[DOCS] Read through Actin Filament Network Literature

[Alejandro1400]

Initially we are using a ridge detection method to analyze the actin filament networks in cell images, but it has come to my attention that there is a field in actin fialments which uses Stretching Open Active Contours (SOACs) algorithms to study the filamentous biopolymer networks in cells imaged by confocal microscopy.

This issue is going to serve as a thread about the literature review.

[Alejandro1400]

1. SOAX

SOAX is based under the method Stretching Open Active Contours (SOACs). For an input image multiple SOACs are initiallized. These evolve by moving, stretching, merging and forming junctions with one another. The final network topology is configured by cutting and joining contours such that SOACs do not end or bend sarply at junctions.

The software provides a great way for extracting and quantifying biopolymer networks imaged by confocal microscopy in 3D. It is accompanied by an optimization method for parameter selection.

After an initial try of the software we can detect that it is able to accurately detect what it calls 'snakes' but its quantitative analysis is more focused on cell mechanical properties. It outputs: Curvature, Length, Point density, radial orientation, spherical orientation. Unfortunately since our focus on fluorescent labeling, we are more interested on analyzing things that can give us an insight of degree of labeling, SNR. The outputs CSV isn't easy to post-process due to its output (Every file has different organization and has no snake index so it is hard to understand each filaments' properties)

Still have to check outputing the image if it is possible to analyze it. Not sure how automating the process would work...

• Paper: SOAX: A software for quantification of 3D biopolymer networks
• Software Download: SOAX

[Alejandro1400]

I found a manual for SOAX where it is better explained what each parameter means, way to call in batches and it may be possible to use the SOAX from python to automate the process of recollecting the images.

The output we receive can be helpful but it needs to be read through the batch option so that it saves a SNAKES file. I still have to look if it is possible from Python, if so then we still have to look at ROI selection to standarize the image selection

This possibility offers a more precise solution at the expense of requiring more post-processing analysis.

• SOAX manual

[Alejandro1400]

SOAX Algorithm Explanation

In the Section DESCRIPTION OF SOAX SOFTWARE it is explained how the algorithm works:

1. Initialization Stage: Multiple short SOACs are automatically placed along intensity ridges of the image, centerlines of filaments. The selection of this initial filaments is correspondent to a ridge tresholg that describes a minimal intensity steepness for a ridge to host an initial SOAC.
2. Evolution Stage: SOACs evolve one after another and stretch according to the local intensity contrast, weighted by a global parameter (kstr) that controls how easil SOACs elongate overall. SOACs stop stretching at a filament end or when their tip collides with the body of another SOAC to form a T-junction (they are parallel). Checks are performed to eliminate SOAC overlapping.
3. Final stage: Nearby T-junctions are clustered into single junctions followed by configuring the local connectivity of SOACs by cutting and splicing them so they do not end or bend sharply at junctions (It is expected that they will continue their path). This stage results in a set of junction points and re-configured SOACs

Network extraction depends on the SNR of the image, for which the ridge threshold (How intense it needs to be) and stretch factor (how easily SOACs elongate) are estimated relying on the availibility of ground thruth, which is obtained through supervised machine learning if not available.

The SOAX software goes through iterations of modifying these 2 parameters to evaluate the F-function that evaluates an extraction result using only the image and the result. F=-Ltotal + cL, where Ltotal is the length of SOACs in the extraction result, and L is the length of SOACs in regions where image has a low SNR. This approximates to the Ground truth

[Alejandro1400]

Proposed Analysis by SOAX

1. Spatial Distribution: Distribution of filament orientation, curvature and density This can help reflect the concentration and thickness distribution of actin filament bundles.

2. The orientation can be used to detect and select sections of the cell of interested. As it is documented in SOAX they are able to analyze that the orientation of filaments can be interpreted as sections of the cell (It would be interesting to see if this can be a processing step to select ROIs)

Still, the analysis we are interested in is more so related to DOL (analyzing SNR, Gaps, line length...)

[Alejandro1400]

Automatic extraction of actin networks in plants

In this paper they describe an algortihm, I haven't found it yet though. But the interesting thing is their Quantitative network measures:

• Skeleton Density: Can be used to highlight how tightly packed the network iis
• Filament width: Provides information about bundling activity
• Filament Length: May be indicative of growth, depolymerisation, or severing activity

[Alejandro1400]

Ridge Detection Wiki

[Alejandro1400]

Ridge Detection Line Width function

-Width