How to integrate an AI algorithm?
Our new students Sander, Alina and Kamil try to answer this question in their bachelor thesis.
Hello,
Our names are Sander, Alina and Kamil, and we are currently in our last year before completing our bachelor’s degrees in computer science at Western Norway University of Applied Sciences. We have been fortunate enough to have the opportunity to write our bachelor thesis in collaboration with and with help from PiV.
Our bachelor thesis focuses on digital pathology and whether it is possible to effectively integrate AI algorithms into pathologists' workflows by using the EMPAIA API as an intermediary. An API, or application programming interface, can be described as a way for a part of an application to be run or used from a different application. In our case, it involves utilizing an AI algorithm from within the image management system used by pathologists for making diagnoses.
A screenshot of the local EMPAIA APP Test Suite. With an APP that calculates the fibrosis score of a specified area of a WSI.
The main focus is to determine whether EMPAIA is a viable option or if it is too complicated to be used effectively. We are working with a multidisciplinary team, therefore our challenges include not only making the system work for ourselves but also communicating the process to others in an understandable way. Hopefully, this will prove to be possible, and we will be able to automate this process for simplified use for both developers and pathologists.
Another focus is the manipulation of whole slide images (WSI). By using machine learning algorithms we want to remove the background and certain structures (glomeruli) from WSI' of kidneys, this is to optimize these images for our intended use. Down the line, this will be part of an automated process that will work in conjunction with other AI algorithms through the aid of EMPAIA.
We are lucky to have Patrick and Sabine as our supervisors. Their support and guidance have been important to the advancements of this thesis.
Hopefully, our work will prove useful. Thanks to everyone at PiV for their willingness to help, and the friendship they provide.