A matched-filter algorithm to detect amperometric spikes resulting from quantal secretion

• Published in: Journal of neuroscience methods 2018-01, Vol.293, p.338-346
• Main Authors: Balaji Ramachandran, Supriya, Gillis, Kevin D.
• Format: Article
• Language: English
• Subjects: Action Potentials; Adrenal Glands - physiology; Algorithms; Animals; Carbon; Carbon Fiber; Carbon-fiber amperometry; Cattle; Cells, Cultured; Chromaffin Cells - physiology; Exocytosis; False Positive Reactions; Least-Squares Analysis; Mice; Microelectrodes; Neurons - physiology; Pattern Recognition, Automated - methods; Quantal exocytosis; ROC Curve; Signal processing; Signal Processing, Computer-Assisted; Software; Spike detection
• ISSN: 0165-0270; 1872-678X
• DOI: 10.1016/j.jneumeth.2017.10.019

•A “matched filtering” approach for detecting amperometric spikes is proposed.•The approach relies on the use of a library of prototypical spike templates.•Data segments are matched against each template to yield a criterion score.•Spikes are detected when the criterion score exceeds a threshold.•The algorithm outperforms the derivative-threshold detection approach. Electrochemical microelectrodes located immediately adjacent to the cell surface can detect spikes of amperometric current during exocytosis as the transmitter released from a single vesicle is oxidized on the electrode surface. Automated techniques to detect spikes are needed in order to quantify the spike rate as a measure of the rate of exocytosis. We have developed a Matched Filter (MF) detection algorithm that scans the data set with a library of prototype spike templates while performing a least-squares fit to determine the amplitude and standard error. The ratio of the fit amplitude to the standard error constitutes a criterion score that is assigned for each time point and for each template. A spike is detected when the criterion score exceeds a threshold and the highest-scoring template and the time of peak score is identified. The search for the next spike commences only after the score falls below a second, lower threshold to reduce false positives. The approach was extended to detect spikes with double-exponential decays with the sum of two templates. Receiver Operating Characteristic plots (ROCs) demonstrate that the algorithm detects >95% of manually identified spikes with a false-positive rate of ∼2%. ROCs demonstrate that the MF algorithm performs better than algorithms that detect spikes based on a derivative-threshold approach. The MF approach performs well and leads into approaches to identify spike parameters.