Enabling Clinical Trials of Artificial Intelligence: Infrastructure for Heart Failure Predictions

Abstract

The last decade has seen a large increase in artificial intelligence research within healthcare. However, relatively few attempts of clinical trials have been made for such configurations. One of the main challenges arise in the extensive infrastructure necessary, both for development, but particularly to run prospective studies. In this paper, infrastructural requirements are first presented, together with constraints due to underlying production systems. Then, an architectural solution is presented, with the aim of both enabling clinical trials and streamline model development. Specifically, the suggested design is intended for research of heart failure prediction from ECG, but is generalizable to projects using similar data protocols and installed base.