> ## Documentation Index
> Fetch the complete documentation index at: https://docs.autocleaneeg.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Derivatives: Stage Folders

> Learn how AutoCleanEEG writes numbered stage folders under derivatives for step-by-step inspection.

## Stage Folders under `derivatives/`

AutoCleanEEG writes **numbered stage folders directly under `bids/derivatives/`**. These are **dynamically generated** — only the steps you enable and configure for export will appear.

For example, if you skip ICA or don’t export filtering outputs, the corresponding step folder (`08_ica/` or `03_filter/`) won’t exist in your run.

### Example Layout

```
derivatives/
├── 01_import
├── 02_resample
├── 03_filter
├── 04_trim
├── 05_bad_channels
├── 06_rereference
├── 07_artifact_detection
├── 08_ica
├── 09_pre_ica
├── 10_epochs
├── 11_drop_bad_epochs
├── 12_epochs
├── 13_gfp_clean
└── 14_comp
```

### How It Works

* **Step numbering** reflects the pipeline order.
* **Exported steps only**: If you request export after filtering but not after trimming, you’ll see `03_filter/` but not `04_trim/`.
* **Flexible granularity**: This lets you save just the checkpoints you care about (for inspection or debugging), while avoiding unnecessary disk usage.

### Step Highlights

| Folder                | Purpose                                                                          |
| --------------------- | -------------------------------------------------------------------------------- |
| **01\_import**        | Original raw data converted into BIDS‑compatible format.                         |
| **02\_resample**      | Data resampled (e.g., 1000 Hz → 250 Hz).                                         |
| **03\_filter**        | Filtered signals (HP, LP, notch).                                                |
| **05\_bad\_channels** | Snapshot of data after noisy channels are dropped.                               |
| **08\_ica**           | Data after ICA decomposition and component removal.                              |
| **10\_epochs**        | Segmented data into epochs.                                                      |
| **13\_gfp\_clean**    | Epochs cleaned with Global Field Power metrics.                                  |
| **14\_comp**          | Final compiled output before writing to `derivatives/` and task‑root `exports/`. |

<Info>
  ### ICA Export Format

  The **ICA stage** is exported in both **EEGLAB (.set)** and **MNE (.fif)** formats.\
  This is a special case:

  * In **EEGLAB**, ICA weights are stored directly inside the `.set` dataset.
  * In **MNE-Python**, ICA weights are stored separately from the EEG data (the `.fif` ICA file).

  This separation allows you to:

  * Reapply ICA weights later to the same dataset.
  * Redo component artifact removal and continue the pipeline from that point (feature coming soon).
</Info>

### Why It Matters

* **Customizable exports**: You decide which steps produce saved files.
* **Debug-friendly**: Inspect any stage of the pipeline without re-running the entire process.
* **Storage-aware**: Avoid large intermediate folders by disabling exports for steps you don’t need.

<Info>
  Stage folders are **optional checkpoints**. You can enable or disable exports per step in your task configuration. Once validated, these folders can be safely deleted to reclaim space.
</Info>