# RTX-Neuro Python

High-performance MEG/EEG analysis library with Python bindings.

RTX-Neuro provides Rust-powered neuroimaging analysis with a Python-friendly API, offering significant performance improvements over pure Python implementations.

## Installation

### From PyPI (when published)
```bash
pip install rtx-neuro
```

### From Source
```bash
cd crates/specialized/rtx-neuro-python

# Development build (editable)
maturin develop

# Release build
maturin build --release
pip install target/wheels/rtx_neuro-*.whl
```

## Quick Start

```python
import rtx_neuro as rtx
import numpy as np

# Read MEG/EEG data
raw = rtx.io.read_raw_edf("recording.edf")
print(f"Channels: {raw.n_channels}, Sample rate: {raw.sfreq} Hz")

# Get data as numpy array
data = raw.get_data(tmin=0.0, tmax=10.0)

# Statistical analysis
t_stat, p_value = rtx.stats.ttest_1samp(data[0], popmean=0.0)

# Multiple comparison correction
p_values = np.array([0.01, 0.04, 0.03, 0.20, 0.001])
rejected, corrected = rtx.stats.fdr_correction(p_values, alpha=0.05)
```

## Supported File Formats

| Format | Function | Description |
|--------|----------|-------------|
| EDF/EDF+ | `read_raw_edf()` | European Data Format |
| FIF | `read_raw_fif()` | Elekta/Neuromag MEG |
| CTF | `read_raw_ctf()` | CTF MEG (.ds directories) |
| BTi | `read_raw_bti()` | 4D-Neuroimaging/BTi MEG |
| KIT | `read_raw_kit()` | Yokogawa/KIT/Ricoh MEG |
| EGI | `read_raw_egi()` | EGI (.raw, .mff) |

## Statistical Functions

### Parametric Tests
- `ttest_1samp(data, popmean)` - One-sample t-test
- `ttest_ind(a, b)` - Independent samples t-test (Welch's)
- `ttest_rel(a, b)` - Paired samples t-test

### Non-parametric Tests
- `permutation_t_test(a, b, n_permutations, seed)` - Permutation test

### Multiple Comparison Correction
- `fdr_correction(p_values, alpha)` - Benjamini-Hochberg FDR
- `bonferroni_correction(p_values, alpha)` - Bonferroni correction

## Examples

See the `examples/` directory for complete tutorials:
- `quickstart.py` - Getting started guide
- `reading_data.py` - Reading various file formats
- `statistical_tests.py` - Statistical analysis workflows

## Performance

RTX-Neuro leverages Rust's performance advantages:
- Native compiled code (10-100x faster than pure Python)
- Parallel processing with Rayon
- Memory-efficient data structures
- Zero-copy numpy array transfers where possible

## API Reference

### RawData Object

Properties:
- `sfreq` - Sampling frequency (Hz)
- `n_channels` - Number of channels
- `n_samples` - Number of samples
- `duration` - Recording duration (seconds)
- `ch_names` - List of channel names
- `path` - File path

Methods:
- `get_data(tmin=None, tmax=None)` - Get data as numpy array
- `pick_channels(ch_names)` - Get indices for channel names

## Requirements

- Python >= 3.8
- NumPy >= 1.20

## Building from Source

Requirements:
- Rust >= 1.70
- maturin >= 1.0
- Python development headers

```bash
# Install maturin
pip install maturin

# Build and install
cd crates/specialized/rtx-neuro-python
maturin develop  # Development mode
# or
maturin build --release  # Release wheel
```

## License

MIT OR Apache-2.0