Intra-cardiac Mapping Basket Using Biosignal-Based 3D Reconstruction

The patent introduces an intermediary computational layer that processes intracardiac electrogram signals through machine learning algorithms. This intermediary system bridges the gap between raw signal acquisition and precise driver localization, enabling the distinction between active and passive rotors through automated analysis of signal characteristics, wavefront propagation patterns, and spatiotemporal features.

The patent replaces traditional mechanical/manual signal analysis methods with automated machine learning-based computational systems. Instead of relying on manual interpretation of electrogram signals or simple filtering algorithms, the system uses trained neural networks and computational models to automatically identify AF drivers, determine rotor stability, and localize sources with high precision.

The patent enables the mapping system to determine electrode positions and reconstruct 3D cardiac anatomy using only the signals already acquired during the electrophysiological study. The intracardiac electrogram signals serve dual purposes: both for arrhythmia analysis and for navigation/localization. This self-service approach eliminates the need for separate navigation hardware while maintaining high precision through signal-based position determination.

The patent makes the intracardiac electrogram recording system multi-functional by enabling it to perform both its traditional role of arrhythmia signal acquisition and the additional function of 3D anatomical reconstruction and electrode position determination. This universal system uses the same hardware infrastructure for multiple purposes, avoiding additional hardware complexity while improving measurement precision through integrated analysis.

The patent applies preliminary action by pre-training machine learning models and computational algorithms before clinical use. During the actual procedure, these pre-trained systems rapidly process intracardiac signals to determine electrode positions and localize AF drivers. The heavy computational lifting of model training is performed beforehand, allowing fast real-time inference during the procedure without significant time loss.

The patent optimizes processing speed by changing parameters such as sampling rates, analysis window sizes, and computational complexity thresholds. The system adapts processing parameters based on the specific clinical scenario, balancing localization accuracy with processing time requirements. This allows high-precision localization to be achieved within acceptable time frames by dynamically adjusting processing parameters.