Keywords:

Abstract

  • Paper introduces OpenMEA, an open source bioelectronic neural interfacing platform.
  • Advancements in ML and low-power ICs mean closed-loop stimulation is a viable strategy now.
  • Two key questions for meaningful breakthroughs:
    • What markers and algorithms are best suited for detecting biological states?
    • What types of electrical stimuli are optimal for controlling these states?
  • OpenMEA uses a multielectrode array system that interfaces with biological tissue. The implications of this:
    • Reduced experimental complexity
    • Better reproducibility
    • Fewer confounding variables: less unknown/uncontrolled variables?
  • Existing MEA systems have many problems: functional limitations and closed-source designs.

Conclusion

  • Benefits of OpenMEA:
    • Accessible manufacturing, improving equity and access to bioelectronic research and development tools
  • Future improvements: disease-on-chip models as a possible application of OpenMEA.

Discussion

  • Factors that enable OpenMEA
    • Not too important, basically covers technological changes that let OpenMEA fill a cost-conscious niche that is actually possible (PCB manufacturing), etc.
  • Translational device development
    • OpenMEA also integrates hardware, software, and neuroscience in a way that was previously lacking.
    • Fundamental problem with closed-loop algorithms: offline datasets are used to test the algorithm’s performance. This doesn’t effect causal effects of how a detection-triggered stimulus could change proceeding states.
      • Thus, quantifying performance must be done with biological tissue.
      • Animal models are used because manufacturing ASICs and getting regulatory approval for humans is expensive and time-costly.
    • From a hardware POV, closed-loop devices have a few key digital ICs to process the neural signal.
      • This logic needs to be synthesised from an HDL in software.
      • These are then connected to an analogue IC that serves as a front-end sampling neural signals at the implanted electrode site. It also delivers the electrical charge needed for neurostimulation.
      • AFE ICs are able to address some practical challenges (like simultaneous recording and stimulation), but have a key bottleneck in digital control for classifying neural signals into representative brain states and generating an appropriate neurostimulation waveform in response.
      • Some digital IC architectures have been proposed but fail in an online closed-loop setting due to challenges of in vivo testing — OpenMEA can overcome this challenge.
      • OpenMEA uses a modular PCB with an FPGA prototyping device in the loop. This FPGA can be used to implement the low-power IC for in-vitro applications without needing an ASIC.
  • Towards adaptive closed-loop stimulation
    • Relatively little focus has been placed on stimulus adaptation compared to adapting to brain states.
    • The “bi-phasic rectangular waveform” is usually used because it’s “charge-balanced” and easy to generate with a simple current source.
    • More complex waveforms can provide more precise control of neural activity.
    • What’s the challenge? Stimulators must generate waveforms that don’t cause damages.
      • These damages are caused by imbalanced charges during positive/negative stimulation phases.
      • Choosing appropriate waveform parameters for a desired response is a difficult and infeasible computational task.
    • MEAs allow for controlled in vitro exploration of what waveform parameters should be selected.
      • OpenMEA allows for user-programmable charge-balanced stimulus waveform generators in either hardware or software.

OpenMEA

  • Passive MEA
  • Uses a Xilinx Zinq Z-7030 for closed-loop processing, is user programmable