IfThresholds Software Package: score ranking standard threshold methods by comparing images produced by thresholding with manually labeled images

Abstract

This dissertation details my research on threshold methods commonly used in image analysis and the development of the Python package ifThresholds. Imaging in biological and life science research is routinely performed, almost every biological scientist utilizes imaging to gain insight on structures and functions of cells, measure biological parameters, and monitor substrate in real-time. However, quantifying images to draw meaningful conclusions can be challenging. In my dissertation, I showed the background and the motivation of the project along with an overview of the image analysis field followed by descriptions of the images I used to build the package and generate a score ranking table. I also introduced the software I used to manually label images and the programming language I chose to develop ifThresholds. In addition, I presented the development of the software package and the results. So far, ifThresholds assessed and compared 7 global threshold methods that are easily accessible in various image processing software with images obtained from the experiment performed in my lab. The score ranking table gives researchers and scientists insights on how well each threshold method performs in images, thus helps them choose the best threshold method. Users can also run ifThresholds from terminal to analyze their own images and have a score table saved locally. Lastly, I discussed the future directions of ifThresholds which include implementing a decision tree to predict the best threshold method with only a few image data inputs, developing a user-friendly web application, and generalizing the package to multi-channel images.