@article{MRA, author = {John Komp and Aaptha Boggaram and David Kao and Ashutosh Trivedi and Michael Rosenberg}, title = { Strategies and Considerations for Safe Reinforcement Learning in Programming Cardiac Implantable Electronic Devices}, journal = {Medical Research Archives}, volume = {13}, number = {3}, year = {2025}, keywords = {}, abstract = {The programming of cardiac implantable electronic devices, such as pacemakers and implantable defibrillators, represents a promising domain for the application of automated learning systems. These systems, leveraging a type of artificial intelligence called reinforcement learning, have the potential to personalize medical treatment by adapting device settings based on an individual’s physiological responses. At the core of these self-learning algorithms is the principle of balancing exploration and exploitation. Exploitation refers to the selection of device programming settings previously demonstrated to provide clinical benefit, while exploration refers to the real-time search for adjustments to device programming that could provide an improvement in clinical outcomes for each individual. Exploration is a critical component of the reinforcement learning algorithm, and provides the opportunity to identify settings that could directly benefit individual patients. However, unconstrained exploration poses risks, as an automated change in certain settings may lead to adverse clinical outcomes. To mitigate these risks, several strategies have been proposed to ensure that algorithm-driven programming changes achieve the desired level of individualized optimization without compromising patient safety. In this review, we examine the existing literature on safe reinforcement learning algorithms in automated systems and discuss their potential application to the programming of cardiac implantable electronic devices.}, issn = {2375-1924}, doi = {10.18103/mra.v13i3.6363}, url = {https://esmed.org/MRA/mra/article/view/6363} }