Researchers from Turku UAS win top positions in an international AI competition

Can AI help healthcare without the need to share patient data? Can we develop better treatments without risking privacy? Can hospitals collaborate safely – without leaking data? Will we be on board when the global Federated Learning market reaches a potential of more than $250 million in 2027?

The researchers of the Secure and Accurate Federated Learning as a Service (SA-FLaaS) project at Turku University of Applied Sciences have tackled the tangle of problems with an open mind. The pioneering project develops AI solutions in which machine learning can take place in a decentralized manner and without the need to transfer or share confidential data.

The SA-FLaaS project is developing a method in which hospitals and research institutes can train AI models together – without sharing actual patient data. The method is based on distributed machine intelligence learning, where each participant trains their AI model locally and shares only the trained models, not the data itself. This allows for the use of artificial intelligence, but at the same time ensures that the privacy of patients stays protected.

“The technology we have developed is really interesting in terms of scalability and business,” says Heikki Lassila, who oversees the commercialization of the project. “The scalability of our award-winning and patented algorithms does not only end in healthcare or brain injury research. The technology can also be used in other areas of the business where sensitive data is managed.

New AI methods may involve not only challenges in the operating environment but also prejudices. In the global hype, models may appear that work on a theoretical level, but when they are implemented, it is noticed that the solution does not work in the operating environment.

“We can boldly say that our solutions do not work only in theory. The project utilizes two revolutionary innovations patented by Turku UAS, says Lassila.

Finnish higher education institution is fighting for top positions in AI competitions

The FeTS Challenge is a competition that enjoys international attention, sponsored by Intel, one of the world’s largest and most significant technology companies. The competition, which has been held three times, focuses on the process of marking areas of brain tumors in brain scans (MRIs). The solution presented by Turku UAS was awarded in the 2021 competition, in which 125 international teams took part. In last year’s competition, Turku UAS’ innovation achieved second place among more than 80 top international solutions. 

Achieving the top places in the international competition for AI research requires not only top-level research skills but also a new, creative can-do attitude.

“We really have the courage to challenge what we are used to and the perseverance to continue the journey towards a tangle of problems,” Lassila praises the research team.

Turku UAS’ patented and award-winning algorithms and the SA-FLaaS project naturally aim towards goals that are larger than individual competitions.

“Success in competitions serves as a testament to our expertise. The success credited to a tech giant like Intel concretizes talk of “Finnish excellence”. We are getting closer and closer to the Center of Excellence status, emphasizes Elina Kontio, Head of Education and Research at Turku UAS.

As the digitalization of healthcare progresses rapidly, research focusing on ethics and safety is needed. The SA-FLaas project keeps a sharp vision in which hospitals around the world work together to develop better forms of treatment.

“Our perspective on research is global. The goal is to enable secure and privacy-preserving collaboration anywhere in the world. We are paving the way for increasingly exact and ethical AI solutions, emphasize Suleiman A. Khan and Muhammad Irfan Khan, researchers at Turku UAS.