"Voxels.Berlin" is a research group led by Prof. Dr. Schneider and Prof. Dr. Tewes. Together with the radiology department of the Charité, I had the opportunity to work on the development of artificial intelligence (AI) that can evaluate and assess CT and MRI images, specifically, images of the sacroiliac (SI) joints. This project, named SPARTA, intends to make the identification and diagnosis of sacroiliitis (inflammation of SI joints) easier and faster for medical staff.
How does AI learn?
The first step to achieve artificial intelligence in any project is the creation of an artificial neural network. This network has several layers of interconnected neurons that are similar to the structure of the human brain and is capable of learning. In the SPARTA project, its purpose was to be able to recognize specific patterns in CT- and MRI images, that are particular for sacroiliitis. The programming language that we used to create the network was Python, which already had existent libraries that facilitated this task. As soon as the basic structure of the artificial neural network is in place, the training process can begin. To do this, the network is supplied with images where the sacroiliac joints are marked and after a few training runs, the AI is finally able to recognize this disease on new images and to mark it itself.
Previous knowledge needed
During this internship the coding skills learned in Programmieren 1 and Programmieren 2 were fundamental. Even if we only used Python, the logical thinking used while programming in C++ was of great help. I found the knowledge obtained in MGÜ Labor and Bildgebung und Verarbeitung very resourceful as well because it allowed me to distinguish T1- and T2-weighted MRI images, or if the images had any kind of artifacts. With this, one could train the AI to ignore them or, ideally, to recognize them and calculate them independently.
More information
Contact
