Anglegram analysis: baseline code

This project is the implementation of the anglegram and the detection of junctions in the boundary of a clump. The algorithm detects junctions between the boundaries of overlapping objects based on the angles between points of the overlapping boundary.

The anglegram matrix is a 2D matrix with the multiscale angle variation. The anglegram is used to find junctions of overlapping cells.

Quick start guide

In order to use the code as shown in our MIUA publication, you need to do the following:

1. Get a binary image with your clump or cell, let bw0 be such binary image. With it, get the boundary of such binary object

boundy=bwboundaries(bw0);

2. Compute the anglegram. For a simple test with just one clump bw0 and one boundary boundy, nothing else needs to be done:

[anglegram] = computeAngleMatrix(boundy);

3. Get the junction points (we call them candies) via the anglegram matrix. You may need to tweak some of the parameters, but the default values are a good starting point.

[candies, candyhandles] = computeCandidatePoints(anglemgram, boundy);

3. (optional) If you feel you need to tweak the parameters, you can do it via an options structure opts:

opts.mainthr = 150;
opts.offsetvar = 7;
opts.statsfname = 'max';
[candies, candyhandles] = computeCandidatePoints(anglemgram, boundy, opts);

The structure candyhandles contains a lot of useful information from the junctions detected:

• candieshandles.angleSummaryVector Maximum (mean or integral) intensity projection.
• candieshandles.meanAM Mean of the angleSummaryVector.
• candieshandles.stdAM Standard deviation of the angleSummaryVector.
• candieshandles.intensityLocations Locations in the boundary boundy where the candidates are located.
• candieshandles.intensityPeaks Value of the angles where the locations are presented candieshandles.intensityLocations.

Junction slicing (JS)

To compute the Junction Slicing technique from our publication, the (beta) code following from the previous section.

4. Get cell candidates from the junctions found, using using the binary image of the nuclei, nuclei.

cl = candieshandles.intensityLocations; % candidate locations
[bwcells, boundcells, ~] = getCellCandidatesFromSegments(cl, boundy, bw0, nuclei);

Future developments

A future development of the anglegram matrix involves the function computeMultiAnglegram.

Publication

The development of this algorithm was published in the Medical Image Understanding and Analysis Conference MIUA 2017, and then extended on the Journal of Imaging.

Licensing and Citation

This project is part of the GNU GENERAL PUBLIC LICENSE v3, read all the terms and conditions in the LICENSE file. If this code is useful in your research consider citing:

@article{solislemus-mdpi2017,
	title = {Segmentation and {Shape} {Analysis} of {Macrophages} {Using} {Anglegram} {Analysis}},
	volume = {4},
	copyright = {http://creativecommons.org/licenses/by/3.0/},
	url = {http://www.mdpi.com/2313-433X/4/1/2},
	doi = {10.3390/jimaging4010002},
	language = {en},
	number = {1},
	urldate = {2017-12-22},
	journal = {Journal of Imaging},
	author = {Solís-Lemus, José Alonso and Stramer, Brian and Slabaugh, Greg and Reyes-Aldasoro, Constantino Carlos},
	month = dec,
	year = {2017},
	keywords = {macrophages, overlapping objects, segmentation, shape analysis},
	pages = {2},
}