HospitalsHe is the project manager of STRATIF-AI, which has been granted SEK 65 million over a four-year period by Europe Horizon (the European Commission). The project is linked to stroke research, but the digital twins method is expected to have many different applications.
LiU has been granted SEK 11 million of this money. A total of fifteen actors are involved in the project, including five higher education institutions and eight hospitals. These are hospitals in Region Östergötland, Region Västerbotten and similar operations in Romania, the UK, Spain and Germany.
“We’re very happy about this. We submitted the application and are responsible for the design and coordination of the project,” Gunnar Cedersund adds.
Which conditions has LiU studied so far?
“We have, for instance, studied models for blood pressure medicine, or a diabetes medicine that can promote insulin production. We can also simulate how increasing the amount of exercise or changing your diet impacts the risk of diabetes or stroke,” says Gunnar Cedersund.
For several years now, he has been working on simulation of human organs, which is a bit like completing a digital puzzle of the human body. This is done by combining machine learning and mechanistic models. Using machine learning, a computer model gathers extensive research data and knowledge from hospitals, universities and biobanks, to create risk models for stroke.
Machine learningThe method for machine learning is called federated learning, which is a key concept in the project:
“The computer models can jump between the databases, learn what can be learned and take that knowledge with them. This means that there is no need to save all data in one place, which would be ethically and legally impossible. In this way hospitals, for example, can retain their data behind their firewalls.”
The twin also has a mechanistic part describing different processes, such as biochemistry, metabolism, blood flow, brain activity, etc. This part is based on more than 20 years’ research, with experiments both in labs and in clinical environments (health care), and publications on different sub-models.
The combined model, where organs communicate with each other, is what makes up the digital twin. The twin is person-specific and is regularly updated via ongoing analyses of new data generated on this person. This data may be the result of measurements from various test methods, such as portable sensors, new examinations and similar.
“We can simulate the effects of a particular drug on the body and examine possible results. We can also show these simulations to patients, which we hope will improve doctor-patient communication. “
Stroke careOne expected result of this is that patients’ understanding of their own health, and thereby their motivation for change, will increase:
“We will investigate whether more people can be motivated to start exercising, eat more healthily or take their medicines more consistently.”
The twin may be useful also in stroke care, both in emergency care and afterwards in rehab.
“In emergency stroke care, the twin can be used to find out the case history of a patient and decide on what would be the optimal medical intervention. The results of any tests performed in emergency care may prove useful in the future. During rehab, the patient can look inside their twin and follow the rehab process, and this could also increase their motivation to do their rehab exercises,” says Gunnar Cedersund.