Electrical Impedance Tomography Reconstruction With Attention Autoencoders

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing electrical impedance tomography (EIT) methods suffer from low spatial resolution and high sensitivity to noise due to the non-linearity of the inverse problem, leading to inefficient image reconstruction, particularly in high-pressure and high-temperature environments.

Innovation Solution

A neural network-based method using an autoencoder with attention modules and dropout layers, combined with predefined electrode excitation, to enhance feature extraction and spatial resolution by focusing on important regions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional iterative mathematical algorithms (Tikhonov regularization, linear backprojection, Gauss-Newton method) are used to solve the nonlinear inverse problem of EIT, then the method can reconstruct the impedance map inside the object, but the spatial resolution is low and the computation time increases exponentially with the data

Engineering Contradiction:
Improvespatial resolutionVSAvoidcomputation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces traditional iterative mathematical algorithms with a neural network-based computational system. The neural network learns the mapping from boundary measurements to internal impedance distribution through training, substituting the mechanical iterative solving process with a learned model that provides both speed and accuracy improvements.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The neural network is trained in advance using simulated EIT data to learn the inverse problem solution. This preliminary training phase allows the network to capture complex nonlinear relationships, so that during actual EIT operation, the reconstruction can be performed rapidly without iterative computation.

Inventive Principle:
Principle #10Preliminary action

2Reliability

If traditional iterative algorithms are used, then the inverse problem can be solved, but the method shows high sensitivity to noise during measurements

Engineering Contradiction:
Improvenoise sensitivityVSAvoidimage quality
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The neural network replaces the noise-sensitive iterative mathematical algorithms with a robust learned model. During training, the network learns to handle measurement noise and variations, making the reconstruction process more reliable and less sensitive to noisy measurements in practical applications.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Ease of operation

If electrodes are placed on the outer surface of a metal object, then non-intrusive measurement is achieved, but the electrodes must pass through the wall and be in contact with the fluid

Engineering Contradiction:
Improvenon-intrusive measurementVSAvoidelectrode configuration
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The patent develops a universal neural network model that can handle different electrode configurations and object geometries. The model is trained to be adaptable to various measurement setups, including surface-mounted electrodes on metal objects, reducing the need for complex custom electrode arrangements for different applications.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

The method achieves improved spatial resolution and reduced noise sensitivity, resulting in high-quality image reconstruction of electrical properties, including conductivity and permittivity.

Implementation Method 1

Electrical Impedance Tomography (EIT) is a non-invasive, non-destructive technique that allows for real-time, continuous visualization of the interior of an object by measuring its electrical properties (potential and electric current) at its surface

Methodology Applied
Scientific EffectElectrical Impedance Tomography: Electrical Impedance Tomography

Implementation Method 2

The neural network comprises: an autoencoder comprising several levels L1,...,LJ, each having at least one convolution layer and at least one dropout layer, and an attention module comprising attention gates AG1,..., AGJ-1

Methodology Applied
Scientific EffectNeural network processing:

Implementation Method 3

The impedance map inside the object is reconstructed by solving the associated inverse problem, in order to recover the internal distribution of conductivity and permittivity, according to Ohm's law

Methodology Applied
Scientific EffectOhm's law: Ohm's Law

Data Source

PatentEP4647004A1Method for reconstruction of the distribution of electrical properties of materials using electrical impedance tomography
Publication Date: 2025.11.12 COMMISSARIAT A LENERGIE ATOMIQUE ET AUX ENERGIES ALTERNATIVES
  • EP4647004A1 patent drawingFigure 1
  • EP4647004A1 patent drawingFigure 2
  • EP4647004A1 patent drawingFigure 3a

AI summary

The invention relates to a method for reconstructing the distribution of electrical properties of at least one material of a body comprising a cylindrical part containing a fluid, using a neural network and electrical value data of the body previously measured by electrical impedance tomography using electrodes arranged around a periphery of the cylindrical part of the body, each electrode having been excited by a potential of predefined shape, the neural network comprising: - an autoencoder comprising several levels (L1,...,LJ) each comprising at least one convolution layer and at least one dropout layer, and - an attention module comprising attention gates (AG1,..., AGJ-1), each having an associated attention signal (gj).