Core Concepts
Stimulation timing can be used to encode information into in-vitro neural networks, with optimal delays between 36 and 436 ms, and different combinations of readout parameters may be optimal at different parts of the evoked spike response.
Abstract
The study explores using stimulation timing as an encoding method for in-vitro neural networks on micro electrode arrays (MEAs). The goal is to identify the bounds and acuity of stimulation timings that produce linearly separable spike responses, as well as the optimal readout parameters for a linear decoder.
The key findings are:
Network Stability:
Spiking and bursting activity fluctuated over the networks' development, often coinciding with media exchange days.
Evoked responses also differed significantly between experimental days for many conditions, indicating potential instability issues.
Decoder-Readout and Information Encoding:
The network may use both spike-time and rate-based encoding, which require different readout parameters to capture.
Exploring epoch length, time bin size, and epoch offset revealed that optimal settings can vary for different parts of the evoked response.
Upper Separability Bound:
Networks showed mixed performance, but memory lengths could exceed 7 seconds for some conditions and networks.
Accuracy tended to stay stable for a period before declining, with the length of stability varying.
Lower Separability Bound:
At the smallest probe delay of 0 ms, accuracy was up to 75%, suggesting the probe can still affect the network's trajectory.
Accuracy decreased for some networks at intermediate probe delays, indicating potential network saturation effects.
Probe-Probe Acuity:
Stimulation timings between 36 and 436 ms may be optimal for encoding, with higher acuity in the 200-400 ms range.
Accuracy varied across networks, with some performing well and others poorly, indicating network-specific differences.
Overall, the results suggest that stimulation timing can be a viable encoding method for in-vitro neural networks, with optimal delays and readout parameters depending on the specific network and part of the evoked response.
Stats
The number of spikes evoked by each experimental condition was significantly different between days for many networks and conditions.
Quotes
"Stimulation timings between 36 and 436ms may be optimal for encoding and that different combinations of readout parameters may be optimal at different parts of the evoked spike response."
"At the smallest probe delay of 0 ms, A2 had the highest accuracy at 75 % on day 1, this was however decreased to 64 % for the 100 ms probe."