tracking-graph 0.9.0

Algorithm to track clusters across recordings

Tracking_Graph

Tracking_Graph is a tool for track units across spike sorting solutions. This package is part of the PhD This project is part of my PhD research at [University Name]

Installation

Install Tracking_Graph using pip:

pip install tracking_graph

Usage

Creating a Graph

from tracking_graph import run_tg, get_tg_groups, EuclideanClassifier

# Create a classifier
modelcreator = EuclideanClassifier.creator(std_mult=3)  # Classify spikes with length > 3 std

# Run Tracking_Graph
G = run_tg(
    we_list,  # List of WaveformExtractor objects
    outputfile='/home/user/examplefolder/tg_data.hdf5',
    max_len=2,  # Maximum edge length, must be at least 1
    modelcreator=modelcreator
)

# Compute results programmatically (can be replaced by the GUI)
groups, sG, discarded = get_tg_groups(
    G,
    mintrack=3,  # Minimum number of segments for a cluster
    merge=True  # Apply criteria to merge splits
)

# Create a final results table
import pandas as pd

df = []
for gi, g in enumerate(groups):
    for c in g:
        df.append({'segment': c.segment,
                    'cluster': c.unit,
                    'tg_unit': gi})

results_table = pd.DataFrame(df)

Exploring Simplified Graphs with GUI

Launch the graphical interface (Streamlit server) using:

tg_gui

Additional Tools

Tracking_Graph provides a wrapper to load aligned waveforms from Wave_Clus clustering results, addressing limitations in SpikeInterface's waveform interpolation:

from tracking_graph.spikeinterface_addons import Waveclus_Waveforms

path_times_file = '/home/user/examplefolder/times_example.mat'  # Full path to Wave_Clus result
we = Waveclus_Waveforms(path_times_file)  # Object with basic WaveformExtractor interface

Limitations

• Tracking_Graph requires spikeinterface<=0.100 due to its dependency on the WaveformExtractor class.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

About

For more projects and information, visit my GitHub profile: ferchaure