Towards a Fast and Accurate EIT Inverse Problem Solver: A Machine Learning Approach

Abstract

Different industrial and medical situations require the non-invasive extraction of information from the inside of bodies. This is usually done through tomographic methods that generate images based on internal body properties. However, the image reconstruction involves a mathematical inverse problem, for which accurate resolution demands large computation time and capacity. In this paper we explore the use of Machine Learning to develop an accurate solver for reconstructing Electrical Impedance Tomography images in real-time. We compare the results with the Iterative Gauss-Newton and the Primal Dual Interior Point Method, which are both largely used and well-validated solvers. The approaches were compared from the qualitative as well as the quantitative viewpoints. The former was focused on correctly detecting the internal body features. The latter was based on accurately predicting internal property distributions. Experiments revealed that our approach achieved better accuracy and Cohen's kappa coefficient (97.57% and 94.60% respectively) from the qualitative viewpoint. Moreover, it also obtained better quantitative metrics with a Mean Absolute Percentage Error of 18.28%. Experiments confirmed that Neural Networks algorithms can reconstruct internal body properties with high accuracy, so they would be able to replace more complex and slower alternatives.

Application of Deep Neural Network to the Reconstruction of Two-Phase Material Imaging by Capacitively Coupled Electrical Resistance Tomography

“…In addition, the studies of artificial neural networks (ANNs), another popular DNN model, have also attracted lots of interest for ET application. Fernández-Fuentes et al developed an ANN-based inverse problem solver for EIT, which takes the boundary measurements as the input and generates the conductivity value of each mesh of triangular elements of the image [32]. Rymarczyk et al compared some machine learning algorithms for industrial ET, including the ANN, LARS, and elastic net methods, and they used a set of trained subsystems to generate the value of each pixel of the image in parallel [33].…”

Automatic Diabetic Retinopathy Diagnosis Using Adaptive Fine-Tuned Convolutional Neural Network

“…The results of the ResNet152, DensNet121, and ResNeSt50 models, fine-tuned using the same training set and evaluated using the same testing set as for from EyePACS and APTOS2019, are also shown in Table 2. The results show that DeepPCANet-4 and DeepPCANet-16 outperform RsenNet152, DesneNet121, and ResNeSt50 on both datasets in terms of all metrics; in particular, in both cases, the sensitivity and Cohen's Kappa are higher than those of RsenNet152, DesneNet12, and ResNeSt50, Cohen's Kappa is considered more robust statistical measure than accuracy [48,49]. The DeepPCANet-4 has the lowest number of FLOPs (1.36 G) and learnable parameters ( 63.7 K) among all competing models, as shown in Table 2.…”

Spatially-Adaptive Variational Reconstructions for Linear Inverse Electrical Impedance Tomography

“…In particular, particle swarm optimization has been used in the past to improve the reconstruction quality while reducing the number of iterations [30] and to optimize the current injection pattern [38]. More recently, neural networks have been extensively applied to increase image quality and reduce reconstruction time [10,11]. The D-Bar method has lately gained traction for EIT imaging, thanks to a quick non-iterative approach relying on the Fourier and inverse transform [13] which has also been combined with deep learning techniques to improve reconstruction quality [12].…”