Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 10 (4), 046016

A Carbon-Fiber Electrode Array for Long-Term Neural Recording


A Carbon-Fiber Electrode Array for Long-Term Neural Recording

Grigori Guitchounts et al. J Neural Eng.


Objective: Chronic neural recording in behaving animals is an essential method for studies of neural circuit function. However, stable recordings from small, densely packed neurons remains challenging, particularly over time-scales relevant for learning.

Approach: We describe an assembly method for a 16-channel electrode array consisting of carbon fibers (<5 µm diameter) individually insulated with Parylene-C and fire-sharpened. The diameter of the array is approximately 26 µm along the full extent of the implant.

Main results: Carbon fiber arrays were tested in HVC (used as a proper name), a song motor nucleus, of singing zebra finches where individual neurons discharge with temporally precise patterns. Previous reports of activity in this population of neurons have required the use of high impedance electrodes on movable microdrives. Here, the carbon fiber electrodes provided stable multi-unit recordings over time-scales of months. Spike-sorting indicated that the multi-unit signals were dominated by one, or a small number of cells. Stable firing patterns during singing confirmed the stability of these clusters over time-scales of months. In addition, from a total of 10 surgeries, 16 projection neurons were found. This cell type is characterized by sparse stereotyped firing patterns, providing unambiguous confirmation of single cell recordings.

Significance: Carbon fiber electrode bundles may provide a scalable solution for long-term neural recordings of densely packed neurons.

Conflict of interest statement

Disclosures: We declare that we have no conflicts of interest, financial or otherwise.


Figure 1
Figure 1. Array assembly
A, Left, Diagram of the 3D-printed plastic block with wells for 16 carbon fibers. Fibers are threaded through the wells on the top of the block and exit through the hole on the bottom. Middle, To expose the connector-side ends of the fibers from parylene, the fibers are heated by passing them through a gas/oxygen torch. Right, The wells are then filled with conductive silver paint to make electric contact with an Omnetics connector, which slides into the plastic block wells. B, Fire-sharpening of electrode tips. Left, the assembled array is lowered into a water bath with the tips of the carbon fibers protruding above the surface of the water. Bottom, SEM image of a a blunt cut carbon fiber electrode, with insulation frayed near the tip. Middle, A gas/oxygen torch is passed over the surface of the water, burning the carbon and the insulating parylene down to the water surface. Bottom, After passing the torch over the exposed tips, the carbon fiber tapers to a sharp point. Right, The array is then taken out of the water, with the tips pointing down; surface tension acts to bring the carbon fibers into a single tight bundle. C, The assembled array. The 16-electrode bundle is approximately 26 μm in diameter (upper right). The immersion tip burning process exposes approximately 89 μm of carbon (lower right).
Figure 2
Figure 2. Electrode impedances
A, Histogram of the fire-sharpened pre-implant electrode impedance (n = 210 fibers; median = 1.0MΩ). B, Impedance of fibers in 7 implanted arrays measured at various time points after implanting. Colors indicate fibers' grouping into arrays. The pre-implant impedances (in saline) in corresponding colors are shown at Day 0.
Figure 3
Figure 3. Single unit recording in singing bird
Example of a putative interneuron recorded in the pre-motor nucleus HVC aligned to song. Top, the time frequency histogram of aligned renditions of the same song motif. Middle and Bottom, spike raster from a single unit aligned to song and a raw trace from the same channel.
Figure 4
Figure 4. Chronic recording stability in the singing bird
Top, A putative HVC interneuron (single unit by the standard criterion) in bird recorded over 15 days. Bottom, raw traces from the same channel on Days 1, 7 and 14. Signal fading (as on day 14) indicates periods of partial loss of cell isolation.
Figure 5
Figure 5. A principal cell recorded in HVC
Top, Song-aligned spike raster of a putative RA-projecting neuron. Bottom, The raw voltage trace from rendition 199 out of 500 song renditions recorded across 1 hour and 26 minutes. Insets show the average waveform with SD and ISI distribution.
Figure 6
Figure 6. Simultaneously recorded activity
Signal from 15 simultaneously recorded channels during singing (one channel was used for the microphone trace). Fourth trace from the top is the reference electrode. (The reference electrode is always one of the carbon fibers in the bundle.) These signals were not sortable by our criteria.
Figure 7
Figure 7. Example stability of spike features and firing pattern in rigorous single units
Left, The total elapsed time from the first trial (top), peak amplitude (second from top), spike width (second from bottom, in samples at 200 kHz) and root-mean-square error of the average instantaneous firing rate estimated in a sliding 25 trial window (bottom, see Methods) are shown across trials. Right, trial-averaged spike waveform. Colors match the elapsed time shown on the left. In the top example, both spike features and the firing pattern sharply change on the same trial. The bottom example demonstrates the utility of a stable firing pattern (see Figure S10 for the spike raster). Though the spike features drift from trial to trial, the firing pattern remains stable, allowing for reliable unit identification through continuous changes in the waveform. The firing variability measure used here is sensitive to variability in song duration, and the 15% change seen over the course of the day is accurate, and not the result of imprecision in spike identification.
Figure 8
Figure 8. Stability of sorted multi-units and single-units
Top, Stability of sorted multi-units. These points represent single units by standard criteria. The 107 day example from Figure S9 is excluded from this plot. Bottom, Stability of rigorous single-units, which were isolatable based on threshold alone.

Similar articles

See all similar articles

Cited by 45 PubMed Central articles

See all "Cited by" articles

Publication types