Back

Optimized Neuron Tracing Using Post Hoc Reanalysis

Azzouz, S.; Walker, L. A.; Doerner, A.; Geisel, K.; Rodriguez-Rivera, A. K.; Li, Y.; Roossien, D. H.; Cai, D.

2022-10-13 neuroscience
10.1101/2022.10.10.511642 bioRxiv
Show abstract

Over the last decade, the advances in Brainbow labeling allowed labeling hundreds of neurons with distinct colors in the same field of view of a brain [1, 2]. Reconstruction (or "tracing") of the 3D structures of these images has been enabled by a growing set of software tools for automatic and manual annotation. It is common, however, to have errors introduced by heuristics used by tracing software, namely that they assume the "best" path is the highest intensity one, a more pertinent issue when dealing with multicolor microscope images. Here, we report nCorrect, an algorithm for correcting this error by reanalyzing previously created neuron traces to produce more physiologically-relevant ones. Specifically, we use a four dimensional minimization algorithm to identify a more-optimal reconstruction of the image, allowing us to better take advantage of existing manual tracing results. We define a new metric (hyperspectral cosine similarity) for describing the similarity of different neuron colors to each other. Our code is available in an open source license and forms the basis for future improved neuron tracing software.

Matching journals

The top 6 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.