Publication of the stEVE paper in Computers in Biology and Medicine
We are pleased to share that our paper, “Learning-based autonomous navigation, benchmark environments and simulation framework for endovascular interventions,” has been accepted for publication in Computers in Biology and Medicine. In this work, we present stEVE (simulated EndoVascular Environment), a modular simulation framework designed to advance research in autonomous endovascular navigation. Current research in this field suffers from incomparable evaluation environments, with each study employing unique setups that hinder progress. Our framework addresses this challenge by providing standardized benchmark environments and open-source tools. We introduce three distinct benchmark interventions – BasicWireNav, ArchVariety, and DualDeviceNav – that systematically evaluate fundamental navigation capabilities from single-instrument control to complex dual-device coordination. Using deep reinforcement learning, our autonomous controllers achieved success rates up to 98% in simulation, with successful transfer to physical test benches demonstrating 97% success rates under real-world conditions. The framework’s modular design enables researchers to adapt vessel geometries, instrument configurations, and imaging systems to their specific requirements while maintaining reproducibility across institutions. By providing open-source benchmarks and simulation tools, this work reduces barriers to entry and establishes a foundation for meaningful comparison of autonomous navigation approaches in endovascular interventions.
https://www.sciencedirect.com/science/article/pii/S0010482525011953
