Neuron_Tracking

This repository has been moved to: https://github.com/janelia-TDHarrisLab/Yuan-Neuron_Tracking, please refer to it for the latest updates

This repository contains the code used for the manuscript: "Multi-day Neuron Tracking in High Density Electrophysiology Recordings using EMD". (doi: https://doi.org/10.1101/2023.08.03.551724)

All features are implemented in MATLAB R2023b.

Required packages:

• npy-matlab(https://github.com/kwikteam/npy-matlab): need for reading .npy files
• C_Waves (https://billkarsh.github.io/SpikeGLX/#post-processing-tools): to generate mean waveforms
• ecephys(https://github.com/jenniferColonell/ecephys_spike_sorting): for KS2.5 post-processing and metrics

Directories:

• Pipeline: code for tracking neurons
• Example: example code with 5 datasets from animal AL032
• Figure: code for reproducing figures in the manuscript

1. Pipeline

• Function 1 - NT_main.m
  Goal: Match units between 2 datasets.
  Input: Kilosort cluster label(only include clusters in the region of interest), channel map, mean waveforms (created with C_Waves or other tool, from the KS2.5 drift corrected data)
  Output: Unit match assignment

• Function 2 - chain_summary.mat
  Goal: Find chains using output from all datasets.
  Input: Unit assignment from multiple datasets
  Output: Summary of chains within distance threshold

• Function 3 - chain_stats.mat
  Goal: Summarize L2, FR, and locations of all fully tracked chains.
  Input: Summary of chains Output: Plots(waveforms, firing rate, xz locations) of selected chains

• Additional Functions:

  • ks2_working_toBinary: Converts the KS2.5 temp_wh.dat file, which has been drift corrected, to a standard, correctly scaled binary to be input to C_Waves
  • dist_mat.m: A function used to generate a 'distance matrix' that gives an idea of whether to exclude certain datasets with poor trackability. It is more meaningful to run on datasets with many comparisons.
  • acc.m: A function used to compute recovery rate(without threshold) and accuracy(with threshold). Only use if there are validation information. Will use a matrix with correct-match cluster labels as input. The directory 'supplementary' contains one example validation table 'truth.mat'(from animal AL032 shank 1 dataset 1 and 2).

• Documentation:
  output.all_results (column code) = [reference(0 for datasets without validation info), d2 clu label, d1 clu label, EMD distance, location distance, waveform distance, vertical distance] output.results_wth = unit assignment with threshold applied

2. Example

Function - Example_run.m
Goal: Demonstrate an example with 5 datasets Input: Animal AL032 shank 1 dataset 1 to 5
Output:

• result_chain.mat - A matlab file contains the summary of chains within distance threshold
• chain_stats.mat - A matlab file contains the summary of L2 weight, firing rate, and xz locations of fully tracked chains
• Plot waveforms, firing rate, and xz locations of a chain of choice

3. Figure - code for reproducing figures