Neutralisation of myoelectric interference from recorded nerve signals using models of the electrode impedance.
Doctoral thesis, UCL (University College London).
Any form of paralysis due to spinal cord injury or other medical condition, can have a significant impact on the quality and life expectancy of an individual. Advances in medicine and surgery have offered solutions that can improve the condition of a patient, however, most of the times an individual’s life does not dramatically improve. Implanted neuroprosthetic devices can partially restore the lost functionalities by means of functional electrical stimulation techniques. This involves applying patterns of electrical current pulses to innervate the neural pathways between the brain and the affected muscles/organs, while recording of neural information from peripheral nerves can be used as feedback to improve performance. Recording naturally occurring nerve signals via implanted electrodes attached to tripolar amplifier configurations is an approach that has been successfully used for obtaining desired information in non-acute preparations since the mid-70s. The neural signal (i.e. ENG), which can be exploited as feedback to another system (e.g. a stimulator), or simply extracted for further processing, is then intrinsically more reliable in comparison to signals obtained by artificial sensors. Sadly, neural recording of this type can be greatly compromised by myoelectric (i.e. EMG) interference, which is present at the neural interface and registered by the recording amplifier. Although current amplifier configurations reduce myoelectric interference this is suboptimal and therefore there is room for improvement. The main difficulty exists in the frequency-dependence of the electrode-tissue interface impedance which is complex. The simplistic Quasi-Tripole amplifier configuration does not allow for the complete removal of interference but it is the most power efficient because it uses only one instrumentation amplifier. Conversely, the True-Tripole and its developed automatic counterpart the Adaptive-Tripole, although minimise interference and provide means of compensating for the electrode asymmetries and changes that occur to the neural interface (e.g. due to tissue growth), they do not remove interference completely as the insignificant electrode impedance is still important. Additionally, removing interference apart from being dependent on the frequency of the interfering source, it is also subject to its proximity and orientation with respect to the recording electrodes, as this affects the field. Hence neutralisation with those two configurations, in reality, is not achieved in the entire bandwidth of the neural signal in the interfering spectrum. As both are less power efficient than the Quasi-Tripole an alternative configuration offering better performance in terms of interference neutralisation (i.e. frequency-independent, insensitive to the external interference fields) and, if possible, consume less power, is considered highly attractive. The motivation of this work is based on the following fact: as there are models that can mimic the frequency response of metal electrodes it should be possible, by constructing a network of an equivalent arrangement to the impedance of electrodes, to fit the characteristic neutralisation impedance – the impedance needed to balance a recording tripole – and ideally require no adjustment for removing interference. The validity of this postulation is proven in a series of in-vitro preparations using a modified version of the Quasi-Tripole made out of discrete circuit components where an impedance is placed at either side of the outer electrodes for balancing the recording arrangement. Various models were used in place of that impedance. In particular, representing the neutralisation impedance as a parallel RC reduced interference by a factor of 10 at all frequencies in the bandwidth of the neural signal while removed it completely at a spot frequency. Conversely, modelling the effect of the constant phase angle impedance of highly polarisable electrodes using a 20 stages non-uniform RC ladder network resulted in the minimisation of interference without the initial requirement of continuous adjustment. It is demonstrated that with a model that does not perfectly fit the impedance profile of a monopolar electrochemical cell an average reduction in interference of about 100 times is achieved, with the cell arranged as a Wheatstone bridge that can be balanced in the ENG band.
|Title:||Neutralisation of myoelectric interference from recorded nerve signals using models of the electrode impedance|
|Open access status:||An open access version is available from UCL Discovery|
|UCL classification:||UCL > School of BEAMS > Faculty of Engineering Science > Electronic and Electrical Engineering|
Archive Staff Only