Algorithms Serving Medicine: USI-IDSIA Research to Enhance Atrial Fibrillation Treatments

A collaboration between researchers from the Dalle Molle Institute for Artificial Intelligence (IDSIA USI-SUPSI), led by Cesare Alippi, a Full Professor at the Faculty of Informatics at Università della Svizzera italiana (USI), and Imperial College London has resulted in the development of a neural network model. This model aims to enhance the precision and personalisation of surgical interventions for resolving atrial fibrillation. Ticino Scienza published an article about the project.

Professor Cesare Alippi leads the Graph Machine Learning Group (GMLG), where both fundamental research and various projects are developed to apply informatics and Artificial Intelligence across different fields. This has led to prestigious international collaborations, including partnerships with Imperial College London and Oxford University. One notable outcome of this collaboration is FibMap, a neural network model designed to reconstruct the dynamics of atrial fibrillation, a heart rhythm disorder characterised by uncoordinated electrical signals. The goal of FibMap is to personalise and enhance the efficiency of surgical interventions for this condition. Suppose drugs and therapy prove ineffective in the treatment of atrial fibrillation, which can cause blood clots and increase the risk of stroke. In that case, it is possible to intervene surgically and perform an ablation, i.e. to specifically lesion the areas of the heart that trigger the arrhythmia. "Before the ablation procedure, data is collected using a series of sensors that capture electrical signals," Professor Alippi explained. "Using this data, the electrical profile of the atria is reconstructed, allowing doctors to determine the optimal point for intervention with ablation. However, obtaining an accurate electrical reconstruction of the atrial profile is challenging because the sensors only capture limited data."

It is to overcome this problem that Professor Alippi and his research group, which includes Andrea Cini, Assistant Professor with a PhD from the Faculty of Informatics (and now at Oxford University), and Alexander Jenkins, PhD student at Imperial College London, visiting the group, developed the method called FibMap. "FibMap requires only 10 per cent of the atrial surface to function effectively. With this limited data, the method can reconstruct the overall electrical map of the atria with an accuracy of 210 per cent greater than current techniques. The algorithm takes advantage of both spatial dimensions—measuring points on the atrial surface—and temporal dimensions—capturing electrical signals over time at various spatial locations. Using this spatiotemporal data, the method effectively reconstructs the atrial electrical profile."

To date, the algorithm, which, as the USI professor explained, learns and improves performance the more it is used, has been validated on a sample of 51 patients: "The algorithm has -in fact- learned group characteristics that are then customised (recalibrated) to the unique characteristics of each new patient. The more patients are treated, the more the reconstruction capacity and accuracy of the method will improve."

While FibMap has the potential to be a useful tool in treating atrial fibrillation, it will take some time before it can be implemented on a large scale. In fact, the research team has just filed for a patent, as explained by Andrea Foglia, the Head of USI Transfer: "The initial patent application was submitted in the UK. From this point, you have one year to determine your next steps. Alternatively, you can file an international PCT (Patent Cooperation Treaty) application, which allows you up to 30 months to decide in which additional countries you want to seek protection for your invention. It is important to note that the process from patent filing to the final realisation of the new product can take anywhere from three to five years."

The complete interview with Professor Cesare Alippi, conducted by Ticino Scienza, is available at the following link. (Italian only)