Researchers from DZNE and the Vascular Neurology department of the Bonn University Hospital (UKB) want to develop a computer model based on artificial intelligence (AI) to help doctors treat patients with stroke. It serves as a digital auxiliary system, it is intended to predict the long -term outcome of patients after minimal invasive treatment (mechanical thrombectomy) and potential complications, which helps doctors about the best possible therapy. A proof-of-concept study will now be carried out to determine whether this is feasible using data from the “German stroke register” and extra brain images. The project is based on an AI technology called “Swarm Learning”, which takes a new path in the safe analysis of distributed medical data and is intended to lay the foundation for a network of clinics in Germany and beyond. Cispa Helmholtz Center for Information Security is also involved in this endeavor, which is financed by the Helmholtz Association with 250,000 euros.
A stroke is manifested by neurological symptoms, such as speech shortages or paralysis. The most common cause are blood clots: plugs in brain vessels that impede blood flow and therefore oxygen supply. This situation is called “ischemic” stroke.
In such an event, millions of brain cells die every minute unless countermeasures are taken quickly. This is very time critical. Time is brain, as they say. ”
Dr. Omid Shirvani, Arts and Dzne Scientist
AI for personalized medicine
Possible measures are, for example, medicinal resolving the blood clot or mechanical thrombectomy, a minimally invasive procedure that aims to remove vessel blocking by means of a special catheter. “The type of treatment is decided on a case -by -case basis, depending on factors such as, for example, the size of the closed? Doel? Do you emphasize:” We do not want a black box, the predictions of our computer model must be understandable for doctors, so that they can make a well -considered decision for the individual patient. That is, our AI must have what is called “explanation” and its characteristics are based.
Combining different types of data
AI relies on algorithms that are trained on large amounts of data to recognize patterns. The greater the pool of training data, usually the better the AI will learn. The researchers therefore intend to combine data from the “German stroke register” with extra brain images generated by magnetic resonance image formation (MRI) or computer tomography (CT). This central register contains data on the treatment of ischemic strokes of more than 20 hospitals throughout Germany. It contains thousands of things. “This information comes from the initial investigation and follow-up care after a thrombectomy up to three months after intervention. These are mainly detailed entries from the medical data. However, associated MRI or CT images of the brain are not included. In general, these are generally held at the DZNE. is for training our AI. That is why we want to link this local data to the information from the Central Register. “
Travel algorithm
This is where “Swarm Learning” comes into play. The innovative AI technology is the center of current effort. “Traditionally, image data would be collected centrally. Given the enormous amounts of data, this is complex and difficult to scale if the network of partners has to grow. And because these are personal data, it requires the legal regulations that take a considerable amount of time to meet. Key role in the project. Instead, we send the algorithm to the data via the internet. We let the AI travel from place to place, to speak, to learn. That is the core idea behind Swarm Learning. “
Collective Learning
This approach was developed by DZNE in collaboration with IT company Hewlett Packard Enterprise and is currently being used in various DZNE projects. The term “swarm” refers to the partners who interact within the network. “When learning swarms, everyone from the Collective Datapool benefits without having to share their own data. This data remains on site and confidentially in accordance with the data protection regulations. This is because the algorithm only extra -held parameters without personal references,” says Prof. Joachim Schultze, also a Professor at the University of Medicine. “The result is a trained AI that has learned at all network nodes. It has assimilated collective knowledge and can even evolve as new data is introduced. In our specific case we would then have an AI-based computer model that doctors could support in treating strokes.
International perspective
Starting with three clinics, including Bonn University Hospital, the researchers are planning to gradually extend their approach to other members of the “German stroke register”. For testing purposes, they will start with multicentric data from the “German battle registration” that is available in Bonn and use it to simulate a swarm in the Dzne computer center before transferring the system to geographically individual locations. “We want to lay the foundation for a national network,” says Aschenbrenner. “Furthermore, we are already in conversation with partners in the UK to continue our concept internationally. I think there is a lot of room for development.”