There is a strong need for a non-invasive measurement technique that is capable of accurately identifying the physiological condition change or heterogeneity of subcutaneous adipose tissue (SAT) by localizing the abnormalities within the compartment. This paper aims to investigate the feasibility of Electrical Impedance Tomography (EIT) to assess the interstitial fluid in subcutaneous adipose tissue as an enhancement method of bioelectrical impedance spectroscopy (BIS). Here, we demonstrate the preliminary result of EIT with a wearable 16 electrodes sensor. The image-based reference EIT with fat weighted threshold method is proposed. In order to evaluate the performance of our novel method, a physiological swelling experiment is conducted, and Multi-Frequency Bioelectrical Impedance Analysis (MFBIA) is also applied as a comparison with EIT results. The experimental results showed that the proposed method was able to distinguish the physiological swelling condition and effectively to remove the unexpected background noise. Furthermore, the conductivity variation in the subcutaneous layer had a good correlation with extracellular water volume change from MFBIA data; the correlation coefficient R2 = 0.927. It is concluded that the proposed method provides a significant prospect for SAT assessment.
- Subcutaneous adipose tissue assessment
- Electrical Impedance Tomography
- Wearable sensor
- Physiological swelling
- Multi-Frequency Bioelectrical Impedance Analysis
Variation of the physiological condition of subcutaneous adipose tissue (SAT) has been diagnosed as a part of metabolic syndrome (MetS) , lymphedema , Dercum’s disease , and obesities . Multi-Frequency Bioelectrical Impedance Analysis (MFBIA) or Bio-Electrical impedance spectroscopy (BIS) is proposed as a rapid tool for SAT assessment as it has demonstrated a significant impedance difference between healthy and abnormal SAT [2,4,5]. MFBIA uses 8 resistive electrodes that attached on the ankle and wrist through multi-frequency impedance measurements on the whole body and its portability and noninvasive characteristic makes MFBIA more attractive for routinely physiology measurement .
MFBIA is used based on the fact that the current flow through biological tissue is frequency-dependent . At low frequency, the membrane cell acts as high impedance. It makes the current flows only at extracellular fluid. Meanwhile, at the increase of frequency, the impedance of the cell membrane is decreasing and makes the current possible to penetrate into the intracellular fluid. Several Anthropometric parameters need to be considered to relate the measured impedance with the water content at interstitial fluid, such as body weight, body height, gender, age, postures . MFBIA technique assumes a human body as a single object to be measured in which only single volumes of particular compartments are measurable . However, there are a number of potential limitations to MFBIA, such as: an inability to assess heterogeneity of SAT [9,10], an inability to differentiate accurately edema type and albumin accumulation location [9,10], and the performance depends on the accuracy of the controlled clinical condition . These issues make a shortcoming and leads to the modification of MFBIA to localize body segment impedance evaluation [10,12].
Our objective is, therefore, to develop the non-invasive and low-cost medical equipment using the impedance measurement with several multi-projection, i.e., using Electrical Impedance Tomography (EIT). EIT is a noninvasive medical imaging method that can show the conductivity distribution of the human body [13,14,15]. In EIT, the domain (e.g., bone, muscle) is excited electrically by the electrical currents injected through the electrodes, which are discretely placed along the domain boundary (e.g., skin). The exciting electric field in the domain is determined in accordance with the geometry, the boundary conditions, and the internal distribution of the electrical properties such as conductivity. The excited voltages on the electrodes can be measured, and then used for the reconstruction of the internal conductivity distribution. As the human body has been demonstrated to have different electrical properties , a cross-sectional view can be reconstructed with EIT to identify any abnormalities; on the subject of the early detection of lymphedema, which is the protein accumulation in subcutaneous adipose tissue. The relative value is mostly used for the image reconstruction, i.e., a homogeneous data is taken as a reference to image the difference owing to the existence of the abnormality. However, it is troublesome considering the homogeneous reference data cannot be taken in the biomedical instrumentation as they always vary over different subjects, and the human body is composed of complex properties.
This paper aims to investigate the feasibility of EIT to assess interstitial fluid in subcutaneous adipose tissue as a preliminary step in lymphedema early-detection. We measure the right lower leg to examine to what extent our EIT system with novel threshold technique has a quality to show the morphologically same image as a sample of the MRI image. A physiological swelling experiment is then carried out to scrutinize the relationship between MFBIA and our EIT system. In order to effectively focus on the varying water contents, image-based reference EIT is proposed.
In EIT, as typically the given domain Ω is assumed as quasistatic condition, which has a smooth boundary
As the human body is very complex and has inhomogeneous electrical properties, there are problems not only to refer a homogeneous reference conductivity but also obtaining the prior conductivity distribution. Thus, the problems can be compensated by considering the first measurement voltage data as the initial conductivity distribution and subtracting the conductivity distribution at a different time point relative to the initial conductivity. The initial conductivity distribution
Schematic flow of image-based reference EIT.
Considering the fact that skin is a highly resistive material and the human body has a complex conductivity distribution, we first apply the nonlinear absolute Gauss-Newton imaging , where we set
We implemented the quasi-adjacent stimulation and measurement pattern to accelerate the sensitivity near the boundary and the sensitivity map area  as the subcutaneous layer is located in the peripheral side of the domain as depicted in Figure 1.
In order to compare the quality of the proposed method, a time-difference EIT was carried out , where the reference was taken from the simulation data
As the human body is not a precise circle and the shape vary significantly over the subjects, the sensor configuration is often troublesome, especially for the human subjects. We, therefore, developed a wearable EIT sensor, which has a row of 16 equidistant silver yarn electrodes and these electrodes are attached on a flexible cloth made up of polyester. The electrodes are shielded by thermoplastic polyurethane (TPU) as a countermeasure for the high contact impedance . It is worth mentioning that this sensor is excellent in adhesion to the skin so that contact impedance can be reduced. Furthermore, this sensor has the capability to adapt its size and shape to any subject with less discomfort.
We used an impedance analyzer (Impedance analyzer IM3570, HIOKI) and a multiplexer (multiplexer 34970A, Agilent) as the data acquisition system, which is connected to the wearable EIT sensor and a PC using coaxial cable and USB cable, respectively, as can be seen in Figure 2. The PC has installed LabVIEW (LabVIEW 2015, National Instruments) in order to visualize the spectrum and control the data acquisition system. The amplitude of the sinusoidal current is set as 1 mA, and 1 kHz frequency is selected to focus on the extracellular fluid as the low-frequency current penetrates only in the interstitium [23,24].
A multi-frequency impedance body composition analyzer (InBodyS10, InBody Japan Inc.) was used to measure extracellular fluid volume in the right leg to acquire the varying water contents, which pertain specifically to the conductivity change.
These experiments were carried out repeatedly for 5 days to attain the reliability of the experimental data. As MFBIA data is prone to be influenced by physiological information, the subject was strictly controlled regarding life activity such as wake-up time and breakfast menu throughout the experiment days.
Graduate School of Science and Engineering, Chiba University, Chiba, Japan, approved the study, and informed consent was obtained from the subjects. The research related to human use has been complied with all relevant national regulations, institutional policies and in accordance with the tenets of the Helsinki Declaration, and has been approved by the authors’ institutional review board or equivalent committee.
The normalized conductivity distribution reconstructed by absolute EIT on each day of one individual at the first measurement is shown in figure 4 (a), which demonstrably showed expected results. Whereas figure 4 (b) and (c) represent the time series of the reconstructed difference conductivity distribution on each day at t = 30 mins, 60 mins, and 90 mins with the proposed method, and with the conventional time difference EIT, respectively. Figure 4 (a) reveals the predicted subcutaneous layer based on the proposed fat weighted threshold method. It has satisfactory quality with regard to the morphological relationship with an MRI image. The reconstructed SAT layer can be identified well throughout the experiment days. Although some noise can still be observed in Figure 4 (b), in which the proposed method was used, conductivity at the SAT layer is increased as time went by. This indicates water increase in the layer. On the other hand, there is almost no change in Figure 4 (c), in which the conventional time difference EIT was used; therefore, time variation cannot be observed. No change can be seen in the muscle and bone area in both methods and the impedance artefacts in the periphery arise from the influence of the measurement and the modelling errors. Similar results were acquired over the course of experiment days.
Figure 5 shows the average temporal variation of normalized segmental conductivity in the predicted subcutaneous layer (white part in Figure 1) and segmental extracellular water volume in the right leg from EIT and MFBIA data, where the average of 5 days was taken at each time point. The effect of long-term standing and exercise can be seen clearly. The correlation coefficient of EIT and MFBIA is achieved to
The images of the SAT layer were successfully reconstructed by absolute EIT; therefore, the fat weighted threshold has a good potential to assess the interstitial fluid in the subcutaneous adipose tissue. In fact, even though there is much background noise in the center of the images before the threshold process, this proposed method was able to reduce it sufficiently. On the other hand, as the thresholding result is calculated based on the average and the standard deviation, it may be affected when the signal-to-noise ratio (SNR) of the measurement voltage becomes lower . Moreover, this thresholding can be performed based on the elements selection of the FEM mesh, hence, it highly depends on the modelling.
We realize that the MRI data used in Figure 1 is not always available for routinely monitoring by using EIT. We used one MRI data as a piece of prior known information in order to consider the inaccuracy of the reconstruction algorithm and the boundary geometry effect on the reconstructed image. Considering this prior known information, then we can obtain the average and standard deviation of SAT
In this study, image-based reference EIT was applied to see the variation over reconstructed conductivity distribution, where the reference was obtained using the absolute EIT at the 1st measurement to focus on the relative value as time goes by. Even though the images changed and the conductivity increased in most cases, there is a problem in the stability.
One of the reasons is the mathematical error in Equation (6). It is not enough just to subtract the value and should be compensated by introducing the penalty term or the weighted function . The other reason for the lack of stability is measurement error. The subject wore the wearable sensor in the same position throughout the experiments days. It is known that even very small changes such as moisture on the skin significantly change the conductivity. Thus, these problems should be calibrated mathematically and empirically.
Conventional time difference EIT was also implemented to compare the results, where the reference was taken from the simulation data. One of the reasons why the conductivity change was too small to see the difference in the conventional time difference EIT, arises from a huge span of the magnitudes of the impedance. In the simulation, the ideal case was considered, where the electrodes were properly attached to the skin and the currents flowed through only on the two-dimensional surface. In addition to that, the contact impedance was set to a relatively small value. In the practical case, these conditions should be considered well, e.g. the electric currents are not in general confined to a two-dimensional plane  and the contact impedance is relatively large especially in the medical applications . This same mistake about current goes for the difference of the impedance magnitude of EIT and MFBIA. The former has impedance different by four orders of magnitude from that of the latter. However, it does not necessarily mean those two data sets cannot be compared as the change of the extracellular matrix due to the physiological swelling generally affects the lower leg.
Although EIT data and MFBIA data had a good correlation in Figure 5, EIT images failed to sufficiently elucidate difference over time. Concerning high resistivity in the human body, especially the skin, the injected current might penetrate in an undesirable pathway. Figure 6 shows the magnitude variation of impedance at 0 min and 60 min on Day 1. Impedance decreased due to the rise in the extracellular water volume. In the quasi-adjacent stimulation, the magnitude should be the biggest at the 7 + 12 * (
In this paper, we investigated the feasibility of our EIT system to obtain a reconstructed image, which has a good correlation with an MRI image and the MFBIA data. We carried out a physiological experiment using a wearable 16 electrodes sensor and acquired the image using fat weighted threshold and image-based reference EIT. The absolute EIT images with the proposed threshold technique were able to visualize the subcutaneous adipose tissue (SAT) layer, which is morphologically similar to an MRI image of the right lower leg. Besides, the time series of EIT images have a high agreement with the MFBIA data regarding varying water content of the subcutaneous fat tissues in the lower leg due to the effect of the physiological swelling over five experiment days; these achieved a correlation coefficient