Localization of neurons from extracellular footprints

Area of Science:

• Neuroscience
• Computational Neuroscience
• Bioengineering

Background:

• High-density microelectrode arrays (HD-MEAs) enable simultaneous recording of hundreds of neurons.
• Extracellular recordings necessitate algorithms for estimating neuron positions and tracking movements.

Purpose of the Study:

• Evaluate and compare the performance of neuron localization algorithms using MEA recordings.
• Provide guidance for researchers on selecting optimal neuron localization methods.

Main Methods:

• Compared center-of-mass, monopolar, and grid-based algorithms for neuron localization.
• Utilized artificial and mouse retina MEA recordings for algorithm evaluation.
• Assessed algorithms based on accuracy, variability, and computational cost.

Main Results:

• The grid-based algorithm demonstrated the most satisfactory outcomes for neuron localization.
• Center-of-mass method had minimal computational cost but suboptimal localization accuracy.
• Monopolar algorithms achieved <10μm error but exhibited high variability and computational cost.

Conclusions:

• The grid-based method offers a balance of satisfactory performance, low variability, and low computational cost.
• Previous benchmarking of localization methods lacked rigorous testing with ground-truth recordings.
• This study provides essential guidance for selecting effective neuron localization techniques.