Analysis in Python of photographs of spheroids from the optical microscope.
SpheroidSnap is a script written in Python that analyzes the geometry of spheroids from photographs taken from an optical microscope (even with a regular smartphone through the lens!).
Utilizing publicly available python libraries such as scikit-image, numpy, scipy, shapely, etc. SpheroidSnap detects spheroids in your photos and calculates their perimeter, area, circularity and quantifies their likeness to a regular polygon in a quick and efficient way, while also giving you the option to output computed data to .json file for further processing.
Having scikit-image as its main processing unit, SpheroidSnap reads the image specified by the user and turns it to HSV colorspace, where it applies a triangle threshold to the saturation channel, thus creating a mask that removes the white background from the spheroids. Afterwards, it detects closed contours on the masked image and guides the user on how to select only those that need to be studied and aren't noise.
In the next step, the convex hull of each selected contour is generated and analyzed with both scipy and shapely modules. The analysis contains the calculation of the perimeter, area, number of sides, centroid coordinates and circularity via the Polsby-Popper algorithm.
Assuming that one of the forementioned convex hulls is an irregular polygon with n sides, then, if it was a regular n-gon, the area of all the triangles formed by joining the centroid to the vertices of each edge, would be equal with one another and with the total area divided by the number of sides. Thus, the mean and median of the distribution of those fractional areas would be equal to total area / n. Based on this fact, SpheroidSnap computes the "area distribution" for each convex hull and performs a One sample Wilcoxon Signed Rank test for the null hypothesis that the median of the difference of the computed "area distribution" and the "area distribution" of a regular n-gon (with n less than or equal to the number of sides of the hull) is zero.
Finally, using matplotlib and seaborn, SpheroidSnap presents the results to the user, in a concise and friendly interface, while also providing the option to save all generated results to a .json file, using the eponymous Python module.
SpheroidSnap was developed by undergraduates at the University of Thessaly's Department of Veterinary Medicine and is free to use in any way that promotes science and learning, as long as you cite the github repository it resides in.
You can always download and run SpheroidSnap localy, however that requires that you have installed Python along with all the dependencies on your local computer. Since this may not always be something that an average non-programmer has, we provide another option.
You can run SpheroidSnap on the go! How ? You can download the .ipunb file and run it on Google Colab (there are instruction on Google on how to do this). Along with uploading said file, upload the image you want to analyze in the same folder. Keep in mind that you might need to erase old analyzed images if the Google Colab space is full. Afterwards, run every cell consecutively and follow the instructions it shows. Keep in mind that you may need to manually install some packages using the command "%pip install packagename".