Segmentation (segmentation_para)

Non-Specific Parameters (nonspecific_para)

  • nuc_radius (integer) — Average radius (in pixels) of a nucleus.

    This parameter can be estimated by the area of a typical cell nucleus.

  • allowed_nuc_size (1x2 integer array) — Acceptable areas (in pixels) of a nucleus.

    Lower and upper limits. Mask components outside of this range will be removed.

  • allowed_ellipse_size (1x2 integer array) — Acceptable areas (in pixels) of an ellipse.

    Lower and upper limits. Ellipses outside of this range will be removed.

  • max_ellipse_aspect_ratio (double | advanced parameter) — Maximal aspect ratio of an ellipse.

    An aspect ratio is defined as the ratio of the major axis to the minor axis. Must be greater than 1. Ellipses with greater aspect ratios will be removed.

  • max_hole_size_to_fill (integer | advanced parameter) — Maximal area (in pixels) of a hole to fill.

    In masks, a hole is defined as a set of background pixels surrounded by foreground pixels. Holes with smaller areas will be converted to foreground pixels.

  • blur_radius (integer | advanced parameter) — Radius (in pixels) of the disk for image smoothing.

Image Binarization (image_binarization_para)

  • if_log (binary) — Whether to log-transform the images.

    Value Suggested Situation
    1 Nuclei have heterogeneous brightness.
    0 Nuclei have homogeneous brightness.

  • background_subtraction_method (string) — Method of Background Subtraction.

    Suggested for images with bright backgrounds. If running, every image is first background-subtracted and then used for Segmentation.

    Option Description
    ‘none’ No background subtraction is performed.
    ‘min’ Minimal intensity of the image background is subtracted from the image.
    ‘median’ Median intensity of the image background is subtracted from the image.
    ‘mean’ Mean intensity of the image background is subtracted from the image.

  • binarization_method (string) — Method of Image Binarization.

    Option Description Suggested Situation
    ‘threshold’

    Thresholding. A threshold is applied

    to the image intensities.

    Nuclei have homogeneous

    brightness.
    ‘blob’

    Blob Detection. A threshold is applied

    to the hessian of image intensities.

    Nuclei have heterogeneous

    brightness.

  • blob_threshold (double) — Blob Detection only. Threshold of the hessian.

Active Contour (active_contour_para)

  • if_run (binary) — Whether to run Active Contour.

    Value Suggested Situation
    1 Accurate nuclear boundaries have not been detected by Image Binarization.
    0 Accurate nuclear boundaries have been detected by Image Binarization.

    Setting the following parameters only if running Active Contour.

  • if_log (binary) — Whether to log-transform the images.

    Value Suggested Situation
    1 Nuclei have heterogeneous brightness.
    0 Nuclei have homogeneous brightness.

  • active_contour_method (string) — Method of Active Contour

    Option Description Suggested Situation
    ‘local’

    Applied to the neighborhood

    of every nucleus.

    		 Nuclei have heterogeneous brightness.
    ‘global’ Applied to the entire image. Nuclei have homogeneous brightness.

Watershed (watershed_para)

  • if_run (binary) — Whether to run Watershed.

    Value Suggested Situation
    1 Nuclei overlap frequently.
    0 Nuclei do not overlap frequently.

Ellipse Fitting (ellipse_para)

The following descriptions are adapted from Zafari et al. 2015. All parameters are advanced. It is often unnecessary to modify their values.

  • k (integer | advanced parameter) — Consider up to k-th adjacent points to the corner point.

  • thd1 (integer | advanced parameter) — Distance (in pixels) between the ellipse centroid of the combined contour segments and the ellipse fitted to each segment.

  • thd2 (integer | advanced parameter) — Distance (in pixels) between the centroids of ellipse fitted to each segment.

  • thdn (integer | advanced parameter) — Distance (in pixels) between contour center points.

  • C (double | advanced parameter) — Minimal aspect ratio for corner detection.

    Must be greater than 1.

  • T_angle (integer | advanced parameter) — Maximal angle (in degrees) of a corner.

  • sig (integer | advanced parameter) — Standard deviation (in pixels) of the Gaussian filter.

  • Endpoint (binary | advanced parameter) — Whether to add the end points of a curve as corner.

  • Gap_size (integer | advanced parameter) — Maximal length of gaps (in pixels) in the contours to fill.

Correction with Training Data (seg_correction_para)

  • if_run (binary) — Whether to run Correction with Training Data.

    Value Suggested Situation
    1

    Training datasets are available and well-predict the

    number of nuclei in each ellipse.
    0 Training datasets are not available or not suitable.

  • min_corr_prob (double | advanced parameter) — Minimal probability (0 to 1) for correction.