Catheter Electrode Tissue Proximity Calibration With Reinforcement Learning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing medical systems face challenges in accurately determining tissue proximity of catheter electrodes connected to auxiliary signal processing units due to differences in computation methods between main and auxiliary signal processing units, leading to inconsistent and potentially inferior tissue proximity indicators (TPIs).

Innovation Solution

Employing reinforcement learning, specifically Q-learning or Deep Q-learning, to derive tissue proximity indicators (TPIs) for electrodes connected only to the auxiliary signal processing unit based on TPIs computed by the main signal processing unit, using a calibration electrode connected to both units and applying a reinforcement learning agent to train the auxiliary unit's TPI calculations to match the main unit's results.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If TPIs are computed separately by main and auxiliary signal processing units using different computation methods, then each unit can independently process electrode signals, but the TPIs become inconsistent and potentially inferior in precision

Engineering Contradiction:
Improveindependent signal processing capabilityVSAvoidtissue proximity indicator consistency
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The system uses a calibration electrode connected to both main and auxiliary signal processing units to generate reference TPIs. The reinforcement learning agent continuously compares TPIs from both units and adjusts the auxiliary unit's computation parameters to minimize the difference between its TPIs and the reference TPIs from the main unit, creating a closed-loop feedback system that ensures consistency

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The auxiliary signal processing unit replicates the TPI computation functionality of the main unit by using a reinforcement learning agent to learn and copy the main unit's computation methodology. The agent trains the auxiliary unit to produce TPIs that match the main unit's output, effectively creating a consistent copy of the computation process

Inventive Principle:
Principle #26Copying

2Measurement precision

If reinforcement learning is used to train auxiliary signal processing unit to match main unit's TPI calculations, then TPI consistency is improved, but system complexity increases

Engineering Contradiction:
Improvetissue proximity indicator consistencyVSAvoidsignal processing system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

A calibration electrode serves as an intermediary element that connects both main and auxiliary signal processing units. This single physical component enables the system to generate reference TPIs that both units can compare against, facilitating the reinforcement learning training process without requiring complex inter-unit communication infrastructure

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The auxiliary signal processing unit performs self-calibration and self-adjustment through the reinforcement learning agent, which automatically tunes its computation parameters to match the main unit's output. The system uses its own operational data from the calibration electrode to train and improve itself, reducing the need for external calibration procedures

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP4014847B1Accurate tissue proximity
Publication Date: 2025.12.31 BIOSENSE WEBSTER (ISRAEL) LTD
  • EP4014847B1 patent drawingFigure 1
  • EP4014847B1 patent drawingFigure 2
  • EP4014847B1 patent drawingFigure 3

AI summary

In one embodiment, a method to find tissue proximity indications includes inserting a catheter into a body part of a living subject such that electrodes of the catheter contact tissue at respective locations within the body part, receiving signals provided by the electrodes, selectively rewarding and penalizing a reinforcement learning agent over reinforcement learning exploration phases to learn at least one tissue proximity policy responsively to at least one of the received signals, applying the reinforcement learning agent in reinforcement learning exploitation phases to find respective tissue-proximity actions to be taken that maximize respective expected rewards responsively to the at least one tissue proximity policy, and providing respective derived tissue-proximity indications of proximity of a given one of the electrodes with the tissue responsively to the found respective tissue-proximity actions.