Bayesian anatomically-driven, artificial-intelligence based intracardiac echocardiography object detection and prediction

Abstract

The systems and methods perform Bayesian anatomically-driven, artificial-intelligence based intracardiac echocardiography object detection and prediction. The system proposes hypotheses for what features and objects are in view based on prior knowledge informed by previously acquired or generated models. During a cardiac interventional procedure, the system uses a per-frame object identification artificial intelligence (AI); the gathered data is analyzed within a Bayesian framework to create updated posterior prediction(s) of what feature(s) may be predicted (with a computed confidence level) to be in view and their respective location(s). Inferences that pass an accept/reject threshold for object identification and meet minimum threshold settings for temporal stability and spatial location are used to display labels and boundaries. The system predicts what features and objects are in-frame, where out-of-frame objects/features may be located, and how to navigate to them.