From a research idea to a life-saving patent: the case of FibMap

Cesare Alippi is a full professor at the Faculty of Informatics at USI and affiliated with the Dalle Molle Institute for Artificial Intelligence (IDSIA USI-SUPSI), where he serves as Scientific Director.

His main research interests include Graph Machine Learning, adaptation and learning in non-stationary environments, and intelligence for embedded, cyber-physical and IoT systems. Professor Alippi also heads the Graph Machine Learning Group (GMLG), within which, thanks to a collaboration with Imperial College London, FibMap was developed.

Professor Alippi, how did the idea for FibMap come about?

"The FibMap idea originates from an international collaboration between Imperial College London and the University of Oxford, addressing the medical challenge of atrial fibrillation, one of the most common and complex arrhythmias to treat. My research group focuses on basic research to drive innovation and applies it to real-world scenarios, combined with the most advanced Artificial Intelligence technologies. In the case of FibMap, we developed a methodology capable of reconstructing missing data. From our meeting with Prof. Danilo Mandic and PhD student Alexander Jenkins from Imperial College, we realised that we could adapt this methodology to reconstruct the dynamics of atrial fibrillation in the human heart. After testing our model on mice, we aimed to enhance the effectiveness and personalisation of ablation surgery. This is how FibMap was created."

Could you explain, in a way that is simple and understandable even to non-experts, how FibMap works and what distinguishes it from previously available systems?

"FibMap is an artificial intelligence algorithm designed to reconstruct a complete electrical map of the cardiac atrium using only a small portion of related data arranged in graph structures. Specifically, FibMap can "touch" only 10% of the atrial surface, yet it is capable of reconstructing the entire electrical map with 210% greater accuracy than existing techniques.

The algorithm exploits both the spatial and temporal dimensions in the reconstruction, i.e., the three-dimensional surface of the atrium and the evolution of electrical signals that can be acquired at spatial points over time. With appropriate processing, the algorithm is able to complete the missing data and identify with great accuracy the critical points where ablation is most effective."

How do you think this project can and will benefit patients?

"FibMap offers significant advantages in managing atrial fibrillation, a cardiac arrhythmia characterised by uncoordinated electrical signals. If left untreated, this condition can lead to strokes and severely impact one's quality of life. Currently, ablation, the surgical procedure aimed at "eliminating" the problematic areas, is not always effective because it relies on an accurate atrial electrical map, which is often unavailable.

With FibMap, we can significantly enhance the likelihood of successful ablation while minimising the risks associated with ineffective or inaccurate procedures. Additionally, our algorithm adapts to each individual patient by recalibrating itself based on personal characteristics after analysing the behaviour of a group of patients. The more patients we gather, the more accurate the system will become over time, leading to improved reconstruction performance. What we have achieved is a concrete example of personalised medicine made possible by artificial intelligence.

What steps did you take to patent your idea, and what are the next steps?

"We initially filed the patent application in the United Kingdom. From that point, we have one year to decide on the next steps. In our case, we plan to submit an international PCT (Patent Cooperation Treaty) application, which will protect the invention for up to 30 months. Given the significance of the Swiss market, we will also extend our protection to our country as well.

Technology transfer is a crucial step, and in this context, the role and support of USI Transfer have been instrumental. Some people may assume that smaller universities provide less structured services than larger institutions, but that is not the case. In my extensive experience, the "human scale" of USI Transfer fosters close relationships, attention to detail, and a speed of response that is often lacking in larger contexts. This distinctive advantage has significantly impacted the process of protecting and promoting FibMap. I would like to express my gratitude to Andrea Foglia, the Head of USI Transfer, for his valuable support.

The patent is a first step in protecting the idea, but bringing FibMap into the clinical setting is another challenge, requiring full compliance with health and regulatory standards. The algorithm works, and the results are very promising. The next task is to transform this innovation into an operational clinical tool in the field.