FlightPlan for Liver

FlightPlan for Liver is an advanced medical image processing software designed to enhance the planning phase of liver embolization procedures by providing physicians with tools to visualize and analyze hepatic vasculature from 3D X-ray angiography images. It employs both deep learning for automated liver segmentation and non-deep learning algorithms for virtual parenchyma visualization, offering adjunctive information to aid in the evaluation of hepatic arterial anatomy and the identification of arteries near hypervascular lesions.

FlightPlan for Liver is a post-processing software package designed to assist physicians in analyzing 3D X-ray angiography images of the liver. Its primary purpose is to aid in the visualization of hepatic vasculature and the identification of arteries leading to the vicinity of hypervascular lesions, thereby supporting the planning phase of endovascular embolization procedures. The software, an evolution of a predicate device, introduces a 'Parenchyma Analysis' option, which includes two new algorithms: a deep learning-based Liver Segmentation algorithm for automated segmentation of the entire liver from injected Cone Beam CT (CBCT) acquisitions, and a non-deep learning Virtual Parenchyma Visualization (VPV) algorithm. These enhancements allow for multi-modality 3D review, semi-automatic segmentation of the liver arterial tree, operator-defined targeting, and interactive display of vascular structures. The methodology involves using CBCT images as input. The Liver Segmentation algorithm's output serves as a boundary for the VPV algorithm, which then selectively displays the estimated distal liver region adjacent to the vessel segmentation. The software also incorporates a 'Live Tracking' feature for selective display of proximal and distal vessels. Non-clinical testing included engineering bench tests for the new algorithms, with the deep learning algorithm trained and tested using a database of CBCT liver acquisitions. The VPV algorithm's performance was evaluated against selective contrast-injected CBCT exams as ground truth. Clinical assessment involved interventional radiologists evaluating image pairs, demonstrating that the device aids physicians in visualizing and analyzing 3D X-ray angiography images, estimating distal liver regions, and planning embolization procedures, including the selection of injection points. The software operates on GE’s Advantage Workstation and AW Server platforms.

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 deep learning-based Liver Segmentation algorithm vary with different qualities or types of input CBCT acquisitions, such as variations in contrast levels or patient anatomical differences?; What were the specific predefined acceptance criteria for the Virtual Parenchyma Visualization (VPV) algorithm, and how was its performance quantitatively measured against the selective contrast-injected CBCT exams used as the ground truth?; The clinical assessment utilized a 5-point Likert scale; what specific aspects of visualization, analysis, and planning utility for embolization procedures were evaluated by the interventional radiologists, and what were the key findings or average scores?; How does the 'Live Tracking' feature, incorporated from a reference device, functionally integrate with the new Parenchyma Analysis option, particularly the Virtual Parenchyma Visualization, to enhance the interactive planning of liver embolization procedures?; What is the typical processing time required for the Parenchyma Analysis option, encompassing both the deep learning Liver Segmentation and the Virtual Parenchyma Visualization algorithms, when run on the GE Advantage Workstation or AW Server platforms?