Back

Lussac: a fully-automated consensus method that increasesthe yield and quality of spike-sorting analyses

Wyngaard, A. J. G.; Llobet, V.; Barbour, B.

2025-04-28 neuroscience
10.1101/2022.02.08.479192 bioRxiv
Show abstract

The rise in site counts of multi-electrodes used in extracellular recordings in the brain has driven the development of increasingly automated spike-sorting packages. However, post-processing is still largely manual and it remains difficult to determine the optimal package and parameters for a given recording. It has recently been shown that different packages produce quite disparate outputs, suggesting that a combination of analyses might be beneficial. Here, we describe the formalization of existing and new metrics of unit quality and comparison, then build upon these to automate the creation of a consensus output from multiple analyses. We validated our package against synthetic and real ground truths. Compared to individual analyses, our package increased the yield and quality of correct units (doubling the yield of Purkinje cells in our recordings) and, crucially, eliminated numerous incorrect units that were impossible to identify in a single analysis. These improvements also increase analytical objectivity and reduce manual effort.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.