Image reconstruction in EIT with unreliable electrode data using random sample consensus method

Abstract

In electrical impedance tomography (EIT), it is important to acquire reliable measurement data through EIT system for achieving good reconstructed image. In order to have reliable data, various methods for checking and optimizing the EIT measurement system have been studied. However, most of the methods involve additional cost for testing and the measurement setup is often evaluated before the experiment. It is useful to have a method which can detect the faulty electrode data during the experiment without any additional cost. This paper presents a method based on random sample consensus (RANSAC) to find the incorrect data on fault electrode in EIT data. RANSAC is a curve fitting method that removes the outlier data from measurement data. RANSAC method is applied with Gauss Newton (GN) method for image reconstruction of human thorax with faulty data. Numerical and phantom experiments are performed and the reconstruction performance of the proposed RANSAC method with GN is compared with conventional GN method. From the results, it can be noticed that RANSAC with GN has better reconstruction performance than conventional GN method with faulty electrode data.

Noise-robust bioimpedance approach for cardiac output measurement

“…The corresponding kernel strides were (2,3) for the first 2 layers and (1,3) for the third layer. The output sizes of the convolutional layers are (18,75), (9,25) and (9,9), respectively. All convolutional layer activations were rectified linear units (ReLU), preceded by mini-batch normalization.…”

“…'bad' electrodes. There have been a number of studies that have attempted to address this problem by 1) compensating for bad electrodes by simultaneously solving for both internal conductivity and the electrode contact impedances [14,15,16], 2) automatically detecting bad electrodes and simply removing them [17,18], or 3) automatically detecting bad electrodes and adjusting their measurements in some way to mitigate their effect [19,20,21]. It was concluded in [14] that the simultaneous reconstruction of conductivity and contact impedance was not practical in situations without a homogeneous (or known) measured reference frame.…”

Surface reconstruction method for measurement data with outlier detection by using improved RANSAC and correction parameter

“…The RANSAC algorithm, presented by Fischler and Bolles, 32 can robustly estimate the model parameters. It has been applied to many fields, especially in computer vision, 33,34 due to good performance in handling the data with a tremendous level of outliers and remarkably simple structure. The main idea of the RANSAC is to calculate parameters of hypothesized model by randomly sampling the subsample from the entire dataset, 35 and then the model is performed on the entire dataset.…”

Compensation of volatile-distributed current due to variance of the unknown contact impedance in an electrical impedance tomography sensor

“…In order to overcome this difficulty, simultaneous compensation of the unknown contact impedance variance due to the volatile-distributed current I 1 is necessary at each position of the transmitter and receiver electrodes [17]. Even though some authors have proposed innovative methods to compensate for the unknown contact impedance variance, such as the uniform conducting medium method [18,19], the compensation error method [20], the statistical estimation method [21][22][23][24], skin preparation (cleaning the skin, applying conductive gel and using bandages) [14] and covering the electrode with a resistive material [25,26], none of these methods consider the accuracy of the frequency-dependent behavior in the reconstructed images. Therefore, simultaneous compensation of the unknown contact impedance variance due to the volatile-distributed current I 1 is still a significant challenge in actual EIT applications.…”