This repository has the code for the bacterial cellulose fiber analysis (using SEM) that has been published as:
"Silicone polyether surfactant enhances bacterial cellulose synthesis and water holding capacity"
International Journal of Biological Macromolecules, 2022
https://doi.org/10.1016/j.ijbiomac.2022.03.124

SEM_fiber_analysis

This repository contains python scipts to perform fiber analysis in SEM images. It uses scikit-image and quanfima¹ for the analysis, dask for lazy-loading and parallelization, and napari for intermediate outcome visualization.

Step-by-step example workflow

The Jupyter notebook Example_Fiber_analysis.ipynb provides a step-by-step example of the image analysis workflow, using a representative SEM of bacterial cellulose fibers, provided in the fiber_data folder. An exported PDF version is also provided.

Visualizing the segmentation and skeletonization using napari

The python script Visualize_Segment_Skeleton.py permits visualization of the outputs of segmenting the fibers, followed by skeletonization. These are crucial steps for the downstream quanfima analysis.
First, images are loaded using dask-image with each image being a block in a dask array. Then segmentation and skeletonization steps are applied using map_blocks, a dask.array method that applies a function to an array (here stack of images) block-wise (here image by image).
Using napari these steps can be visualized lazily, on a per-image basis, without pre-computing for every image. The computations are performed when a slice is selected.

Computing the fiber properties (porosity, orientations, and diameters) using quanfima

The python script Analyze_fibers.py calculates the fiber parameters using quanfima morphology module.
As previously, images are loaded using dask-image with each image being a block in a dask array. Then segmentation is obtained using map_blocks, as previously described.
Next, a function is defined to calculate the porosity using quanfima and the calculation is parallelized using dask. The outputs are paired up with the names of the input images and returned as a dict.
Finally, the segmented images are used to obtain the fiber orientations and diameters again using quanfima. As previously, the calculation is parallelized using dask.
Note: it can take ~1 min per image to run this, so it's recommended to consider using the processess scheduler or the dask.distributed scheduler. See the Dask scheduler docs for details.
Summary values (means and standard deviations) are computed, paired up with the input images names, and returned as a dict.

Footnotes

1. Note: In python3, the quanfima module has some problems due to relatively old dependencies. A fork capable of running 2D image fiber analysis in python3 is available: https://github.com/psobolewskiPhD/quanfima ↩