Tracking single units in chronic, large scale, neural recordings for brain machine interface applications

Abstract

In the study of population coding in neurobiological systems, tracking unit identity may be critical to assess possible changes in the coding properties of neuronal constituents over prolonged periods of time. Ensuring unit stability is even more critical for reliable neural decoding of motor variables in intra-cortically controlled brain-machine interfaces (BMIs). Variability in intrinsic spike patterns, tuning characteristics, and single-unit identity over chronic use is a major challenge to maintaining this stability, requiring frequent daily calibration of neural decoders in BMI sessions by an experienced human operator. Here, we report on a unit-stability tracking algorithm that efficiently and autonomously identifies putative single-units that are stable across many sessions using a relatively short duration recording interval at the start of each session. The algorithm first builds a database of features extracted from units' average spike waveforms and firing patterns across many days of recording. It then uses these features to decide whether spike occurrences on the same channel on one day belong to the same unit recorded on another day or not. We assessed the overall performance of the algorithm for different choices of features and classifiers trained using human expert judgment, and quantified it as a function of accuracy and execution time. Overall, we found a trade-off between accuracy and execution time with increasing data volumes from chronically implanted rhesus macaques, with an average of 12 s processing time per channel at ~90% classification accuracy. Furthermore, 77% of the resulting putative single-units matched those tracked by human experts. These results demonstrate that over the span of a few months of recordings, automated unit tracking can be performed with high accuracy and used to streamline the calibration phase during BMI sessions. Our findings may be useful to the study of population coding during learning, and to improve the reliability of BMI systems and accelerate their deployment in clinical applications.

Long term recordings with immobile silicon probes in the mouse cortex

“…For example, using such a metric in studies of plasticity would lead to biased results (an underestimate of receptive field changes). To determine which recorded units are stable, therefore, a method based solely on waveform shapes should be used, such as those proposed by Tolias et al (2007), or Eleryan et al (2014).…”

Biomimetic and Non-biomimetic Extraction of Motor Control Signals Through Matched Filtering of Neural Population Dynamics

“…Since a typical decoder does not assign the same weights to all units, it may be desirable – from a pragmatic standpoint – to have more unit contributions to the computation of the decode. This can help in maintaining a decoder fixed across multiple days – even if some of the decoder units disappear for a few days ( Heliot et al , 2010; Eleryan et al , 2014). A unit’s contribution to the decode can be quantified in terms of its output z i ( t ).…”

Neuroplasticity subserving the operation of brain–machine interfaces

“…Using recordings from ‘stable’ populations of M1 neurons [39], units were divided into a number of non-overlapping clusters of functionally connected neurons – as measured by spike train correlations during periods of spontaneous activity [72]. An unsupervised, non-biomimetic decoder was built from offline analysis of spontaneous neural activity to enable the monkeys to control multiple DOFs [40].…”

“…As such, a stable transformation not only requires fixing the decoder mapping, but also assigning units to the BMI that have stationary spike waveform shapes that does not alter the day-to-day spike sorting outcome. Only a few neurons (on the order of tens), however, can be stably isolated across a period of days to weeks in typical chronic microelectrode recordings [39, 41, 76, 89-91], although Taylor et al empirically reported maintaining stability of neurons—as determined by their tuning properties to overt movement kinematics—for ~3 years (Taylor, Tillery et al 2002). Consequently, with few exceptions [92], most BMI studies to date rely on identification of new units and decoder recalibration before each session.…”

“…Consequently, with few exceptions [92], most BMI studies to date rely on identification of new units and decoder recalibration before each session. Novel techniques that track units across days could be useful to stabilize the mapping as learning progresses [39, 93]. …”