FluoSTA

Structure Tensor Analysis for axon bundle tracking on fluorescence images from cleared brain tissue.

Workflow

1. Compute structure tensor (sta_flow.py)

Estimates local fiber orientation by computing the structure tensor eigenvector field (v3) and a coherence map from a 3D fluorescence image.

python sta_flow.py --input image.nii.gz --brainmask brain.nii.gz --seedmask seed.nii.gz \
    --dog 0.5 --gauss 2.0 --outdir results

Outputs are written to results/dog0.5gau2.0/: v3.nii.gz, coherence.nii.gz, fsl_tensor.nii.gz, bmask.nii.gz, smask.nii.gz.

2. Track streamlines (sta_track.py)

Propagates streamlines from seed voxels along the v3 direction field using Numba-accelerated nearest-neighbor sampling.

python sta_track.py --v3 v3.nii.gz --seeds smask.nii.gz --mask bmask.nii.gz \
    --out streamlines.tck --step 1e-3 --angle 35 --density 0.1

3. Filter streamlines (sta_filter.py)

Prunes streamlines by suffix-mean intensity, coherence, and/or turning angle.

python sta_filter.py --in_tck streamlines.tck --img FA.nii.gz \
    --threshold 0.2 --angle_max 45 --out_tck filtered.tck

Dependencies

numpy nibabel scipy scikit-image dipy numba

Install with:

pip install numpy nibabel scipy scikit-image dipy numba

numba is optional for sta_track.py — tracking will fall back to pure Python if unavailable.

Usage help

Each script prints detailed usage with python <script>.py -h.

---

Tianbo Qi, Scripps Research, 2025
Part of the code was adapted from MIRACL