainardi, Luca T., et al. "Pole-tracking algorithms for the extraction of time-variant heart rate variability spectral parameters." IEEE Transactions on Biomedical Engineering 42.3 (1995): 250-259.
[3] Thanagasundram, Suguna, Sarah Spurgeon, and Fernando Soares Schlindwein. "A fault detection tool using analysis from an autoregressive model pole trajectory." Journal of Sound and Vibration 317.3 (2008): 975-993.
[4] Sigl, Jeffrey C., and Nassib G. Chamoun. "An introduction to bispectral analysis for the electroencephalogram." Journal of clinical monitoring 10.6 (1994): 392-404.
Feature extraction
method
Bispectral Analysis TV-AR Modelling
Classification
algorithm
K-NN SVM SVM
Accuracy
98.62% 97.01% 92.3%
Table 1: Accuracy results of the classification between heart perforation and artefact using K-nearest
Neighbor (K-NN) and Support Vector Machine (SVM) classification for bispectral analysis gained features
and SVM for features based on TV-AR modelling.
Obtained audio signals of three different events, of perforation and artifacts, are
shown in the figure above. The TV-AR spectrum of perforations shows stable
frequency components compared to the more disperse frequencies of artifact
signals. The TV-MEP image of the perforation shows characteristic segments which
does not occur with artifact signals; an overshoot at the beginning of the event,
followed by a plateau.
The figures above show a 3-dimensional feature space where each point
represents a audio signal. The left figure represents a separation of the
data by TV-AR based features. To the right of this, the data is projected
on features of bispectral analysis.
Maximal
energy
pole
Time [s]
TV AR
Spectrum
(3D view)
Original
Signal
TV AR
Spectrum
Frequency [Hz]
Heart perforation Friction
Frequency [Hz]
Time [s]
Frequency [Hz]
Time [s]
Bump
Frequency [Hz]
Time [s]
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.