With the help of machine learning methods and based on real, anonymised patient data, around 35 students and doctoral students from the University of Bonn developed predictive models to support doctors in making key decisions in everyday clinical practice. The first #Hack4Surgery was organised by BOSTER together with the Transfer Center enaCom and the Bonn-Aachen Institute for Information Technology (b-it).
Support from prediction models offers advantages in diagnostics
"Every day, doctors make decisions about people's lives: What treatment is required for certain key symptoms? Will emergency surgery be performed for acute abdominal pain or not? Machine learning-based prediction models can help provide fast and precise support for challenging questions in everyday clinical practice," emphasises organiser Dr. Jonas Henn from BOSTER. Participants in the hackathon were given a keynote speech by Dr. Nils Sommer from the Clinic and Polyclinic for General, Visceral-, Thoracic and Vascular Surgery, which clearly illustrated the everyday challenges of clinical diagnostics. At #Hack4Surgery, participants had one day to work in teams of two and three on such decision models. In addition to this practical task, the participants had the opportunity to gain exclusive insights into the modern real-world laboratories and workspaces at BOSTER and to network with the experts from various disciplines there. BOSTER is an interdisciplinary competence center for digital surgery and medical robotics affiliated with the UKB.
Working with real data from the emergency center offered students practical insights
"It is particularly noteworthy that the modelling teams were able to work with real data from the UKB's emergency center as part of #Hack4Surgery. They also presented real solutions that have the potential for practical application," explains co-initiator Dr. Daniela Treutlein, Innovation Scout at the Transfer Center enaCom. During #Hack4Surgery, Prof. Dr. Holger Fröhlich, Sophia Krix and Manuel Lentzen from b-it and Fraunhofer SCAI were also on hand to advise the participants and act as jury members. “We are delighted about the great interest in this hackathon and are very impressed by the results,” says Prof. Fröhlich. “For our students it is of immense value to understand the impact of AI and specifically machine learning in medicine. In this regard, the possibility to work with real data is a fantastic chance.”
Following the intensive work phase, the three best teams selected by the jury - undefined, fsk and Code Geass - pitched their solutions on the big stage. Niklas Dobberstein and Amr Moustafa of the team undefined were then chosen as the winners by the jury and received their trophy and jumper from the University of Bonn on stage. The two M.Sc. Computer Science students of the University of Bonn achieved the best prediction performance and put the results into a target-oriented medical context. At the end of the afternoon, all participants went home with numerous inspirations and exciting insights - the best prerequisites for a successful repetition of this interdisciplinary event at the interface between computer science and medicine.
