The overarching goal of this project is to design a cell-counting algorithm in Python based off of open-source code written in C++. This is intended as a way to learn about, experiment with, and, of course, produce a working machine learning program.
The idea for this program came from the connectomics research lab that I work in at the Rockefeller Neuroscience Institute at West Virginia University. syGlass, a software company from this lab, was derived from the need to obtain a better visualization of the data (mostly in the form of 3D images) used on a day-to-day basis. Although intended for any field that could benefit from better 3D image visualization (for example, NETL utilizes the software to analyze the interaction of liquids and gases in sub-millimeter rock pores), syGlass is especially excellent at viewing images of stained tissue. One particularly useful application for syGlass in this domain would be the ability to count the number of cells occuring within the volume. This use case is present in other technologies similar to syGlass, such as Vaa3D. However, most of these programs aren't as strongly-supported as syGlass, and the algorithms have limited efficacy. Currently, Vaa3D is used to count and mark as many cells as the algorithm can find before being sent to syGlass to complete the marking by a human. It would be much more convenient to have this cell counting feature within syGlass instead of having to use another program and import the results into syGlass, and it is likely that the cell counting algorithm itself could be improved upon.
The work being done on this project is purely exploratory, and will most likely not be used within syGlass. This project will enable me to learn about various feature selection, dimensionality reduction, and machine learning methods currently being used, as well as concepts including k-fold testing and data visualization in software. Any results produced from this project can be used by syGlass as a means to direct their future implementation.
The code this project is based off of comes from Vaa3D, a free open-source software developed by researchers to view 3D images. More information on syGlass can be found at https://www.syglass.io/.