spikesift 0.1.1

Real-time spike sorting for high-density extracellular recordings

SpikeSift

SpikeSift is a fast, drift-resilient spike sorting algorithm for high-density extracellular recordings. It delivers accurate, real-time spike sorting from raw binary data using only a single CPU core.

---

Features

• Real-time performance on thousands of channels
• Drift-aware segmentation and robust merging
• Parallelizable across segments
• Modular design - sort, merge, split, and compare segments
• Minimal parameter tuning, even for short recordings

---

Installation

Install with pip:

pip install spikesift

Or install from source:

git clone https://github.com/vasilisgeorgiadis/spikesift.git
cd spikesift
pip install -e .

---

Quickstart

from spikesift import Recording, perform_spike_sorting

# Define probe layout (example)
import numpy as np
probe = np.load("probe.npy")

# Load raw data
recording = Recording(
    binary_file="recording.bin",
    data_type="int16",
    probe_geometry=probe,
    sampling_frequency=30000
)

# Run sorting
result = perform_spike_sorting(recording)

# Access spike times
for cid in result.cluster_ids():
    spikes = result.spikes(cid)
    print(f"Cluster {cid}: {len(spikes)} spikes")

For more examples, see the User Guide or Example Usage.

---

Documentation

Full documentation is available at:

https://spikesift.readthedocs.io/en/latest/index.html

---

Performance

SpikeSift is over 20x faster than GPU-based sorters like Kilosort, and up to 300x faster when all run on a single CPU core. It handles thousands of channels, fragmented recordings, and real-time pipelines with ease.

---

Citing SpikeSift

(Preprint coming soon)

---

License

MIT (c) 2025 Vasileios Georgiadis