Metadata-Version: 2.4
Name: spikesift
Version: 0.1.1
Summary: Real-time spike sorting for high-density extracellular recordings
Home-page: https://github.com/vasilisgeorgiadis/spikesift
Author: Vasileios Georgiadis
License: MIT
Classifier: Development Status :: 4 - Beta
Classifier: Intended Audience :: Science/Research
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Cython
Classifier: Operating System :: OS Independent
Requires-Python: >=3.7
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: numpy
Requires-Dist: scipy
Requires-Dist: tqdm
Dynamic: license-file

# 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:

```bash
pip install spikesift
```

Or install from source:

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

---

## Quickstart

```python
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](https://spikesift.readthedocs.io/en/latest/user_guide.html) or [Example Usage](https://spikesift.readthedocs.io/en/latest/example_usage.html).

---

## Documentation

Full documentation is available at:

[https://spikesift.readthedocs.io/en/latest/index.html](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](LICENSE) (c) 2025 Vasileios Georgiadis
