Fast signal processing techniques for surface somatosensory evoked potentials measurement
FAST SIGNAL PROCESSING TECHNIQUES FOR SURFACE SOMATOSENSORY EVOKED POTENTIALS MEASUREMENT Submitted by
Shing Chun Benny LAM
for the degree of Master of Philosophy at The University of Hong Kong
in August 2003
Somatosensory evoked potential (SEP) testing has been widely applied to intraoperative spinal cord integrity monitoring, diagnosis of various neurological disorders, and nerve conduction velocity measurements. However, the SEP recorded using surface electrodes is buried in noises that are both electrical and biological in nature. Hence, the noninvasive measurements of these potentials suffer from very poor signal-to-noise ratios (SNR). Some means of signal processing is required to extract SEP signal from strong background noise.
The most commonly used technique of SEP extraction is ensemble averaging (EA). The conventional EA method usually requires several hundred to thousands of raw SEP input trials to produce an identifiable waveform for latency and amplitude measurement. This is time-consuming and may lead to the failure to detect the dynamic behaviour of the evoked potentials. Therefore, a fast and accurate SEP extraction technique is needed to reduce the measurement time.
An adaptive Signal Enhancer (AS E) was applied to extract the weak and noisy SEP signal from anesthetized subjects during surgery. ASE has a self-learning ability in that the weights of the ASE are able to adjust according to each input trial. This ability makes the ASE suitable to track the noisy and time-varying SEP in fewer input trials than EA. The best estimation of the SEP signal can be obtained upon the
convergence of the ASE. ASE with 50 input trials provided results comparable to those extracted by conventional EA. In order to examine the ability of ASE in detecting SEP during spinal cord compression, an animal study simulating different level of spinal cord compression was conducted. ASE with 50 input trials successfully detected the SEP during the normal situation and spinal cord.
During neurological diagnosis of conscious subjects, noises are much more complicated and severe than those III anesthetized subjects. On-going electroencephalography, electromyography, visual evoked potentials, and brain stem auditory evoked potentials continuously add to the surface SEP waveform during the data acquisition process. ASE was insufficient to extract an SEP for latency and amplitude measurement. A Multi-Adaptive Filtering (MAF) technique was developed for this purpose. This technique is a combination of an Adaptive Noise Canceller (ANC) and the ASE in which the raw surface recorded SEP is first processed by ANC with a reference noise channel of background noise for adaptive subtraction before entering ASE. The purpose of the ANC is to eliminate the correlated noises so that the SNR is increased before ASE processing. The MAP was theoretically developed and verified by filtering simulated SEP signals III which electroencephalography and Gaussian noise of different SNRs were added. The technique was also applied to track surface SEP recorded from conscious human subjects. It was found that the MAP provided similar SEP detection to the conventional averaging method in much less data acquisition time. Efficient and effective surface SEP measurement for neurological diagnosis is beneficial to clinicians and patients.
School:The University of Hong Kong
School Location:China - Hong Kong SAR
Source Type:Master's Thesis
Keywords:university of hong kong dissertations evoked potentials electrophysiology signal processing
Date of Publication:01/01/2004