Reinforcement learning (RL) approaches are starting to be applied in a few medical applications, mainly for landmark detection. In patients suffering cardiac arrhythmias, the detection of the optimal points to apply radio-frequency ablation (RFA) is key to stop the clinical problem. Nevertheless, it is a challenging therapy from which success is often based on the clinician experience and trial-and-error, thus sometimes being long-time procedures. The main goal of this project is to develop RL techniques to predict the optimal RFA targets based on information provided by pre-operative delay-enhancement images, as well as ECGs and intra-operative electrophysiological information. This project will be performed in collaboration with renowned electrophysiologists from Teknon-Quirón Salud and engineers from the ADAS3D company.
Supervisor: Oscar Camara