An Open-source, Flexible, Plug-and-Play Inline CMR Image Segmentation Platform

Motivation: Cardiac magnetic resonance (CMR) image segmentation remains a time-consuming task. Deep learning (DL) segmentation models have advanced considerably in recent years, but they lack flexibility for manual adjustments. Goal(s): We sought to develop an open-source, flexible plug-and-play inline CMR image segmentation platform. Approach: We implemented in-house algorithms in a clinical scanner as separate Python modules, allowing new and existing DL models to be integrated into clinical workflow. Results: We assessed the feasibility of the automated plug-and-play platform for inline cine segmentation in 308 patients referred for clinical CMR. Impact: This platform facilitates the rapid development and evaluation of any segmentation algorithm in a transparent and reproducible fashion. An open-source, flexible, plug-and-play inline CMR segmentation platform will enable rapid testing and evaluation of new segmentation and analysis algorithms.