FlightPlan for Embolization

FlightPlan for Embolization is a medical image processing software that enhances the planning phase of embolization procedures by enabling physicians to analyze 3D X-ray angiography images. It employs both semi-automatic tools and newly introduced AI-powered deep-learning algorithms for automatic vasculature segmentation and root point definition, with its efficacy supported by non-clinical testing and clinical evaluation.

FlightPlan for Embolization is a post-processing software package designed to assist physicians in analyzing 3D X-ray angiography (CBCT) images for planning endovascular embolization procedures. Its primary function is to help visualize vasculature, providing outputs such as segmented vasculature and selective display of proximal and distal vessels from a user-defined reference point. While user-defined data can be exported for use during the guidance phase of a procedure, the software is intended for the planning phase only and not for therapy delivery; injection points must be confirmed independently. The system operates on GE HealthCare's Advantage Workstation (AW) and AW Server (AWS) platforms and is an extension to their Volume Viewer application. The key technological advancement in this version of FlightPlan for Embolization is the introduction of an 'AI Segmentation' option. This option incorporates two new deep-learning algorithms: one for automatic segmentation of vasculature and another for automatic definition of the root point. These AI-driven features are intended for the adult population. The algorithms were developed and qualified using contrast-injected CBCT scans from GE HealthCare's interventional X-ray system. Performance was evaluated using a segregated test set of 207 scans, demonstrating a 93.7% success rate for the vessel extraction algorithm and a 95.2% success rate for the root definition algorithm against pre-defined acceptance criteria. Clinical assessment by interventional radiologists further substantiated the utility of these AI features in procedure planning.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the performance of the AI-driven automatic segmentation compare to experienced human operators in terms of time efficiency and accuracy for complex or atypical vascular anatomies?; What were the specific criteria defining success for the 90% threshold in the vessel extraction and root definition algorithm testing, and how were these criteria clinically validated?; Are there plans to expand the validation of the AI Segmentation option beyond the adult population, for instance, to pediatric cases or specific challenging anatomies not heavily represented in the test set?; How does the system handle image artifacts or variations in CBCT acquisition protocols (beyond GE HealthCare systems), and what is the impact on the AI algorithms' reliability and accuracy?; What is the typical learning curve for clinicians to proficiently use the new AI Segmentation features, and how does it integrate with existing downstream applications like GEHC’s Vision Applications for 3D road-mapping?