The relation between a biological process and the changes in passive electrical properties of the tissue is often non-linear, in which developing prediction models based on bioimpedance spectra is not trivial. Relevant information on tissue status may also lie in characteristic developments in the bioimpedance spectra over time, often neglected by conventional methods. The aim of this study was to explore possibilities in machine learning methods for time series of bioimpedance spectra, where we used organ ischemia as an example. Based on published data on the development of the bioimpedance spectrum during liver ischemia, a simulation model was made and used to generate sets of synthetic data with different levels of organ-to-organ variation, measurement noise and drift. Three types of artificial neural networks were employed in learning to predict the ischemic duration, based on the simulated datasets. The simulated prediction performance was very dependent on the amount of training examples, the organ-to-organ variation and the selection of input variables from the bioimpedance spectrum. The performance was also affected by noise and drift in the measurement, but a recurrent neural network with long short-term memory units could obtain good predictions even on noisy and drifting measurements. This approach may be relevant for further exploration on several applications of bioimpedance having the purpose of predicting a biological state based on spectra measured over time.
This paper gives a basic overview of relevant statistical methods for the analysis of bioimpedance measurements, with an aim to answer questions such as: How do I begin with planning an experiment? How many measurements do I need to take? How do I deal with large amounts of frequency sweep data? Which statistical test should I use, and how do I validate my results? Beginning with the hypothesis and the research design, the methodological framework for making inferences based on measurements and statistical analysis is explained. This is followed by a brief discussion on correlated measurements and data reduction before an overview is given of statistical methods for comparison of groups, factor analysis, association, regression and prediction, explained in the context of bioimpedance research. The last chapter is dedicated to the validation of a new method by different measures of performance. A flowchart is presented for selection of statistical method, and a table is given for an overview of the most important terms of performance when evaluating new measurement technology.
Impedance cardiography (ICG) is a non-invasive method of hemodynamic measurement, mostly known for estimation of stroke volume and cardiac output based on characteristic features of the signal. Compared with electrocardiography, the knowledge on the morphology of the ICG signal is scarce, especially with respect to age-dependent changes in ICG waveforms. Based on recordings from ten younger (20–29 years) and ten older (60–79) healthy human subjects after three different levels of physical activity, the typical interbeat ICG waveforms were derived based on ensemble averages. Comparison of these waveforms between the age groups indicates the following differences: a later initial upward deflection for the younger group, an additional hump in the waveform from many older subjects not presented in the younger group, and a more pronounced second wave in the younger group. The explanation for these differences is not clear, but may be related to arterial stiffness. Further studies are suggested to determine whether these morphological differences have clinical value.
A circuit is presented that enables measurement of skin electrical conductance, susceptance, and potential simultaneously beneath the same monopolar electrode. Example measurements are shown to confirm the function of the circuit. The measurements are also in accordance with earlier findings that changes in skin conductance and potential do not always correspond and hence contain unique information.
Alternating current methods have the potential to improve the measurement of electrodermal activity. However, there are pitfalls that should be avoided in order to perform these measurements in a correct manner. In this paper, we address issues like the choice of measurement frequency, placement of electrodes and the kind of electrodes used. Ignoring these factors may result in loss of measurement sensitivity or erroneous measurements with artifacts that contain little or no physiological information.
Accurate assessment of experienced pain is a well-known problem in the clinical practices. Therefore, a proper method for pain detection is highly desirable. Electrodermal activity (EDA) is known as a measure of the sympathetic nervous system activity, which changes during various mental stresses. As pain causes mental stress, EDA measures may reflect the felt pain. This study aims to evaluate changes in skin conductance responses (SCRs), skin potential responses (SPRs), and skin susceptance responses (SSRs) simultaneously as a result of sequences of electrical (painful) stimuli with different intensities. EDA responses as results of painful stimuli were recorded from 40 healthy volunteers. The stimuli with three different intensities were produced by using an electrical stimulator. EDA responses significantly changed (increased) with respect to the intensity of the stimuli. Both SCRs and SSRs showed linear relationship with the painful stimuli. It was found that the EDA responses, particularly SCRs (p < 0.001) and SSRs (p = 0.001) were linearly affected by the intensity of the painful stimuli. EDA responses, in particular SCRs, may be used as a useful indicator for assessment of experienced pain in clinical settings.
Sixteen volunteers each drank 700 ml sugar-containing soft drink during two successive periods and the blood sugar was measured at 10 min intervals together with electrical impedance spectroscopy and near infrared spectroscopy (NIR). A maximum correlation of 0.46 was found for the electrical measurements but no clear separation between low and high blood glucose levels were found in the NIR measurements. The latter was attributed to the experimental design where the NIR probe was removed from the skin between each measurement.