Finding Spots with TrackpyLocalMaxPeakFinder

RNA FISH spots are usually bright point spread functions in a greyscale image. Rolonies, which are rolling-circle amplicons produced in certain assays (e.g. in situ sequencing), are approximately 1-um diameter Gaussian spots. Generally, the recommended FindSpotsAlgorithm to use in a starfish pipeline is BlobDetector because it accounts for the intensity profile of a spot rather than just thresholding pixel values. But for some images BlobDetector may not be satisfactory so starfish also provides alternatives.

TrackpyLocalMaxPeakFinder finds Gaussian spots by an implementation of the Crocker-Grier centroid-finding algorithm. Local maxima are treated as centroids and then the locations are refined to obtain sub-pixel accuracy. Unlike LocalMaxPeakFinder, TrackpyLocalMaxPeakFinder measures spot attributes like size and eccentricity but it also requires the user to manually set a minimum intensity threshold.

Before running TrackpyLocalMaxPeakFinder, the image must be preprocessed to smooth noise and remove background. This can be done by setting preprocess=True or by running Bandpass filter separately. If done in TrackpyLocalMaxPeakFinder, noise_size defines the width of the Gaussian blurring kernel for removing high frequency noise and smoothing_size defines the width of boxcar smoothing kernel. There is no formula for picking noise_size, but larger values will lead to more blurring and it should be less than smoothing_size. And smoothing_size should generally be the diameter of a spot rounded up to the nearest odd integer. The background should be set to zero with a threshold set in Bandpass or by ref:clipping background to zero <howto_clip_percentile_to_zero>.

For finding spots, spot_diameter is used to identify all possible spots and should be set to the diameter of the spots on zero background rounded up to the nearest odd integer. Then thresholds such as min_mass (integrated brightness), max_size (radius of gyration), percentile (relative peak brightness), and separation (distance between spots) are used to filter out spots. Plotting these attributes can help ballpark cutoff values but visually assessing results is needed to validate the parameter settings before running in batch.


TrackpyLocalMaxPeakFinder does not support finding spots on independent 2D slices of a volume (i.e., is_volume = False).


TrackpyLocalMaxPeakFinder is not compatible with cropped data sets.

from starfish import data
from starfish import FieldOfView
from starfish.image import Filter
from starfish.spots import FindSpots

experiment = data.allen_smFISH(use_test_data=True)
img = experiment['fov_001'].get_image(FieldOfView.PRIMARY_IMAGES)

# filter to remove noise, remove background, blur, and clip
bandpass = Filter.Bandpass(lshort=.5, llong=7, threshold=0.0)
glp = Filter.GaussianLowPass(
    sigma=(1, 0, 0),
clip1 = Filter.Clip(p_min=50, p_max=100)
clip2 = Filter.Clip(p_min=99, p_max=100, is_volume=True), in_place=True), in_place=True), in_place=True), in_place=True)

tlmpf = FindSpots.TrackpyLocalMaxPeakFinder(
    spot_diameter=5,  # must be odd integer
    max_size=2,  # this is max radius
    percentile=10,  # this has no effect when min_mass, spot_diameter, and max_size are set properly
spots =

Gallery generated by Sphinx-Gallery