Computational neuroscienceA new method for spike extraction using velocity selective recording demonstrated with physiological ENG in Rat
Introduction
If it were possible to construct an ideal nerve interface, such a system would allow recording from and stimulation of every single axon in the nerve; it would be stable in time so that each axon, once identified, would have a known function. Every axon would have a physiological label so, for example, some axons in the femoral nerve, after identification, would be labelled vastus lateralis, efferent. However, before these physiological labels could be attached, it would be helpful to know the propagation velocity and direction for every fibre (afferent or efferent) which would greatly reduce its possible range of function. At present of course, no such interface exists. Practical methods start with the axon-specific, such as intra-fascicular devices, fine tungsten needle electrodes (Yoshida et al., 2000) or arrays such as the Utah design (Maynard et al., 1997); these are typically invasive, show poor chronic applicability in peripheral nerves, and give no indication of the axon/action potential characteristics. At the extra-fascicular level, cuffs can be safe for chronic clinical use, but are limited in recording the composite activity from all the axons in the nerve. In this range of types, Longitudinal Intrafasicular Electrode (LIFE) arrays can record from small groups of axons, perhaps within one fascicle, but do not show activity in the whole nerve (Boretius et al., 2010). Micro-channel nerve interfaces have enabled inter-fascicular recording from peripheral nerves with single unit activity resolution, but the number of axons in each micro-channel is quite large (∼100, FitzGerald et al., 2012) and physiological characterisation is still limited. No current interface method allows communication with one axon, or even a group of only a few axons, with physiological labelling.
The method that we call velocity selective recording (VSR) has been applied to cuffs and, by extension, could be applied to micro-channels. By filtering the neural signal in the velocity domain, activity within bands of conduction velocity can be discriminated and, if that band corresponds to a functional group of fibres, such as, for example, the γ efferents which are responsible for muscle spindle contraction, it should be possible to estimate the activity in those fibres. Thus the VSR method should improve both the quantity and quality of the information that can be extracted from the neural signal using practical types of electrode structure. This improvement may yield substantial benefits in a clinical neuroprosthesis if one can provide better resolution at the input. However, in spite of the many potential applications of VSR, to date it has only been demonstrated with electrically-evoked electroneurogram (ENG), i.e. compound action potentials (CAPs) (Schuettler et al., 2011, Schuettler et al., 2013). This is because there are several very significant differences between the requirements of recording electrically-evoked and natural ENG that complicate the process of recording the latter. These differences include much smaller signal amplitudes (typically, using cuffs, 1–10 μV, as opposed to about 100 μV for CAPs) and the need to determine the rate of neural firing in a particular velocity band, rather than the relative amplitudes of activity between bands, which is generally the case for CAPs. This paper presents the first experimental validation of a new VSR-based method using naturally-evoked (physiological) signals. The data were obtained from a strand of intact nerve within a dorsal/sensory root of a rat and a new method was employed that overcomes the difficulties of recording natural ENG and allows neuronal firing rates in specified velocity bands to be computed in real time. We call this the method of velocity spectral density (VSD) (Metcalfe et al., 2014).
The new method was validated by capturing and then manually calculating the propagation velocities of individual spikes (i.e. action potentials-APs) and comparing the firing rates in each of a chosen set of velocity bands with the output from the VSD processor. The recording was made with the nerve resting on hook electrodes immersed in oil. This was convenient for the experiment but we expect that the signals so obtained were similar to the outputs from a row of electrodes in a micro-channel. Both arrangements greatly increase the amplitude of extracellular potentials (Vex) compared to a nerve of diameter 1 mm or greater placed in a cuff, enabling individual ENG spikes to be distinguished and counted: this is essential for the method and also allows validation by inspection of the electroneurograms. The ten electrodes were connected in separate pairs to form five bipolar recording channels before amplification and band-pass filtering.
There is, of course, a long history of analysing neural recordings from microelectrodes in brain. These methods normally identify spikes by the characteristics of their shape. This spike sorting is generally not done in real time and the methods often use substantial computing power (Gibson et al., 2012). By comparison, the proposed VSD method, in common with other VSR-based approaches, can operate in real time and is relatively economical in terms of computational effort. These features are important in certain neuroprosthetic devices such as the “Bioelectronic Medicines” currently being advocated by GlaxoSmithKline (Famm et al., 2013). This is because the devices must be small and low-powered and the firing rates of fibres that serve different functions must be calculated without significant computation delays.
Section 2 considers the VSR approach in general and the modifications and new methods necessary for it to compute VSDs while Section 3 describes the experimental methods employed. Section 4 describes the experimental results and Section 5 provides discussion and conclusions.
Section snippets
Delay-and-add
The essence of VSR is a simple process called delay-and-add that is analogous to beam-forming algorithms used in certain types of synthetic aperture arrays (Huang and Miller, 2004). The channels are delayed relative to the last channel VB1 by an interval that depends on both the electrode spacing and the propagation velocity of the signal. So if the delay between the first two channels (VB1, VB2) is dt the delay between the first and third channels (VB1, VB3) is 2·dt and so on. The general
Surgical procedure
All animal procedures were performed in accordance with the United Kingdom Animal (Scientific Procedures) Act 1986. An adult female Sprague Dawley rat (250 grams) was anaesthetised with 1.5 g/kg urethane (Sigma) administered by the intraperitoneal route. The dorsal spinal cord was exposed via a laminectomy of three of the lumbar spinal vertebrae. The dorsal skin was sutured to an over-hanging rectangular bar, creating a contained pool into which non-conductive mineral oil was poured. The dura
Electrical stimulation
Electrical stimulation was applied to the L5 dermatome to test the recording system and record the CAP from the cutaneous afferents. The stimulation waveform was a fixed width (100 μs) square pulse of variable amplitude and Fig. 7 shows the resulting time domain response, the IVS and the VD waveforms for the peak stimulation current of 4 mA (of length 10 ms). The location of the peaks within the IVS for each stimulation current is given in Table 2.
At lower stimulation currents the peak velocities
Validity of results
Within the nervous system, information is encoded in terms of neuronal firing rates and so an increase in the amplitude of the stimulus results in a correlated increase in the rate of AP generation (Milner-Brown et al., 1973). As an example, the afferent fibres that contain information about the fullness of the human urinary bladder have been observed in man to propagate at a mean velocity of 41 m/s with a base-line firing rate of about 15 APs per 200 ms and a rate representing a full bladder of
Conclusions
A method for extracting neuronal firing rates from physiological ENG based on conduction velocity has been demonstrated using in-vivo recordings in rat. Simple wire hook electrodes were used to form a short recording array in which a micro-dissected but intact fascicle was placed. Data were recorded using commercially available amplifiers and data converters before being processed using basic operations in MATLAB. This method generates a detailed overview of the firing rates of neurons based on
Acknowledgment
This work was generously supported by the Brian Nicholson PhD scholarship.
References (26)
- et al.(2010)
- et al.
Conduction velocity is inversely related to action potential threshold in rat motoneuron axons
Exp brain Res
(2003) - et al.
A microchannel neuroprosthesis for bladder control after spinal cord injury in rat
Sci Transl Med
(2013) - et al.
Concurrent recordings of bladder afferents from multiple nerves using a microfabricated PDMS microchannel electrode array
Lab Chip
(2012 Jul 21) - et al.
Noise and selectivity of velocity-selective multi-electrode nerve cuffs
Med Biol Eng Comput
(2008 Oct) - et al.
Drug discovery: a jump-start for electroceuticals
Nature.
(2013 Apr) - et al.
A regenerative microchannel neural interface for recording from and stimulating peripheral axons in vivo
J Neural Eng
(2012) The classification of nerve fibers
Ohio J Sci
(1941)- et al.
Spike sorting the first step in decoding the brain: the first step in decoding the brain
IEEE Signal Process Mag
(2012) - et al.
Phased-array processing for spike discrimination
J Neurophysiol
(2004)
The Utah intracortical electrode array: a recording structure for potential brain-computer interfaces
Electroencephalogr Clin Neurophysiol
An enhancement to velocity selective discrimination of neural recordings: Extraction of neuronal firing rates
Annu Int Conf IEEE Eng Med Biol Soc
Changes in firing rate of human motor units during linearly changing voluntary contractions
J Physiol
Cited by (27)
Array processing of neural signals recorded from the peripheral nervous system for the classification of action potentials
2021, Journal of Neuroscience MethodsCitation Excerpt :These disadvantages are made clear using comparative simulations. The most basic velocity selective recording array can be constructed from wire hooks, with typically between 5 and 10 electrodes (Metcalfe et al., 2015; Schuettler et al., 2013), these are easy to fabricate and can be made by hand with limited tooling. However, hooks are only suitable for acute experiments and, due to their hand assembly, there is often a large variation in the actual electrode spacing.
Closed-Loop Control of Functional Electrical Stimulation Using a Selectively Recording and Bidirectional Nerve Cuff Interface
2024, IEEE Transactions on Neural Systems and Rehabilitation EngineeringNeural network-based classification of ENG recordings in response to naturally evoked stimulation
2022, Proceedings of the 9th ACM International Conference on Nanoscale Computing and Communication, NANOCOM 2022Tutorial: A guide to techniques for analysing recordings from the peripheral nervous system
2022, Journal of Neural EngineeringA comparison of extraneural approaches for selective recording in the peripheral nervous system
2022, Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS