Glasser-Tian fMRI Atlas

A robust, memory-efficient fMRI parcellation tool using:

• Glasser et al. (2016): 360 cortical regions
• Tian et al. (2020): 54 subcortical regions
  → Total: 414 ROIs

GT-MAP enables reproducible time series extraction across heterogeneous datasets—without discarding subjects due to non-standard TR or short scan durations.
Atlases are bundled with the package, so no external downloads are required.

---

Features

• TR-flexible: Accepts any repetition time (TR)
• Resamples first, standardizes after: Methodologically correct; preserves signal integrity
• No temporal filtering: Avoids padlen errors on short or degenerate runs
• Parallel batch processing: Scales across subjects with joblib
• Memory-safe: Uses nilearn caching; no full-dataset loading
• Robust validation: Automatically skips scans shorter than target duration (no crash)

Designed for ABIDE, UK Biobank, ADHD-200, and other multi-site fMRI studies.

---

📦 Installation

Install directly from GitHub:

pip install git+https://github.com/cimt-unia/gt_map.git

Requirements: Python ≥3.9, nilearn≥0.9, numpy, pandas, scipy, joblib

---

Usage

Example

from gt_map import GlasserTianParcellator

# No atlas_dir needed — uses built-in atlases
parcellator = GlasserTianParcellator()

ts = parcellator.parcellate_subject(
    fmri_path="sub-01_task-rest_bold.nii.gz",
    tr=2.5,                     # actual TR in seconds
    target_tr=2.0,              # desired TR
    target_duration=300.0       # 5 minutes → 150 timepoints
)
# ts.shape → (150, 414)

---

Details

The package includes the following files in MNI152NLin6Asym space:

• glasser_360_MNI152NLin6Asym.nii.gz (cortical, 360 ROIs)
• tian_subcortex_54_MNI152NLin6Asym.nii (subcortical, 54 ROIs)
• roi_labels.csv (414 rows, column: roi_name)

These are derived from publicly available sources (see Acknowledgments).
The bundled data enables zero-configuration usage while ensuring reproducibility.

---

References

• Glasser et al. (2016)
  A multi-modal parcellation of human cerebral cortex
  Nature 536, 171–178. https://doi.org/10.1038/nature18933

• Tian et al. (2020)
  Topographic organization of the human subcortex unveiled with functional connectivity gradients
  Nature Neuroscience 23, 1421–1432. https://doi.org/10.1038/s41593-020-00711-6
  Atlas files available at: https://github.com/yetianmed/subcortex

• Nilearn team for robust, open-source neuroimaging utilities
  https://nilearn.github.io/