StarDist

Parser for StarDist segmentation results.

• Refer to API: StarDistSegmentation

Supported Models

• 2D_versatile_fluo: Versatile fluorescence model

• 2D_versatile_he: Versatile H&E staining model

• 2D_paper_dsb2018: DSB 2018 paper model

• 2D_demo: Demo model

Example of loading stardist segmentation results

from neuralib.stardist import read_stardist

# Load stardist segmentation result (.npz file)
seg = read_stardist("/path/to/stardist_seg.npz")

# Get number of segmented cells
print(f"Number of cells: {seg.n_segmentation}")

# Get image dimensions
print(f"Image size: {seg.width} x {seg.height}")

# Access labeled image (NaN = background)
labels = seg.labels  # Array[float, [H, W]]

# Access polygon coordinates for each cell
cords = seg.cords  # Array[float, [N, 2, E]]

# Get detection probabilities
probs = seg.prob  # Array[float, N]

# Get cell center coordinates
centers = seg.points  # Array[float, [N, 2]]

# Get model used for segmentation
print(f"Model: {seg.model}")

Filter by probability threshold

# Mask out low-confidence detections
seg.mask_probability(threshold=0.5)

Convert to ImageJ ROI

from neuralib.stardist import stardist_point_roi_helper

# Convert segmentation to ImageJ point ROI
stardist_point_roi_helper(
    "/path/to/stardist_seg.npz",
    "/path/to/output.roi"
)

Save as compressed NPZ

# Save segmentation results
seg.to_npz("/path/to/output.npz")