Table of contents

With the advent of miniaturized sensing technology, which can be body-worn, it is now possible to collect and store data on different aspects of human movement under the conditions of free living. This technology has the potential to be used in automated activity profiling systems which produce a continuous record of activity patterns over extended periods of time. Such activity profiling systems are dependent on classification algorithms which can effectively interpret body-worn sensor data and identify different activities. This article reviews the different techniques which have been used to classify normal activities and/or identify falls from body-worn sensor data. The review is structured according to the different analytical techniques and illustrates the variety of approaches which have previously been applied in this field. Although significant progress has been made in this important area, there is still significant scope for further work, particularly in the application of advanced classification techniques to problems involving many different activities.

The purpose of this study was to compare the monopolar electromyographic (EMG) amplitude versus isometric force relationships from three signal processing methods (raw versus notch filtering versus adaptive filtering). Seventeen healthy subjects (mean ± SD age = 24.6 ± 4.3 yr) performed incremental isometric muscle actions of the dominant leg extensors in 10% increments from 10% to 100% of the maximum voluntary contraction (MVC). During each muscle action, a monopolar surface EMG signal was recorded from the vastus lateralis and processed with the three signal processing methods. The linear slope coefficients for the EMG amplitude versus isometric force relationships were equivalent for the three signal processing methods and correlated (r = 0.997–0.999). However, the mean amplitude values for the notch-filtered signals were less than those for the raw and adaptive-filtered signals. Thus, adaptive filtering may be the best method for removing electromagnetic noise from monopolar surface EMG signals.

Due to the importance of motility in a number of gastrointestinal disorders, efforts have been made to evaluate both gastric motility counterparts: electrical activity and mechanical activity. The present work aimed to propose a new approach, associating AC biosusceptometry (ACB) and electrogastrography (EGG), to noninvasively monitoring mechanical and electrical gastric activity, respectively. Fourteen volunteers ingested a test meal and their gastric activity was evaluated by EGG and ACB at a baseline and after 20 mg of i.v. hyoscine butylbromide. ACB and EGG showed a similar signal pattern and high temporal correlation. Hyoscine butylbromide decreased the mechanical motility index (MI) by 50.9%, while for electrical MI the reduction was 36.5%. Delayed times to onset (mean ± SD: 50 ± 15 versus 40 ± 20 s; P < 0.01) and the inhibitory effect (16 ± 4 versus 14 ± 5 min; P < 0.01) were calculated for ACB and EGG, respectively. ACB and EGG emerged due to their interesting nature, noninvasiveness and low cost to evaluate gastric motility. Our approach associating ACB and EGG allowed monitoring and quantification of the effects of an anticholinergic drug in gastric electrical activity and contractile activity in humans.

We propose dynamical models for pulsatile flow and head estimation in an implantable rotary blood pump. Pulsatile flow and head data were obtained using a circulatory mock loop where fluid solutions with different values of viscosities were used as a blood analogue with varying haematocrit (HCT). Noninvasive measurements of power and pump speed were used with HCT values as inputs to the flow model while the estimated flow was used with the speed as inputs to a head estimation model. Linear regression analysis between estimated and measured flows obtained from a mock loop resulted in a highly significant correlation (R² = 0.982) and a mean absolute error (e) of 0.323 L min⁻¹, while for head, R² = 0.933 and e = 7.682 mmHg were obtained. R² = 0.849 and e = 0.584 L min⁻¹ were obtained when the same model derived in the mock loop was used for flow estimation in ex vivo porcine data (N = 6). Furthermore, in the steady state, the solution of the presented flow model can be described by a previously designed and verified static model. The models developed herein will play a vital role in developing a robust control system of the pump flow coping with changing physiological demands.

Integration of electroencephalography (EEG) and functional magnetic imaging (fMRI) resonance will allow analysis of the brain activities at superior temporal and spatial resolution. However simultaneous acquisition of EEG and fMRI is hindered by the enhancement of artifacts in EEG, the most prominent of which are ballistocardiogram (BCG) and electro-oculogram (EOG) artifacts. The situation gets even worse if the evoked potentials are measured inside MRI for their minute responses in comparison to the spontaneous brain responses. In this study, we propose a new method of attenuating these artifacts from the spontaneous and evoked EEG data acquired inside an MRI scanner using constrained independent component analysis with a priori information about the artifacts as constraints. With the proposed techniques of reference function generation for the BCG and EOG artifacts as constraints, our new approach performs significantly better than the averaged artifact subtraction (AAS) method. The proposed method could be an alternative to the conventional ICA method for artifact attenuation, with some advantages. As a performance measure we have achieved much improved normalized power spectrum ratios (INPS) for continuous EEG and correlation coefficient (cc) values with outside MRI visual evoked potentials for visual evoked EEG, as compared to those obtained with the AAS method. The results show that our new approach is more effective than the conventional methods, almost fully automatic, and no extra ECG signal measurements are involved.

The analysis of oxygen desaturations is a basic variable in polysomnographic studies for the diagnosis of sleep apnea. Several algorithms operating in the time domain already exist for sleep apnea detection via pulse oximetry, but in a disadvantageous way—they achieve either a high sensitivity or a high specificity. The aim of this study was to assess whether an alternative analysis of arterial oxygen saturation (SaO₂) signals from overnight pulse oximetry could yield essential information on the diagnosis of sleep apnea hypopnea syndrome (SAHS). SaO₂ signals from 117 subjects were analyzed. The population was divided into a learning dataset (70 patients) and a test set (47 patients). The learning set was used for tuning thresholds among the applied Poincaré quantitative descriptors. Results showed that the presence of apnea events in SAHS patients caused an increase in the SD₁ Poincaré parameter. This conclusion was assessed prospectively using the test dataset. 90.9% sensitivity and 84.0% specificity were obtained in the test group. We conclude that Poincaré analysis could be useful in the study of SAHS, contributing to reduce the demand for polysomnographic studies in SAHS screening.

The bioimpedance technique provides a safe, low-cost and non-invasive alternative for routine monitoring of lung fluid levels in patients. In this study we have investigated the feasibility of bioimpedance measurements to monitor pleural effusion (PE) patients. The measurement system (eight-electrode thoracic belt, opposite sequential current injections, 3 mA, 20 kHz) employed a parametric reconstruction algorithm to assess the left and right lung resistivity values. Bioimpedance measurements were taken before and after the removal of pleural fluids, while the patient was sitting at rest during tidal respiration in order to minimize movements of the thoracic cavity. The mean resistivity difference between the lung on the side with PE and the lung on the other side was −48 Ω cm. A high correlation was found between the mean lung resistivity value before the removal of the fluids and the volume of pleural fluids removed, with a sensitivity of −0.17 Ω cm ml⁻¹ (linear regression, R = 0.53). The present study further supports the feasibility and applicability of the bioimpedance technique, and specifically the approach of parametric left and right lung resistivity reconstruction, in monitoring lung patients.

In this study we examine the baseline characteristics of facial skin temperature, as measured by dynamic infrared thermal imaging, to gauge its potential as a physiological access pathway for non-verbal individuals with severe motor impairments. Frontal facial recordings were obtained from 12 asymptomatic adults in a resting state with a high-end infrared thermal imaging system. From the infrared thermal recordings, mean skin temperature time series were generated for regions of interest encompassing the nasal, periorbital and supraorbital areas. A 90% bandwidth for all regions of interest was found to be in the 1 Hz range. Over 70% of the time series were identified as nonstationary (p< 0.05), with the nonstationary mean as the greatest contributing source. Correlation coefficients between regions were significant (p< 0.05) and ranged from values of 0.30 (between periorbital and supraorbital regions) to 0.75 (between contralateral supraorbital regions). Using information measures, we concluded that the greatest degree of information existed in the nasal and periorbital regions. Mutual information existed across all regions but was especially prominent between the nasal and periorbital regions. Results from this study provide insight into appropriate analysis methods and potential discriminating features for the application of facial skin temperature as a physiological access pathway.