Showing papers presented at "IEEE International Symposium on Medical Measurements and Applications in 2021"

Analysis of Galvanic Skin Response to Acoustic Stimuli by Wearable Devices

Abstract: This paper evaluates the Galvanic Skin Response (GSR) signals to three different acoustic stimuli, collected through a commercial wearable device (Empatica E4) by a group of healthy individuals at rest. The collected GSR signals are analyzed depending on the overall number of peaks in the time domain, as well as on the Power Spectral Density (PSD) in the frequency domain, where three bands of interest are identified. In particular, the proposed paper aims to highlight features related to acoustic stimulation. The outcomes show that the GSR signal presents a higher number of GSR peaks in case of unpleasant and neutral stimuli than in case of pleasant stimulus. Moreover, a larger band than the bands typically considered in literature should be observed in the frequency domain, in order to include meaningful PSD of the GSR signal.

Deterministic Compressed Sensing of heart sound signals

Abstract: This paper proposes the use of the Deterministic Binary Block Diagonal (DBBD) matrix as sensing matrix for compressed sensing of heart sound signals. The use of a deterministic matrix has the advantage of not requiring the generation of random numbers in the acquisition node. Moreover, the DBBD matrix has a very low computational complexity at the compression side, as it only requires a sum of the samples. In the paper, the DBBD sensing matrix is used in combination with the Discrete Cosine Transform and the Mexican Hat wavelet to compress and reconstruct heart sound signal obtained from the PhysioNet database. The results show a lower value of the Percent Root Mean Square Difference compared to that obtained by the random sensing matrix previously used in the literature for heart sound signals.

Distinguishing Stroke patients with and without Unilateral Spatial Neglect by means of Clinical Features: a Tree-based Machine Learning Approach

Abstract: Unilateral Spatial Neglect is a cognitive impairment of neuropsychological interest that is a consequence of stroke able to influence negatively the rehabilitation outcome of patients with stroke. The aim of the study is to explore the feasibility of machine learning to classify stroke patients with and without unilateral spatial neglect using clinical features. We performed the study using a machine learning approach by means the following tree-based algorithms: Decision Tree, Random Forest, Rotation Forest, AdaBoost of decision stumps and Gradient Boost tree using six clinical features both numerical and nominal: Montreal Cognitive Assessment, Functional Independence Measure scale, Barthel Index, aetiology, site of brain lesion and presence of hemiparesis at lower limbs. Tree-based Machine learning analysis achieved interesting results in terms of evaluation metrics scores; the best algorithm was Random Forest with an Accuracy, Sensitivity, Specificity, Precision and Area under the Receiver Operating Characteristic curve equal to 0.92, 0.83, 1.00, 1.00, 0.95 respectively. The study demonstrated the proposed combination of clinical features and algorithms represents a valuable approach to automatically classify stroke patients with and without Unilateral Spatial Neglect. The future investigations on enriched datasets will further confirm the potential application of this methodology as prognostic support to be chosen among those already implemented in the clinical field.

Repeatability and reproducibility in the breathability measurement of surgical masks

Abstract: The pandemic COVID-19 is still requiring several countermeasures to adopted in order to decrease the virus spread. Among others, face masks have shown their potentiality to reduce the person-to-person transmission. For this reason, several research groups and/or industries focused their supply chain in the production of innovative materials for the face mask design. Considering the surgical mask, several parameters have to be checked before their commercialization, as for example the breathability. This study aims at evaluating the repeatability and reproducibility of the breathability measurement in surgical masks, and at assessing the effects induced by the selected measurement points on the mask. Three samples for each type I, II and IIR masks were tested within the experimental protocol. Breathability was measured by following the UNI EN 14683:2019. Fifteen measurement points were identified for each mask and the measure was repeated five times per each point. Standard deviation across the repeated measures on the single point and across the different tested areas were used to evaluate the repeatability and reproducibility of the procedure, respectively. The measurement uncertainty was also computed. In addition, in order to verify the effects induced by the selection of five points out the available fifteen, as requested by the UNI EN 14683:2019, all combinations of points were tested. Results showed a high value of reproducibility error, leading to the consideration that different areas of the same mask are characterized by different values of breathability. Thus, the selection of the five measurement points to perform the breathability measurement according to the standard is a crucial aspect that can influence the result of mask compliance. The results of the present paper could provide useful information for the standardization of the breathability measurement.

Detection of Sleep Apnea from Single-Lead ECG: Comparison of Deep Learning Algorithms

Abstract: Apnea is a prevalent sleep disorder which has detrimental impacts on human health and quality of life. Accurate automatic algorithms for the detection of sleep apnea are needed for analyzing long-term sleep data and monitoring and management of its side effects and consequences. Among different approaches for automatic detection of sleep apnea from biosignals, deep learning algorithms are of particular interest as, unlike conventional machine learning algorithms, they do not rely on expert crafted features. In this paper, we developed and evaluated a number of different deep learning models for the detection of sleep apnea from a single-lead electrocardiogram (ECG) signal. ECG R-peak amplitude and R-R intervals were extracted, and power spectral analysis was performed to align the R-peak amplitude and the R-R intervals in frequency domain. Convolutional neural network (CNN), long short-term memory (LSTM), bidirectional LSTM, gated recurrent unit, and deep hybrid models were implemented and analyzed. The performance of deep learning algorithms was evaluated on an apnea-ECG dataset of 70 recordings divided into a learning set of 35 records and a test of 35 records. The best accuracy, sensitivity, specificity, and F1-score on the test data were 80.67%, 75.04%, 84.13%, and 74.72%, respectively, with a hybrid CNN and LSTM network. The results show promise toward improved apnea detection using deep learning.

Densely Connected Convolutional Networks (DenseNet) for Diagnosing Coronavirus Disease (COVID-19) from Chest X-ray Imaging

Abstract: Since the beginning of the coronavirus disease (COVID-19) pandemic several machine learning and deep learning methods had been introduced to detect the infected patients using the X-Ray or CT scan images. Numerous sophisticated data-driven methods had been introduced to improve the performance and the accuracy of the diagnosis models. This paper proposes an improved densely connected convolutional networks (DenseNet) method based on transfer learning (TL) to enhance the model performance. The results show promising model accuracy.

An Optimal Procedure for Stride Length Estimation Using Foot-Mounted Magneto-Inertial Measurement Units

Abstract: Stride length is often used to quantitatively evaluate human locomotion performance. Stride by stride estimation can be conveniently obtained from the signals recorded using miniaturized inertial sensors attached to the feet and appropriate algorithms for data fusion and integration. To reduce the detrimental drift effect, different algorithmic solutions can be implemented. However, the overall method accuracy is supposed to depend on the optimal selection of the parameters which are required to be set. This study aimed at evaluating the influence of the main parameters involved in well-established methods for stride length estimation. An optimization process was conducted to improve methods’ performance and preferable values for the considered parameters according to different walking speed ranges are suggested. A parametric solution is also proposed to target the methods on specific subjects’ gait characteristics. The stride length estimates were obtained from straight walking trials of five healthy volunteers and were compared with those obtained from a stereo-photogrammetric system. After parameters tuning, percentage errors for stride length were 1.9%, 2.5% and 2.6% for comfortable, slow, and fast walking conditions, respectively.

Gait Analysis using Wearable E-Textile Sock: an Experimental Study of Test-Retest Reliability

Abstract: Sock is a wearable e-textile sock for gait analysis. It is based on the acquisition and digital processing of the angular velocities of the lower limbs. In this paper we focus on the study of test-retest reliability of this system in measuring spatio-temporal gait parameters. The analysis was simultaneously conducted on data acquired by a multicamera system for gait analysis (SMART-DX 700 by BTS), in order to have reference values. A group of healthy subjects, equipped with both systems, performed four repeated walking tests along an 11 m walkway, consecutively and under constant conditions. The four tests were repeated at preferred, slow and fast self- selected walking speed. The Intraclass Correlation Coefficient (ICC) and Minimum Detectable Change (MDC) were evaluated to assess the repeatability of the measures. ICC values range from moderate to excellent for all gait parameters assessed by smart socks. The novel system presents test-retest reliability values comparable to, if not higher than, those shown by the gold standard. Finally, the results of gait reliability as a function of walking speed show excellent ICCs and very low MDCs for all parameters evaluated on trials at fast velocity, supporting the referenced hypothesis that faster movement is more consistent.

Effect of Deep Brain Stimulation Frequency on Gait Symmetry, Smoothness and Variability using IMU

Abstract: Deep brain stimulation (DBS) implant represents an appropriate treatment for motor symptoms typical of Parkinson’s Disease (PD). However, little attention has been given to the effects of different DBS stimulation frequencies on gait outcomes. Accordingly, the aim of this pilot study was to evaluate the effects of two different DBS stimulation frequencies (60 and 130 Hz) on gait spatio-temporal parameters, symmetry, smoothness, and variability in PD patients. The analysis concentrated on acceleration signals acquired by a magnetic inertial measurement unit placed on the trunk of participants. Sessions of gait were registered for three PD patients, three young and three elderly healthy subjects. Gait outcomes revealed a connection with both age and pathology. Values of the Harmonic Ratio (HR) estimated for the three-axis acceleration signals showed subjective effects provoked by DBS stimulation frequencies. Consequently, HR turned out to be suitable for depicting gait characteristics, but also as a monitoring parameter for the subjective adaptation of DBS stimulation frequency. Concerning the Poincare analysis of vertical acceleration signal, PD patients showed a greater dispersion of data compared to healthy subjects, but with negligible differences between the two stimulation frequencies. Overall, the presented analysis represented a starting point for the objective evaluation of gait performance and characteristics in PD patients with a DBS implant.

Analysis of Test-Retest Repeatability of Gait Analysis Parameters in Hereditary Spastic Paraplegia

Abstract: Hereditary Spastic Paraplegia (HSP) is a rare inherited neurological disorder, whose predominant feature is a spastic gait. Gait analysis represents an objective tool to quantify the impairment of gait pattern in patients with HSP, thus supporting diagnosis and monitoring of disease progression. This study contributes to the characterization of HSP pathological gait, providing the assessment of test-retest repeatability of 122 parameters regarding postural sway, anticipatory postural adjustment in step initiation, gait and turn tasks. Data are collected on a cohort of thirty-five HSP patients, performing three consecutive repetitions of the Instrumented Stand and Walk (iSAW) test provided by Mobility Lab gait analysis system by APDM. Intraclass Correlation Coefficient (ICC) is used to assess repeatability. Repeatability Limit (RL) has also been evaluated and compared to the absolute value of difference (DoM) of HSP patients’ measurements mean and normative mean of the same variable, in order to understand which variable can better characterize HSP gait with respect to normal gait. Results show that gait and turn measurements are more repeatable than sway and anticipatory postural adjustments variables. Furthermore, this study confirms previous findings in this field, identifying, among other gait parameters, cadence, gait velocity, stride length and RoM of the shanks as the main distinctive parameters of the pathology. Conversely, the RoM of the knees presents excellent repeatability, but low difference between healthy and pathological subjects.

Innovative Smart Face Mask to Protect Workers from COVID-19 Infection

Abstract: The most frequent prodromes of COVID-19 infection are fever and signs/symptoms of incipient respiratory diseases such as cough and shortness of breath or tachypnea. However, it is not infrequent that in patients infected with COVID-19, in addition to respiratory manifestations, cardiac rhythm alterations are also present which can be an early sign of an acute cardiovascular syndrome. It is therefore of utmost importance, especially for health care and civil protection workers who are most exposed to the infection, to detect the prodromal symptoms of this infection in order to be able to make a diagnosis of possible positivity to COVID-19 infection as quickly as possible and therefore to provide their immediate insertion in the isolation/therapy protocols. Here a prototype of a smart face mask is presented: the AG47-SmartMask. In addition to having the function of both an active and passive anti COVID-19 filter, the latter by an electro-heated filter brought to a minimum temperature of 38°C, the AG47-SmartMask also allows the continuous monitoring of numerous cardio-pulmonary variables. Several specific sensors are incorporated into the mask in an original way that assess the inside mask temperature, relative humidity and air pressure together with the auricular assessment of body temperature, heart rate and percentage of oxygen saturation of haemoglobin. Sensors work in synergy with an advanced telemedicine platform. To validate the device, twenty workers engaged in a vegetable packaging chain tested the tool simulating, while working, both tachypnea and cough, and the AG47-SmartMask faithfully quantified the simulated dyspnoic events.

A Wearable SSVEP BCI for AR-based, Real-time Monitoring Applications

Abstract: A real-time monitoring system based on Augmented Reality (AR) and highly wearable Brain-Computer Interface (BCI) for hands-free visualization of patient’s health in Operating Room (OR) is proposed. The system is designed to allow the anesthetist to monitor hands-free and in real-time the patient’s vital signs collected from the electromedical equipment available in OR. After the analysis of the requirements in a typical Health 4.0 scenario, the conceptual design, implementation and experimental validation of the proposed system are described in detail. The effectiveness of the proposed AR-BCI-based real-time monitoring system was demonstrated through an experimental activity was carried out at the University Hospital Federico II (Naples, Italy), using operating room equipment.

Simultaneous Measurement of Heartbeat Intervals and Respiratory Signal using a Smartphone

Abstract: This paper explores the use of a common smartphone for measuring simultaneously both heartbeat intervals and respiratory cycles. The proposed technique uses the smartphone’s accelerometer to measure the seismocardiographic signal and the acceleration due to breathing movements. The measurement is carried out while the subject is laying down, with the smartphone placed on his/her xiphoid process. In the paper, processing algorithms are presented, that can be used to obtain the heartbeat and the respiratory intervals from the measured signals. As concrete examples of possible application, heartbeat intervals are used to derive Heart Rate Variability and, together with the respiratory signal, to derive a Respiratory Sinus Arrhythmia measure on eight healthy volunteers.

A preliminary study on the dynamic characterization of a MEMS microgripper for biomedical applications

Abstract: This preliminary study concerns the dynamic characterization of a MEMS microgripper for biomedical applications. In particular, a prototype of microgripper, embedded with electrostatic comb-drive actuators, has been powered with a 10V sinusoidal input at different frequencies, 0.5 Hz, 1.0 Hz and 4.0 Hz. The response of the device has been recorded with a trinocular optical microscope, equipped with a digital camera and the recorded videos have been analysed with an in-house software implemented by the authors for the measurement of the comb-drive angular displacement, velocity and acceleration. The uncertainty analysis has been carried out to identify the uncertainty sources that characterize the measurements. Experimental data showed that the maximum angular displacement is (13.2 ± 0.2)•10-3 rad, (13.6 ± 0.2)•10-3 rad and (13.1 ± 0.3)•10-3 rad, the maximum angular velocity is (2.8 ± 0.2)•10-2 rad/s, (5.7 ± 0.4)•10-2 rad/s and (19.9 ± 1.5)•10-2 rad/s, and the angular acceleration is 0.178 ± 0.015 rad/s2, 0.72 ± 0.04 rad/s2 and 6.3 ± 0.7 rad/s2 for 0.5 Hz, 1.0 Hz and 4.0 Hz, respectively. The measurement results have been compared with the expected values from the theoretical model that describes the behaviour of the microgripper: the overall percentage error (PE) between the measured and the expected values at different frequencies is lower than 1%, 1% and 3% for the angular displacement, velocity and acceleration respectively.

PARKIBIP Feedback Wearable Rehabilitation Device: Market Analysis and Enhancements

Abstract: Rehabilitation counteracts motor deficiencies in gait disorder of Parkinson's Disease (PD) patients. PARKIBIP is a wearable feedback device that aims to offer a continuous and personalized rehabilitation tool for such people. A survey and external study of PARKIBIP suggest design enhancements. Exploration of its industrial potential shows direct competitors, a first step to conclude that PARKIBIP is suitable for Technological Transfer to a company for commercial dissemination. PARKIBIP is both a home treatment helping device and a clinical data & feedback capture terminal for the electronic medical record. Being wearable technology, PARKIBIP stands out in the present global context as an affordable robotic element with feedback capability connected to the patient's mobile phone.

Preliminary Study of a Flexible Printed Multi-Sensing Platform for Electromyography and Lactate Measuring during Rehabilitation

Abstract: Real-time measure of muscular fatigue during physical exercise is highly demanded in fields as rehabilitation and physiotherapy, both hospital and home-based. In addition to the well-accepted features extracted from EMG measurements, an increasing interest has been recently addressed to the investigation of biochemical fatigue markers to drive additional information on fatigue evolution. We propose here a multi-sensing non-invasive approach to measure fatigue through EMG and lactate sensing. A printed unobtrusive sensing patch was developed as first prototype by means of the emerging Aerosol Jet Printing technique, ensuring high repeatability and stability even on a flexible substrate. To study the behavior of this patch, preliminary measurements were acquired to perform an analysis of both the printed sensors. EMG electrodes, with skin-electrodes impedance magnitude and phase angle with trends comparable to commercial electrodes, showed the possibility to successfully extract mean and median frequencies from EMG and to detect their decrease after intense exercise. Preliminary results on lactate static measure showed a limit of detection of 3.1±0.3 mM with the highest linearity (R=0.9995) and sensitivity (0.39 µA/mM) in the range 0-20 mM. Furthermore, dynamic tests permitted a preliminary analysis on the ability of the sensor to measure changes in the concentration of lactate continuously. Overall, reported measures represent a promising starting point to develop a patch easily integrable in any wearable for noninvasive and totally customized fatigue measure during physical exercise.

Metrological characterization and signal processing of a wearable sensor for the measurement of heart rate variability

Abstract: This paper presents a methodology for the processing of the Photoplethysmography (PPG) signal measured using a smartwatch during motion tests. For statistical validation, signals from 15 healthy subjects have been collected while the subjects are walking on a treadmill. The motion artifacts (MAs) of the PPG signal have been removed demonstrating that the 37% of the signals are affected by MAs. Then, the experimental performance assessment of the PPG signal, from which the heart rate variability (HRV) has been extracted, by measuring the RR intervals, is compared to the RR intervals extracted from ECG signals measured using a multi-parametric chest belt that is considered as a reference sensor. The uncertainty of the PPG sensor in the measurement of the RR intervals is ± 169 ms, (with a coverage factor k = 2) if compared to the reference method, which in percentage is 30%.

Development of a multi-transduction system for breath analysis in neurodegenerative diseases

Abstract: The purpose of this paper is to report on the development of a multi-transduction gas-sensor interface that is based on capacitive, resistive, and bulk acoustic wave (BAW) devices. The system is designed to be employed as a sensing hardware interface in the breath analysis for biomedical applications. The different transduction typologies are handled by means of electronic boards developed ad hoc. The prototype is compact and provides all the features needed for gas sensors characterization, from the electronic interfaces to the gas testing chamber, passing through the data acquisition system by a single board computer equipped by an Arduino DUE-compatible board. The developed system is very flexible and can be easily expanded to enhance its features through hardware or software upgrades. Here, we report the results of the preliminary activities regarding the measurement system development and its laboratory validation.

Single-Pulse Measurement Electronics for Accurate Dosimetry in X-ray Radiation Therapy

Abstract: This work aims at describing a dose-measurement system based on a high-quality single-crystal synthetic diamond coupled to specifically-designed front-end electronics and timing circuitry. By exploiting the pulsed nature of radiation sourced by LINACs, the electronics performs a synchronous detection of charge-carriers generated by the diamond detector on which pulsed X-rays impinge. Synchronous detection would assure superior level of accuracy, sensitivity, and dynamics in comparison to conventional continuous-integration solutions, since the measurements are affected by environmental noise to a lesser extent. In addition, preliminary characterization described in this paper highlights the feasibility of a measurement system capable of providing information on the dose delivered at the level of a single X-ray pulse. In this context, the proposed system would be able to respond to the specific needs of accurate dose diagnostics for modern radiotherapy treatments where radiation-dose is characterized by a complex shaped distribution and steep gradients.

Pressure sensors used as bioimpedance plantar electrodes: a feasibility study

Abstract: Diabetic Foot Ulcers are ominous consequence of Diabetic Foot. Only general preventive guidelines are available, and ulcers happen with no previous notice. To develop a multidimensional ulcer opening warning device: temperature, pressure, humidity and friction are usually considered. We use standard flexible Force Sensing Resistors FSR 402 to detect not only plantar pressure, but also plantar bioimpedance. Since FSR includes conductive electrodes covered by polymer films, the interface with the subject can be considered a capacitive electrode. A special bioimpedance detection circuit is required to inject current using two homologous FSR 402 contacts and measuring the resulting voltage from the other two contacts available. This circuit is able to detect the cardiac activity from the foot sole. For the first time, pressure sensors are used as bioimpedance electrodes.

Smartwatches selection: market analysis and metrological characterization on the measurement of number of steps

Abstract: This paper is focused on identifying the accuracy of smartwatches (SWs) on the measurement of number of steps. Five SWs have been identified based on technical characteristics and costs from a list of 32 SWs available on the market. A metrological characterization on the selected SWs has been made on six subjects wearing all the SWs and doing walking activity with natural, slow and fast pace. R, R2 and statistical confidence, with coverage factor equal to 2, are computed considering videos as reference system to identify the number of steps. The overall statistical confidence is 4.2% for natural pace, 7.5% for the slow pace and 7.1% for the fast pace.

Kinect-based wearable prototype system for ataxic patients neurorehabilitation: software update for exergaming and rehabilitation

Abstract: The aim of this study is to evaluate the differences introduced by an update of implemented code between the two version of prototype for ataxic patients’ neurorehabilitation device. The rehabilitation consists in virtual exercise for the patient to improve his control of the upper arm during daily movement. The prototype is based on the Microsoft Kinect device to acquire subject’s position and on the Arduino board with accelerometer/gyroscope sensor to acquire kinematics quantities of the wrist during the task. Two subjects were analysed with 2.0 version of the prototype and they were compared to 2 of 20 subjects selected from the first control group to highlight the differences between the two versions.

Cardiorespiratory monitoring using a mechanical and an optical system

Abstract: A continuous monitoring of cardiorespiratory activity can play an essential role in the health prevention since the cardiovascular and ventilatory systems regulate several vital functions of the human body and adapt themselves in response to various stressors. Typically, early detection of cardiorespiratory irregularities is performed by monitoring respiratory and heart rate (RR and HR) at rest. Among several technological solutions, the most promising are based on mechanical and optical systems such as gyroscopes (GYRs) and accelerometers in inertial measurement units, and fiber Bragg gratings (FBGs) embedded into wearable and non-wearable items.In this work, we investigated the capability of a mechanical system (i.e., a GYR) and an optical system (i.e., a flexible sensor based on FBG) to perform the simultaneous RR and HR monitoring. The system placement varied according to the sensor type to ensure the best unobtrusive cardiorespiratory monitoring: the GYR was worn on the chest, and the FBG-based flexible sensor was placed on a chair in contact with the chest back. Results showed similar performances between the mechanical and optical systems when compared to a reference instrument (mean absolute percentage error -MAPE < 7.7% and 6.1% for HR and MAPE ≤ 0.23% and 1.7% for RR for the FBG and the GYR, respectively).

An Efficient Near-lossless Compression Algorithm for Multichannel EEG signals

Abstract: In many biomedical measurement procedures, it is important to record a huge amount of data, to monitor the state of health of a subject. In such a context, electroencephalograph (EEG) data are one of the most demanding in terms of size and signal behavior. In this paper, we propose a near-lossless compression algorithm for EEG signals able to achieve a compression ratio in the order of 10 with a root-mean-square distortion less than 0.01%. The proposed algorithm exploits the fact that Principal Component Analysis is usually performed on EEG signals for denoising and removing unwanted artifacts. In this particular context, we can consider this algorithm as a good tool to ensure the best information of the signal beside an efficient compression ratio, reducing the amount of memory necessary to record data.

Comparison of OpenPose and HyperPose artificial intelligence models for analysis of hand-held smartphone videos

Abstract: Movement assessments are invaluable in clinical practice. However, the feasibility of in-person evaluation has been greatly affected due to the COVID-19 pandemic. To overcome this barrier, a virtual assessment system using artificial intelligence (AI) and patient provided videos is needed. AI models for pose inference have produced viable results for identifying a person’s joint centers. Identifying AI models for pose inference that provide clinically meaningful results is important for designing a virtual motion assessment tool. This study aims to evaluate the clinical usefulness of two popular pose inference models, OpenPose and HyperPose. Videos recorded by two physicians, who independently performed movements they deemed clinically relevant. Keypoint skeletons were generated and manually inspected frame-by-frame to determine which model produced higher-quality pose inferences. OpenPose produced significantly better scores than HyperPose when comparing within videos (p<0.001). Right ankle and right wrist had the poorest performances. Best-practices to be used in the future design of a virtual motion assessment tool are required to improve video "AI-friendliness".

Efficient feature selection for electroencephalogram-based authentication

Abstract: Opposed to classic authentication protocols based on credentials, biometric-based authentication has recently emerged as a promising paradigm for achieving fast and secure authentication of users. Among the several families of biometric features, electroencephalogram (EEG)-based biometrics is considered as a promising approach due to its unique characteristics. Classification systems based on machine learning allow processing of large amounts of data and performing accurate attribution of each signal to the most relevant group, thus representing an invaluable tool for EEG-based biometrics. This paper provides an experimental evaluation of the performance achievable by EEG-based biometrics employing machine learning. We consider several groups of EEG signals and propose a suitable feature extraction criterion. Then, the extracted features are used along with neural network-based classification algorithms, K Nearest Neighbours (KNN), and eXtreme Gradient Boost (XGBoost) for attributing any EEG signal to a subject. A full feature set and a reduced feature sets are considered and tested on three public data sets. The feature selection criteria are based on a correlation map among features, ANOVA F-test, and logistic regression weights. The results show that the reduced feature sets achieves a significant reduction in computation time over the full feature set, while also providing some improvement in performance.

Multi-Segments Kinematic Model of the Human Spine during Gait

Abstract: The complex biomechanical structure of the human spine requires a deep investigation to properly describe its physiological function and its kinematic contribution during motion. The computational approach allows the segmentation of the human spine into several rigid bodies connected by 3D joints. Despite the numerous solutions proposed by previous literature studies based on both inertial and stereophotogrammetric systems, the modelling of the human spine is characterized by some limitations such as the lack of standardization. Accordingly, the present preliminary study focused on the development of a multi-segments kinematic model of the human spine and its validation during gait trials. Three-dimensional spinal angular patterns and ranges of motion of one healthy young subject were considered as outcomes of interest. They were obtained by applying the YXZ Euler angles convention to the custom model. First, results were compared with those of the standard Plug-in-Gait full-body model, which segments the human spine into pelvis and trunk segments. Then, outcomes of the multi-segments model were compared with those obtained using the Tilt-Twist method. Overall, results stressed the importance of the spine segmentation, the major angular contributions of spinal regions during gait (Medium-Lumbar segments for lateral bending and flexion-extension, Thoracic-Medium segments for axial rotation), and the reliability of the proposed custom model (differences between Euler angles method and Tilt-Twist method lower than 0.5° in most cases). Future analysis on a larger healthy population and in the clinical context might be implemented to optimize, standardize and validate the proposed human spine model.

mHealth application for remote health monitoring useful during the COVID 19 pandemic

Abstract: The paper presents an application of mHealth – a mobile app for remote health monitoring, that facilitates using a Bluetooth enabled health measuring device and synchronizing health data to a health care services provider’s web portal. The mobile app uses a public API that allows its integration in a complex platform for home care providers, allowing health monitoring of large groups of patients, monitoring vital functions, including body temperature, respiratory rate and arterial blood oxygen saturation, relevant in monitoring COVID-19 patients.

Ensemble Empirical Mode Decomposition for Efficient R-Peak Detection in Electrocardiograms Acquired by Portable Sensors During Sport Activity

Abstract: Wearable and portable electrocardiographic devices are revolutionizing athlete’s screening through digital health application enabling a continuous monitoring of important cardiac parameters in real-time. Automatic examination of electrocardiogram (ECG) acquired during sport activity is challenging because acquisition conditions often lead to record ECGs with low signal to noise ratio (SNR). The initial issue of automatic ECG analysis is the identification of R peaks. R peaks are fundamental for the estimation of heart rate, which is the primary clinical parameter used by athletes for athletic performance evaluation. Thus, the aim of this research is to propose an R-peak detection algorithm for ECGs acquired during sport activity by portable and wearable sensors dealing with low SNR. The algorithm is based on a noise assisted data analysis method: Ensemble Empirical Mode Decomposition method (EEMD). Localization of R peaks is primarily performed on the first intrinsic mode function extracted by the EEMD. The algorithm was tested on ‘Run on indoor treadmill’ dataset from Physionet. ECGs were acquired during running/light jogging on an indoor treadmill and present a low SNR (1±7 dB). The developed EEMD-based algorithm showed good performances in terms of positive predicted value (91.08%), sensitivity (92.76%), false discovery rate (8.92), false negative rate (7.24%), cumulative statistical index (83.84%) and mean R-peak position error 1.10 [0.46;1.46]ms. EEMD-based algorithm performs efficiently also in computing heart rate. In conclusion, the developed R-peak detection EEMD-based algorithm showed good level of performances even working on low-SNR ECG acquired during sport activity by portable sensors.

Gait events detection from heel and toe trajectories: comparison of methods using multiple datasets

Abstract: Reliable detection of gait events is important to ensure accurate assessment of gait. While it is usually performed resorting to force platforms, methods based uniquely on kinematic analysis have also been proposed. These methods place no restrictions on the number of steps that can be analysed, simplifying setup and complexity of assessments. They also replace the need of annotating events manually when force platforms are not available. Although few methods have been proposed in literature, validation studies are relatively scarce. In this study we present multiple methods for the detection of heel strike (HS) and toe off (TO) in normal walking, and validate the detection against annotated events using three different datasets. The best performing candidates are based on the evaluation of heel vertical velocity (for HS) and toe vertical acceleration (for TO), resulting in relative errors of -12.4 ± 32.9 ms for HS and of -15.5 ± 24.9 ms for TO. The method is compatible with barefoot and shod walking, constituting a convenient, fast and reliable alternative to automatic gait event detection using kinematic data.