A team of AI experts from the Oxford Internet Institute and the Mathematical Institute at the University of Oxford has won the prestigious 2024 George B. Moody PhysioNet Challenge, an international annual machine learning competition that invites participants to develop automated solutions for clinically relevant challenges.

This year’s winning team comprises researchers Felix Krones, Oxford Internet Institute, Ben Walker, Mathematical Institute, Professor Terry Lyons, Mathematical Institute, and Dr Adam Mahdi, Oxford Internet Institute, University of Oxford.

The competition invited participants to design and implement open-source algorithms for digitising and/or classifying physical electrocardiograms (ECGs). The Oxford team rose to the challenge of digitising ECGs, which involved turning images of ECG print-outs into digital time-series data.

By combining classic techniques with modern machine learning, they developed a robust solution to digitise the signals. This allows healthcare providers to digitise decades of historical physical ECGs, widening access to a large amount of previously inaccessible data for modern analysis and research.

The Oxford team achieved the highest score in the digitisation task, winning first prize for their algorithm’s top performance on a “hidden” test dataset not previously shared with participants.  The winners were announced on 11th September 2024 at the Computing in Cardiology Conference held in Karlsruhe, Germany.

Felix Krones, doctoral researcher who led this year’s efforts, said: “This year’s PhysioNet Challenge was especially rewarding as we applied insights from previous competitions to a new task. The great thing about the Challenges is their focus on evaluation criteria that closely mimic clinical needs. I hope we can use the platform we developed to engage further with healthcare practitioners, ensuring practical impact.”

Commenting on the win, Dr Adam Mahdi said: “It’s been great working with Felix and the team over the past three years since we first came together for this competition. Watching the team develop their skills in applying computational models to healthcare problems has been really impressive.”

Mahdi, a senior academic who assembled and supervised the team since its inception, added: “This win is well-deserved and highlights the hard work they’ve put into their data analysis and modelling. I’m proud of what they’ve accomplished and glad to have been part of the journey.”

The 2024 win sees the team build on their previous success in The PhysioNet Challenge 2023, when the Oxford team was awarded Best Preprint for their paper ‘Multimodal deep learning approach to predicting neurological recovery from coma after cardiac arrest,’ authored by Felix Krones, Benjamin Walker, Guy Parsons, Terry Lyons, and Adam Mahdi.

Ben Walker, a doctoral researcher from the Mathematical Institute at Oxford, added: “Being part of this project with colleagues from the Oxford Internet Institute has been a rewarding experience. It’s a great example of how cross-departmental collaboration can drive meaningful advancements and deliver practical solutions to real-world challenges.”

Find out more about The George B. Moody PhysioNet Challenge 2024 and the Computing in Cardiology conference.

Find out more about the winning team:

• Felix Krones, Oxford Internet Institute, University of Oxford
• Ben Walker, Mathematical Institute, University of Oxford
• Professor Terry Lyons, Mathematical Institute, University of Oxford
• Dr Adam Mahdi, Oxford Internet Institute, University of Oxford