DetectSpots

DetectSpots detects puncta as local maxima in a scored 2D intensity image. Supported methods are Difference of Gaussians (dog), Laplacian of Gaussian (log), and direct local_maxima.

Inputs are input_image, method, sigma, sigma_ratio, threshold, and min_distance. Outputs are output_labels, spot_id, y, x, intensity, score, and spot_count. BioImageFlow records one dataframe row per detected spot, so no separate spot CSV artifact is written.

Use it as a deterministic baseline for puncta workflows. It currently supports 2D scalar images only. Invalid methods, non-2D images, missing files, or overly high thresholds are expected failure or empty-result modes.

Dependencies and Core Libraries

BioImageFlow core APIs, imageio, NumPy, and optionally SciPy for gaussian_laplace.

Assumptions

Puncta are local maxima in a 2D scalar image, and threshold is expressed in the filtered score image units.

Minimal Example

from bioimageflow_core import Arguments
from bioimageflow_spot_tools import DetectSpots

DetectSpots().process_row(
    Arguments(input_image="puncta.tif", method="dog", threshold=0.3)
)

Expected Results

The label image marks one pixel per detected spot, and the workflow output dataframe contains one row per accepted local maximum. The label image is written as uint32; background is 0, and detected spots use positive sequential IDs. When no spots pass the threshold, the dataframe is empty and the blank uint32 label image is still published as the node’s output_labels artifact.

Failure Modes

Invalid methods raise ValueError; non-2D inputs raise ValueError; missing files fail through imageio. Producing more labels than uint32 can store raises ValueError.