MOD: A novel machine-learning optimal-filtering method for accurate and efficient detection of subthreshold synaptic...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Understanding neural information coding requires analyzing subthreshold synaptic events and action potentials (APs).
• Detecting excitatory postsynaptic potentials (EPSPs) or currents (EPSCs) in vivo is challenging due to poor signal-to-noise ratio and variable event characteristics.

Purpose of the Study:

• To develop a robust method for detecting synaptic events in neural recordings.
• To improve the analysis of neural activity during different behavioral states.

Main Methods:

• Developed MOD (Machine-learning Optimal-filtering Detection-procedure), combining supervised machine learning and Wiener filtering.
• Trained the algorithm using expert-scored data to optimize filter coefficients and detection thresholds.
• Validated MOD on in vivo (mouse spatial navigation) and in vitro (enhanced transmitter release) data.

Main Results:

• MOD achieved high detection accuracy, with an average Area Under the Curve (AUC) of 0.894 for in vivo and 0.969 for in vitro data.
• Outperformed existing methods (template-fit, deconvolution, Bayesian) by an average factor of 3.13 for in vivo data.
• Demonstrated comparable or higher computational efficacy than previous methods.

Conclusions:

• MOD offers a significant advancement for synaptic event detection.
• This method facilitates large-scale, real-time analysis of synaptic activity.
• MOD is a valuable new tool for neuroscience research.