A point-process matched filter for event detection and decoding from population spike trains.

Area of Science:

• Computational Neuroscience
• Neural Decoding
• Spike Train Analysis

Background:

• Neuronal information encoding is often described by spatio-temporal response fields (STRFs).
• Point process models can describe neuronal activity, but decoding is challenging when event onset times are unknown.
• Existing neural classifiers require behavior- or stimulus-aligned data, limiting real-world applications.

Purpose of the Study:

• To develop a decoder for neurons with STRFs when event times are unknown.
• To enable accurate detection and decoding of neural activity without prior knowledge of event onsets.
• To advance neurotechnologies for decoding cognitive states from neural responses.

Main Methods:

• Developed a novel point-process matched filter (PPMF) to detect events and estimate onset times from population spike trains.
• Devised a point process filter (PPF) for neurons with transient responses characterized by STRFs.
• Combined PPMF with PPF or discrete classifiers for continuous and discrete brain state decoding, respectively.

Main Results:

• Validated the PPMF on simulated and real-world neural data, accurately estimating stimulus and saccade times.
• Demonstrated successful decoding of visual saliency maps without knowing stimulus times.
• Showcased decoding of saccade direction without knowing saccade times using PPMF and a point process classifier.

Conclusions:

• The developed event detection and decoding algorithms are effective for neurons with STRFs and unknown event times.
• These algorithms facilitate the development of neurotechnologies for decoding cognitive states.
• This work overcomes a significant limitation in decoding neuronal activity for real-world applications.