BIOSIGNALS 2020 Abstracts


Full Papers
Paper Nr: 4
Title:

Exploring the Merit of Collaboration in Classification and Compression of Epilepsy EEG Signal

Authors:

Rushda B. Ahmad and Nadeem A. Khan

Abstract: Ambulatory electroencephalogram (EEG), allows collection of patients data over extended periods of time. However, as a small recording requires large memory for storage, and this makes EEG data storage an arduous task. Moreover, classification of EEG for extraction of relevant information is relatively challenging, and selective data retrieval depends on task at hand. Consequently, EEG data storage and classification need to be computationally efficient. This paper presents a combined scheme, for the simultaneous compression and classification of EEG data, which not only decreases the overall computational effort, but also allows selective archiving and retrieval of data. Huffman and Arithmetic coding techniques are employed on CHB-MIT scalp EEG database and the results are presented in form of compression ratio (CR) and percentage root mean square distortion (PDR). For classification, Intelligent Neurologist Support System (INSS), has been used. The classifier output apart from being stored as data, is also used for intelligent data reduction, when only specific information is required, resulting in increased CR and decreased PDR, which is desired. Hence, the results show intelligent compression and reduction of data results in efficient management of EEG data. The signal undergoes state-of-the-art compression such that on reconstruction it almost maintains the same classification accuracy as the original one.

Paper Nr: 6
Title:

Eye-pointer Coordination in a Decision-making Task Under Uncertainty

Authors:

Catia Cepeda, Maria C. Dias, Dina Rindlisbacher, Marcus Cheetham and Hugo Gamboa

Abstract: Eye-tracking (ET) systems, which capture eye movements, are often used to measure human behavior while interacting with a user interface. Given the high costs and challenges of acquiring, installing and ensuring good calibration of ET systems, the use of pointer (or mouse) tracking is gaining interest as a viable alternative in research on human-computer interaction. In this study, we measured and evaluated temporal and spatial relationships between eye and pointer movements in a standardized task that allowed us to examine the relationship between eye and pointer movements while participants made decisions under conditions of high and low uncertainty. We collected data from N=81 participants and applied a range of metrics to a total of 5205 decision trials. The overall findings show that the convergence between eye and pointer movements is consistently high. Importantly, there are differences in levels of convergence depending on the temporal, spatial and combined temporo-spatial metrics used. There are also differences in eye-pointer convergence depending on the relative level of decision uncertainty in the task. In conclusion, the present findings favour the use of pointer tracking to analyse human-computer interaction in more complex tasks.

Paper Nr: 8
Title:

Source-based Multifractal Detrended Fluctuation Analysis for Discrimination of ADHD Children in a Time Reproduction Paradigm

Authors:

Shiva Khoshnoud, Mohammad A. Nazari and Mousa Shamsi

Abstract: Electroencephalography recordings have a scale-invariant structure and multifractal detrended fluctuation analysis (MF-DFA) could quantify the fluctuation dynamics of these recordings in different brain states. However, the channel-based electrical activity of the brain has low spatial resolution and considering the source-level activity patterns is a good answer for this restriction. In this work, the multifractal spectrum parameters of the channel-based EEG, as well as the corresponding source-based independent components in children with Attention Deficit Hyperactivity Disorder (ADHD) and the age-matched control group, has been investigated. Considering the perceptual timing deficit in children with ADHD, for the analysis of the multifractality, two brain states including the eyes-open rest and the time reproduction condition have been considered. The results obtained showed that switching from rest to the time reproduction condition increases the degree of multifractality and so the complexity and non-uniformity of the signal. While the channel-based multifractal properties could not significantly distinguish two groups, the results for the source-based multifractal analysis showed a significantly decreased degree of multifractality for children with ADHD in prefrontal, mid-frontal and right frontal source clusters suggesting reduced activation of these clusters in this group. Utilizing support vector machine (SVM) classifier it is found that, the source-based multifractal features provide a significantly higher accuracy rate of 86.67% in comparison to the channel-based measures.

Paper Nr: 11
Title:

The Effect of Maxblur-pooling in Neural Networks on Shift-invariance Issue in Various Biological Signal Classification Tasks

Authors:

Xianyin Hu, Shangyin Zou, Yuki Ban and Shin’ichi Warisawa

Abstract: Modern neural networks are widely employed in bio-signal processing due to their effectiveness. However, recent research showed that neural networks for image recognition is not shift-invariant as it was assumed, while it is an important property in bio-signal processing. Fortunately, a simple methodology was proposed referred to as Maxblur-pooling to improve the shift-invariance of neural networks for image recognition. However, the corresponding issue in the domain of bio-signal processing remains untouched. To verify the shift-invariance of neural networks when applied to bio-signal processing, we performed two experiments across different tasks and types of bio-signals. One is Atrial Fibrillation (AF) detection from R-R interval and the other is emotion recognition from multi-channel EEG. We were able to show that the lack of shift-invariance also happens in temporal bio-signal classification. In the AF detection task, we succeed to validate the effectiveness of Maxblur-pooling, which demonstrating improvements in both accuracy (2%-13%) and consistency (8%-15%) compared to the baseline. While for the emotion recognition task, we did not observe any improvements using Maxblur-pooling. Our research provided empirical knowledge for developing real-time diagnose systems that is stable to temporal shifts.

Paper Nr: 14
Title:

Recurrent Neural Network for Gait Pathology Detection

Authors:

Jorge Sanchez-Casanova, Judith Liu-Jimenez, Pablo Fernandez-Lopez and Raul Sanchez-Reillo

Abstract: This work presents a pathology detection system on the lower train. For this, a database of healthy subjects has been captured. Due to the nonexistence of pathological gait databases, pathology walks have been simulated. The users used sole padding in order to simulate clubfoot walk. The database consists of acceleration, angular acceleration, magnetic field signals and the angles between the joints. The algorithm extracts fragments of the signals which are used to train a recurrent neural network (RNN). To optimize the results, hand-tuning method was used to modify the hyperparameters. Using the best configuration, we have a 97% accuracy training with 90% of the database. Although, if we train with only 50% of the data the accuracy reaches at 91%. The results obtained show the solution feasibility, although further research should be done using real lower train pathologies.

Paper Nr: 18
Title:

An Autoregressive Multiple Model Probabilistic Framework for the Detection of SSVEPs in Brain-Computer Interfaces

Authors:

Rosanne Zerafa, Tracey Camilleri, Owen Falzon and Kenneth P. Camilleri

Abstract: This work investigates a novel autoregressive multiple model (AR-MM) probabilistic framework for the detection of steady-state visual evoked potentials (SSVEPs) in brain-computer interfaces (BCIs). The proposed method is compared to standard SSVEP detection techniques using a 12-class SSVEP dataset recorded from 10 subjects. The results, obtained from a single-channel analysis, reveal that the AR-MM probabilistic framework significantly improves the SSVEP detection performance compared to the standard single-channel power spectral density analysis (PSDA) method. Specifically, an average classification accuracy of 82.02 ± 16.21 % and an information transfer rate (ITR) of 48.22 ± 17.25 bpm are obtained with a 2 s period for SSVEP detection with the AR-MM probabilistic framework. These results are found to be on average only 2.29 % and 3.73 % lower in classification accuracy compared to the state-of-the-art multichannel SSVEP detection methods, specifically the canonical correlation analysis (CCA) and the filter bank canonical correlation analysis (FBCCA) methods, respectively. In terms of training, it is shown that the proposed approach requires only a few seconds of data to train each model. This study revealed the potential of using the AR-MM probabilistic approach to distinguish between different classes using single-channel SSVEP data. The proposed method is particularly appealing for practical use in real-world BCI applications where a minimal amount of channels and training data are desirable.

Paper Nr: 24
Title:

Explaining the Ergonomic Assessment of Human Movement in Industrial Contexts

Authors:

Sara Santos, Duarte Folgado, João Rodrigues, Nafiseh Mollaei, Carlos Fujão and Hugo Gamboa

Abstract: The repetitive nature of manufacturing processes is identified as a risk factor for the onset of musculoskeletal disorders. For prevention, the operator’s exposure risk is measured through ergonomic risk scores which are often associated with a workstation, ignoring the variability among operators. Moreover, the score values hinder a comprehensive interpretation by occupational physicians. Observation methods require significant effort, preventing accurate and continuous evaluation. The conducted study developed a solution using inertial sensors for automatic operator risk exposure in the manufacturing industry. Two experimental assessments were conducted: laboratory validation, performed by 14 subjects, using an optical motion capture system as a reference; and field evaluation, with 6 participants, acquired on a real automotive assembly line, served as the basis for an ergonomic risk evaluation study. Through the research, it was implemented an upper-body motion tracking algorithm relying on inertial information, to estimate the angular orientation of anatomical joints. An adjusted ergonomic risk score, based on direct measurements was developed allowing an ergonomic evaluation which also has an explanation approach, based on the comprehensive analysis of the angular risk factors. Direct measurements fasten the ergonomic feedback, consequently, the evaluation can be extended to more operators, ultimately preventing work-related injuries.

Paper Nr: 29
Title:

Delayed Mutual Information to Develop Functional Analysis on Epileptic Signals

Authors:

Victor B. Tsukahara, Pedro B. Jeronymo, Jasiara Carla de Oliveira, Vinicius R. Cota and Carlos D. Maciel

Abstract: Epilepsy is the second most prevalent brain disorder affecting approximately 70 million people worldwide. A modern approach to develop the brain study is to model it as a system of systems, represented by a network of oscillators, in which the emergent property of synchronisation occurs. Based on this perspective, epileptic seizures can be understood as a process of hyper-synchronisation between brain areas. To investigate such process, a case study was conducted applying Delayed Mutual Information (DMI) to perform functional connectivity analysis, investigating the channel capacity (C) and transmission rate (R) between brain areas — cortex, hippocampus and thalamus — during basal and infusion intervals, before the beginning of generalised tonic-clonic behaviour (TCG). The main contribution of this paper is the study of channel capacity and transmission rate between brain areas. A case study performed using 5 LFP signals from rodents showed that the applied methodology represents an another appropriate alternative to existing methods for functional analysis such as Granger Causality, Partial Directed Coherence, Transfer Entropy, providing insights on epileptic brain communication.

Paper Nr: 33
Title:

Image Evaluation in Magnetic Resonance Cholangiopancreatography

Authors:

K. P. Pinho, P. M. Gewehr, A. C. Pinho, A. M. Gusso and C. A. Goedert

Abstract: The objective of this study is to evaluate the image quality with two different contrast agents in MRCP compared to medical evaluation and by using the software Image J®. Natural juices and pulps of different types (açai liquid and powder; and blend) were selected. The selection of patients (31 women and 33 men) was performed at Clinical Hospital, which provides general care in Curitiba city (Brazil). The application of the MRCP protocol followed a sequence tested in healthy volunteers and for the samples described. For image analysis, 2 radiologists participated and were identified as evaluator 1 (E1) and 2 (E2), in order to identify the effect of the contrasts on the images. For the 6 samples tested, only 2 samples remained dark on T2 weighting, which prevents their use as contrast agent. The evaluation of the images was performed separately for each evaluator on different days and places, to identify an appropriate action for the contrasts (A and B). The use of the software (Image J®) allowed a less subjective analysis of the image quality when compared to the evaluation of radiologists and, for the examples presented, a quantitative assessment since the chosen images were submitted to the software analysis.

Paper Nr: 41
Title:

Saliency Maps of Video-colonoscopy Images for the Analysis of Their Content and the Prevention of Colorectal Cancer Risks

Authors:

Valentine Wargnier-Dauchelle, Camille Simon-Chane and Aymeric Histace

Abstract: The detection and removal of adenomatous polyps via colonoscopy is the gold standard for the prevention of colon cancer. Indeed, polyps are at the origins of colorectal cancer which is one of the deadliest diseases in the world. This article aims to contribute to the wide range of methods already developed for the prevention of colorectal cancer risks. For this, the work is organized around the detection and the localization of polyps in video-colonoscopy images. The aim of this paper is to find the best description of a bowel image in order to classify a patch, that is to say a image fragment, as polyp or not. The classification is achieved thanks to an SVM (Support Vector Machine) using a bag of features. Different types of features extraction will be compared. Thus, the traditional SURF (Speeded-Up Robust Features) extractor will be compared to local features extractors like HOG (Histogram of Oriented Gradient) and LBP (Local Binary Pattern) but also to an original extractor based on the structural entropy.

Paper Nr: 56
Title:

Prediction of the Impact of Physical Exercise on Knee Osteoarthritis Patients using Kinematic Signal Analysis and Decision Trees

Authors:

M. Mezghani, N. Hagemeister, M. Kouki, Y. Ouakrim, A. Fuentes and N. Mezghani

Abstract: The evaluation of knee biomechanics provides valuable clinical information. This can be done by means of a knee kinesiography exam which measures the three-dimensional rotation angles during walking, thus providing objective knowledge about knee function (3D kinematics). 3D kinematic data is quantifiable information that provides opportunities to develop automatic and objective methods for personalized computer-aided treatment systems. The purpose of this study is to explore a decision tree based method for predicting the impact of physical exercise on a knee osteoarthritis population. The prediction is based on 3D kinematic data i.e., flexion/extension, abduction/adduction and internal/external rotation of the knee. Experiments were conducted on a dataset of 309 patients who have engaged in physical exercise for 6 months and have been grouped into two classes, Improved state (I) and not-Improved state (nI) based on their state before (t0) and after the exercise (t6). The method developed was able to predict I and nI patien with knee osteoarthritis using 3D kinematic data with an accuracy of 82%. Results show the effectiveness of 3D kinematic signal analysis and the decision tree technique for predicting the impact of physical exercise based on patient knee osteoarthritis pain level.

Paper Nr: 57
Title:

On the Metrics and Adaptation Methods for Domain Divergences of sEMG-based Gesture Recognition

Authors:

István Ketykó and Ferenc Kovács

Abstract: We propose a new metric to measure domain divergence and a new domain adaptation method for time-series classification. The metric belongs to the class of probability distributions-based metrics, is transductive, and does not assume the presence of source data samples. The 2-stage method utilizes an improved autoregressive, RNN-based architecture with deep/non-linear transformation. We assess our metric and the performance of our model in the context of sEMG/EMG-based gesture recognition under inter-session and inter-subject domain shifts.

Short Papers
Paper Nr: 1
Title:

Feature Space Reduction for Multimodal Human Activity Recognition

Authors:

Yale Hartmann, Hui Liu and Tanja Schultz

Abstract: This work describes the implementation, optimization, and evaluation of a Human Activity Recognition (HAR) system using 21-channel biosignals. These biosignals capture multiple modalities, such as motion and muscle activity based on two 3D-inertial sensors, one 2D-goniometer, and four electromyographic sensors. We start with an early fusion, HMM-based recognition system which discriminates 18 human activities at 91% recognition accuracy. We then optimize preprocessing with a feature space reduction and feature vector stacking. For this purpose, a Linear Discriminant Analysis (LDA) was performed based on HMM state alignments. Our experimental results show that LDA feature space reduction improves recognition accuracy by four percentage points while stacking feature vectors currently does not show any positive effects. To the best of our knowledge, this is the first work on feature space reduction in a HAR system using various biosensors integrated into a knee bandage recognizing a diverse set of activities.

Paper Nr: 7
Title:

Impact of Mental Fatigue during Repetitive Exercises of a Visual P300 Speller

Authors:

Patrick Schemrbi, Mariusz Pelc and Jixin Ma

Abstract: In this paper, we investigate the effect that mental fatigue during repetitive exercises of a visual P300 Speller has on the P300 component, in terms of accuracy, amplitude, latency, signal morphology, and overall signal quality. This work is part of a larger EEG based project and is based on the P300 speller BCI (oddball) paradigm and the xDAWN algorithm, with eight healthy subjects; while using a non-invasive Brain-Computer Interface (BCI) based on low fidelity electroencephalographic (EEG) equipment. Herein, eight channels through the initial task (6 minutes), additional tasks (50 minutes) and final task (6 minutes) states, recorded the subjects’ signal. Our results show that the accuracy was best for the initial task (IT) at 100%, followed closely by the final task (FT) at 98%. In addition, our ANOVA analysis showed that the amplitude exhibited a statistical significance between IT and FT, while the latency did not indicate any statistical difference. This paper provides initial results into the practicability of the aforementioned P300 speller methodology and low-cost equipment to be used repetitively and continually and the effect thereof on accuracy and signal characteristics. Our aim is to assess the effect of prolonged usage and exposure to the aforementioned methodology and equipment, with the aim of broadening its use in a real-world context.

Paper Nr: 10
Title:

A Public Dataset of Overground and Treadmill Walking in Healthy Individuals Captured by Wearable IMU and sEMG Sensors

Authors:

Harald Loose, Laura Tetzlaff and Jon L. Bolmgren

Abstract: The paper presents our public Gait Analysis Data Base (http://gaitanalysis.th-brandenburg.de), which contains 3D walking kinematics and muscle activity data from healthy adults walking at normal, slow or fast pace on the flat ground or at an incremental speeds on treadmill. The acceleration, angular velocity and magnetic rate vectors are measured using XSens MTw sensors attached to both feet, shanks, thighs and the pelvis. EMG recordings are acquired using PLUX sEMG sensors applied at various leg muscles. The paper gives not only a detailed description of the data base, its webpage and the used terms (scenario, proband, experiment and trial), but also an overview about the experimental setup, the acquisition of data and the procedure of the experiments, the data processing and evaluation. Results of exemplary applications are described in the second part of the paper. Here the focus is set on the performance of walking: the individual ability to control, to repeat and to reproduce the pace or the dependence of gait parameters on the pre-set velocity.

Paper Nr: 12
Title:

Changes in a Chaotic Fluctuation of Eye Movement Produced by Stiff Shoulder Treatment

Authors:

Eri Shibayama, Kasumi Tanaka and Taira Suzuki

Abstract: Many Japanese people experience physical symptoms known as Katakori (shoulder stiffness or neck pain), which is considered a psychosomatic phenomenon that is strongly correlated with psychological stress and stress caused by human relationships. This study examined changes in the chaos of eye movement accompanied by changes in depression by treating shoulder stiffness. Participants scoring over 1SD from the mean score having high depression and shoulder stiffness were included in the intervention group and provided stretching intervention. The control group included participants having high depression and shoulder stiffness, but they were not provided with the stretching intervention. Moreover, participants scoring lower than 1SD from the mean score having low depression and no shoulder stiffness were included in the Low group. The experimental group exercised using a neck and shoulder relaxation technique using a stretch pole recommended by the stretch pole official site. The results of the experiment indicated that eye movement LLE(Largest Lyapunov Exponent) changed more significantly when the depression level was high. Moreover, LLE and the degree of change in LLE decreased after receiving treatment for the stiff shoulder. It is suggested that the chaotic fluctuation of eye movement might decrease when depression improved, i.e., in people with low depression.

Paper Nr: 17
Title:

Assessing the Impact of Idle State Type on the Identification of RGB Color Exposure for BCI

Authors:

Alejandro A. Torres-García, Luis A. Moctezuma and Marta Molinas

Abstract: Self-paced Brain-Computer Interfaces (BCIs) are desirable for allowing the BCI’s user to control a BCI without a cue to indicate him/her when to send a command or message. As a first step towards a self-paced color-based BCI, we assessed if a machine learning algorithm can learn to distinguish between primary color exposure and idle state. In this paper, we record and analyze the EEG signals from 18 subjects for assessing the feasibility of distinguishing between color exposure and idle states. Specifically, we compare separately the performances obtained in the classification of two different types of idle states (one relaxation-related and another attention-related) and color exposure. We characterize the signals using two different ways based on discrete wavelet transform and Empirical Mode Decomposition (EMD), respectively. We trained and tested two different classifiers, support vector machine (SVM) and random forest. The outcomes provide experimental evidence that a machine learning algorithm can distinguish between the two classes (exposure to primary colors and idle states), regardless of the kind of idle state analyzed. The more consistent outcomes were obtained using EMD-based features with accuracies of 92.3% and 91.6% (considering a break and an attention-related task as the idle states). Also, when we discard the epochs’ onset the performances were 91.8% and 94.6%, respectively.

Paper Nr: 21
Title:

Two-stage Artificial Intelligence Clinical Decision Support System for Cardiovascular Assessment using Convolutional Neural Networks and Decision Trees

Authors:

Shahab Pasha, Jan Lundgren, Marco Carratù, Patrik Wreeby and Consolatina Liguori

Abstract: This paper describes an artificial-intelligence–assisted screening system implemented to support medical cardiovascular examinations performed by doctors. The proposed system is a two-stage supervised classifier comprising a convolutional neural network for heart murmur detection and a decision tree for classifying vital signs. The classifiers are trained to prioritize higher-risk individuals for more time-efficient assessment. A feature selection approach is applied to maximize classification accuracy by using only the most significant vital signs correlated with heart issues. The results suggest that the trained convolutional neural network can learn and detect heart sound anomalies from the time-domain and frequency-domain signals without using any user-guided mathematical or statistical features. It is also concluded that the proposed two-stage approach improves diagnostic reliability and efficiency.

Paper Nr: 22
Title:

Analysis of ECG and PCG Time Delay around Auscultation Sites

Authors:

Xinqi Bao, Yansha Deng, Nicholas Gall and Ernest N. Kamavuako

Abstract: Phonocardiogram (PCG) and Electrocardiogram (ECG) are the two important signals for cardiac preliminary diagnosis. Using ECG as a reference for segmenting the PCG signal is a simple but reliable technique for the devices with integration capability of PCG and ECG recording. The aim of this work is to analyse the time delay between ECG and PCG at each auscultation site. To do so, we performed the experiments on 12 healthy subjects, where the ECG and PCG signals were collected simultaneously at two sites at each time. Our results reveal that 1) the inter-distance of the electrodes for ECG does not affect the occurrence time of the R-peak. 2) The delay between R-peak and onset of first heart sound (S1) depends on the auscultation site e.g. S1 onset occurs before the R-peak at auscultation site M. This study suggests that small integrated ECG-PCG devices can be made by reducing the distance between the ECG electrodes. In the meantime, distinguishing the auscultation location is necessary for performing more precise PCG segmentation using ECG as reference.

Paper Nr: 23
Title:

An Attention-based Architecture for EEG Classification

Authors:

Italo Zoppis, Alessio Zanga, Sara Manzoni, Giulia Cisotto, Angela Morreale, Fabio Stella and Giancarlo Mauri

Abstract: Emerging studies in the deep learning community focus on techniques aimed to identify which part of a graph can be suitable for making better decisions and best contributes to an accurate inference. These researches (i.e., “attentional mechanisms” for graphs) can be applied effectively in all those situations in which it is not trivial to capture dependency between the involved entities while discharging useless information. This is the case, e.g., of functional connectivity in human brain, where rapid physiological changes, artifacts and high inter-subject variability usually require highly trained clinical expertise. In order to evaluate the effectiveness of the attentional mechanism in such critical situation, we consider the task of normal vs abnormal EEG classification using brain network representation of the corresponding EEG recorded signals.

Paper Nr: 27
Title:

Detection of Abnormalities in Electrocardiogram (ECG) using Deep Learning

Authors:

João Pestana, David Belo and Hugo Gamboa

Abstract: The Electrocardiogram (ECG) cyclic behaviour gives insights on a subject’s emotional, behavioral and cardiovascular state, but often presents abnormal events. The noise made during the acquisition, and presence of symptomatic patterns are examples of anomalies. The proposed Deep Learning framework learns the normal ECG cycles and detects its deviation when the morphology changes. This technology is tested in two different settings having an autoencoder as base for learning features: detection of three different types of noise, and detection of six arrhythmia events. Two Convolutional Neural Network (CNN) algorithms were developed for noise detection achieving accuracies of 98.18% for a binary-class model and 70.74% for a multi-class model. The development of the arrhythmia detection algorithm also included a Gated Recurrent Unit (GRU) for grasping time-dependencies reaching an accuracy of 56.85% and an average sensitivity of 61.13%. The process of learning the abstraction of a ECG signal, currently sacrifices the accuracy for higher generalization, better discriminating the presence of abnormal events in ECG than detecting different types of events. Further improvement could represent a major contribution in symptomatic screening, active learning of unseen events and the study of pathologies to support physicians in the future.

Paper Nr: 30
Title:

Inferring Low-level Mental States of Mobile Users from Plethysmogram Features by Regression Models based on Kernel Method

Authors:

Toshiki Iso

Abstract: To infer user’s response when using mobile services without direct interrogation, we propose an algorithm that analyzes earlobe plethysmograms to determine low-level mental states such as ‘relax’, ‘concentration’, ‘awake’. We use subject’s responses acquired in a subjective evaluation as indicative of low-level mental states when subjects use some mobile contents. In order to draw an inference of low-level mental states based on plethysmogram features, our proposed algorithm uses a kernel-based regression model such as Gaussian Process Regression (GPR) or Support Vector Regression (SVR). Our evaluations show that features effective for inferring user’s low-level mental states can be extracted from plethysmograms by using regression and Automatic Relevance Determination (ARD); the regression performance of GPR and SVR are described.

Paper Nr: 32
Title:

Algorithm for Extracting Initial and Terminal Contact Timings during Treadmill Running using Inertial Sensors

Authors:

Laura Prijot, Cédric Schwartz, Julien Watrin, Alex Mendes, Jean-Louis Croisier, Bénédicte Forthomme, Vincent Denoël, Olivier Brüls and Mohamed Boutaayamou

Abstract: Inertial measurement units (IMUs) are now considered as an economical solution for long term assessment in real conditions. However, their use in running gait analysis is relatively new and limited. Detecting the timing at which the foot strikes the ground (initial contact, IC) and the timing at which the foot leaves the ground (terminal contact, TC) gives access to many relevant temporal parameters such as stance, swing or stride durations. In this paper, we present an original algorithm to extract IC and TC timings and associated parameters from running data. These data have been measured using a newly developed IMU-based hardware system in ten asymptotic participants who ran at three speeds (slow, normal, and fast) with different running patterns (natural, rearfoot strike, mid-foot strike, and forefoot strike). This algorithm has been validated against a 200 Hz video camera based on 7056 IC and TC timings and 6861 temporal parameters. This algorithm extracted ICs and TCs with an accuracy and precision of (median [1st quartile; 3rd quartile]) 5 ms [-5 ms, 15 ms] and 0 ms [-5 ms, 5 ms], respectively. The relative errors in the extraction of stride and stance durations are -1.56 ± 3.00% and 0.00 ± 1.32%, respectively.

Paper Nr: 40
Title:

Clustering of Voice Pathologies based on Sustained Voice Parameters

Authors:

Alessa Anjos de Oliveira, Maria E. Dajer, Paula O. Fernandes and João P. Teixeira

Abstract: Signal processing techniques can be used to extract information that contribute to the detection of laryngeal disorders. The goal of this paper is to perform a statistical analysis through the boxplot tool from 832 voice signals of individuals with different laryngeal pathologies from the Saarbrücken Voice Database in order to create relevant groups, making feasible an automatic identification of these dysfunctions. Jitter, Shimmer, HNR, NHR and Autocorrelation features were compared between several groups of voice pathologies/conditions, resulting in three identified clusters.

Paper Nr: 43
Title:

Fractional Order Analysis of the Activator Model for Gene Regulation Process

Authors:

Hisham H. Hussein, Shaimaa A. Kandil and Khadeeja Amr

Abstract: Mathematical modeling for gene regulation process is very important for future prediction and control of diseases on the hereditary level. This paper presents a complete fractional dynamical analysis for an activator gene regulation model. The study of the system's phase planes portraits and the variables' transient responses starting from different initial points are presented and discussed. The effect of the fractional parameter within the differential operator is investigated. The simulation results show that the fractional parameter (𝛼) is effective in the process of synthesizing proteins and the gene regulation process stability.

Paper Nr: 44
Title:

Methods of the Pulse Wave Velocity Estimation based on Pneumatic Blood Pressure Sensor Data and Synchronous ECG Records

Authors:

V. E. Antsiperov, G. K. Mansurov and A. S. Bugaev

Abstract: The article discusses a new method for diagnosing atherosclerosis with the help of a pneumatic blood pressure sensor recently developed by the authors. Since atherosclerosis is a progressive disease characterized by the deposition of cholesterol and certain fractions of lipoproteins on the walls of blood vessels, it is always accompanied by an increase in stiffness of the artery walls. One technique to assess arterial stiffness, is the measurement of arterial pulse wave velocity, that is the distance traveled by blood flow divided by the time it takes to travel that distance. So the direct method for estimating pulse wave velocity is to measure the transit time of a pulse wave between a pair of artery points by means, for example, of any tonometric sensors. In this connection the paper discusses the possibility of using a new type of sensors developed by the authors —– pneumatic sensors —to measure the pulse wave transit time. However, given the existing features of these sensors and, accordingly, the special features of pressure measurements, it was necessary to significantly modify the direct method for estimating pulse wave velocity. The main modification characterizing the new, indirect method consists in evaluating the delay of the pulse wave at the points of the artery with respect to some characteristic moment of a synchronous ECG (e.g. the time moment of R–peak, that corresponds to heart ventricles contractions). The details of this method and its modification in the form of a simplified single–point method of estimating pulse wave velocity form the main content of the work.

Paper Nr: 45
Title:

Improving Dysarthric Speech Intelligibility using Cycle-consistent Adversarial Training

Authors:

Seung H. Yang and Minhwa Chung

Abstract: Dysarthria is a motor speech impairment affecting millions of people. Dysarthric speech can be far less intelligible than those of non-dysarthric speakers, causing significant communication difficulties. The goal of our work is to develop a model for dysarthric to healthy speech conversion using Cycle-consistent GAN. Using 18,700 dysarthric and 8,610 healthy Korean utterances that were recorded for the purpose of automatic recognition of voice keyboard in a previous study, the generator is trained to transform dysarthric to healthy speech in the spectral domain, which is then converted back to speech. Objective evaluation using automatic speech recognition of the generated utterance on a held-out test set shows that the recognition performance is improved compared with the original dysarthic speech after performing adversarial training, as the absolute SER has been lowered by 33.4%. It demonstrates that the proposed GAN-based conversion method is useful for improving dysarthric speech intelligibility.

Paper Nr: 48
Title:

Predicting Function Related Pain Outcomes using Comorbidity and Age Dependent Model

Authors:

Aleksandar Jeremic, Dejan Nikolic, Milena Kostadinovic and Milena S. Milicevic

Abstract: Effective pain management can significantly improve quality of life and outcomes for various types of patients (e.g. elderly, adult, young). In order to improve our understanding of patients’ response to pain we need to develop adequate signal processing techniques that would enable us to understand underlying interdependencies. To this purpose in this paper we develop several different algorithms that can predict function related pain outcomes using a large database obtained as a part of the national health survey. As a part of the survey the respondents provided detailed information about general health care state, acute and chronic problems as well as personal perception of pain associated with performing two simple talks: walking on the flat surface and walking upstairs. We model the correspondent responses using parametric and non-parametric models and use health indicators (both chronic and acute) as explanatory variables. For the binomial model we propose parametric age dependent model and then compare its performance to the performance of the multinomial and histogram models.

Paper Nr: 50
Title:

Cell Segmentation by Image-to-Image Translation using Multiple Different Discriminators

Authors:

Sota Kato and Kazuhiro Hotta

Abstract: This paper presents a cell image segmentation method by improving the pix2pix. Pix2pix improves the accuracy by competing a generator and a discriminator. The relationship of generator and discriminator is likened as follows. A generator is a fraudster who creates a fake image to fool the discriminator. A discriminator is a police officer who checks the fake image created by the generator. If we increase the number of police officers and different police officers are used, they have different roles and various viewpoints are used to check the fake image. In experiments, we evaluate our method on segmentation problem of cell images. We compared our method with conventional pix2pix using one discriminator. As a result, the accuracy will be improved. Thus, we propose to use multiple different discriminators to improve the segmentation accuracy of pix2pix. We confirmed that our proposed method outperformed conventional pix2pix and pix2pix using multiple same discriminators.

Paper Nr: 52
Title:

Investigating the Gait of Lower Limb Amputees Regarding the Present Classification of Mobility Grades

Authors:

Katja Orlowski, Kai-Uwe Mrkor, Harald Loose, Stephanie John and Kerstin Witte

Abstract: The mobility grade determined for German patients with a lower limb amputation based on the profile survey, which is a subjective classification in one of the five mobility grades (0 to 4). It is recommendable to establish objective examinations to determine the mobility grade of lower limb amputees. Gait parameters captured with mobile sensors could be suitable for the distinction between amputees of the different groups (grade G2, G3 or G4). Within a study, standard gait parameters were determined with the InvestiGAIT system based on inertial sensors. A descriptive analysis of the data of the twenty-one subjects (G2: 4, G3: 6, G4: 11) indicates that there are gait parameter (especially gait velocity, step and stride length) which can be used to make the classification of the three mobility grades. The temporal gait parameters (stride duration, swing and stance phase, one-leg-stance and double-leg-support) as well as angles during heel strike and toe off can be additionally used for the classification. Nevertheless, further investigations have to done to get a larger database in order to confirm the presented results regarding generalization and to check, whether the found classification can be implemented as a kind of decision support system.

Paper Nr: 3
Title:

A Study on Variation in EMG Trends under Different Muscular Energy Condition for Repeated Isokinetic Dumbbell Curl Exercise

Authors:

Rushda B. Ahmad, Nadeem A. Khan and Muhammad U. Rizwan

Abstract: Quantification and detection of accumulated muscle fatigue to assess muscle condition is of pivotal importance in sports and injury prevention. Existing research on assessment of exercise-induced muscle fatigue provides information regarding change in EMG waveform of the muscle during a single set of exercise performed. However, currently there is no study which discusses the variation in EMG activity of same subject under different muscle conditions when the exercise is repeated at constant force level. This paper investigates the changes in muscular energy under different conditions using endurance time, mean power frequency (MPF) and integrated EMG (IEMG) as metrics. This paper presents an initial study on EMG data acquired from subjects subjected to repeated isokinetic contractions. The aim of the study is to focus on inter-subject and intra-subject variations in EMG data. Such studies are very limited. Most of the studies focus on isometric contraction and make use of average data of the subjects from which data is collected once per protocol. Such reports do not bring forward the inter-subject and intra-subject variability. Because of huge variability, this study does not validate use of Global Fatigue Index for determination of muscle force and fatigue. Study of this variability is important if any autonomous system is developed for individuals using the EMG data that can accurately make detection and prediction on individual basis. Moreover this study suggests, peak MPF value and average slope of MPF curve during transition to fatigue stage as useful features to predict remaining time to fatigue till failure point.

Paper Nr: 5
Title:

High Performance Multi-class Motor Imagery EEG Classification

Authors:

Gul H. Khan, M. A. Hashmi, Mian M. Awais, Nadeem A. Khan and Rushda Basir

Abstract: Use of Motor Imagery (MI) in Electroencephalography (EEG) for real-life Brain Computer Interface applications require high performance algorithms that are both accurate as well as less computationally intensive. Common Spatial Pattern (CSP) and Filter Bank Common Spatial Pattern (FBSP) based methods of feature extraction for MI-classification has been shown very promising. In this paper we have advanced this frontier to present a new efficient approach whose variants out compete in accuracy (in terms of kappa values) with the existing approaches with the same or smaller feature set. We have demonstrated that use of one mu band and three beta sub-bands is very ideal both from the point-of-view of accuracy as well as computational complexity. We have been able to achieve the best reported kappa value of 0.67 for Dataset 2a of BCI Competition IV using our approach with a feature vector of length 64 directly composed out of FBCSP transformed data samples without the need of further feature selection. The feature vector of size 32 directly composed from FBCSP data is enough to outcompete existing approaches with regard to kappa value achievement. In this paper we also have systematically reported experiments with different classifiers including kNN, SVM, LDA, Ensemble, ANN and ANFIS and different lengths of feature vectors. SVM has been reported as the best classifier followed by the LDA.

Paper Nr: 13
Title:

Evaluation of a New Functional near Infrared Spectroscopy (fNIRS) Sensor, the fNIRS Explorer™, and Software to Assess Cognitive Workload during Ecologically Valid Tasks

Authors:

Bethany K. Bracken, Colette Houssan, John Broach, Andrew Milsten, Calvin Leather, Sean Tobyne, Aaron Winder and Mike Farry

Abstract: Medical personnel and first responders are often deployed to dangerous environments where their success at saving lives depends on their ability to act quickly and effectively. During training, non-invasive measurement of cognitive performance can provide trainers with insight into medical students’ skill mastery. Functional Near-Infrared Spectroscopy (fNIRS) is a direct and quantitative method to measure ongoing changes in brain blood oxygenation (HbO) in response to a person’s evolving cognitive state (i.e., cognitive workload or mental effort) that has only recently received significant attention for use in the real world. The work presented here includes data collection with a new, more portable, rugged design of an fNIRS sensor to test the functionality of this new sensor design and our ability to measure cognitive workload in a medical simulation training environment. To assess sensor and model accuracy, during breaks from the training, participants completed a gold-standard, laboratory task and during training in a medical simulation environment. Linear mixed model ANOVA showed that when we accounted for fixed effects of intercept and slope in our model, there was a significant difference in the HbR Ch1 model for n-back load (coef=0.009, p=0.034), intercept (coef=0.96, p=1.21e-07***), and load (slope) (coef=-0.09, p=0.03). Future work will present results of the data collected during the disaster response medical simulation training.

Paper Nr: 20
Title:

Electroencephalography-based Motor Hotspot Detection

Authors:

Ga-Young Choi, Chang-Hee Han, Hyunmi Lim, Jeonghun Ku, Won-Seok Kim and Han-Jeong Hwang

Abstract: The motor-evoked potential (MEP) induced by transcranial magnetic stimulation (TMS) has been generally used to identify a motor hotspot, and it has been used as a target location for transcranial electrical stimulation (tES). However, the traditional MEP-based method needs a bulky TMS device, and it involves the empirical judgement of an expert. In this study, we propose a machine-learning-based motor hotspot identification method using electroencephalography (EEG) that is portably acquired in a tES device. EEG data were measured from ten subjects while they performed a simple finger tapping task. Power spectral densities (PSDs) were extracted from the EEG data as features, and they were used to train and test artificial neural network (ANN). The 3D coordinate information of individual motor hotspots identified by TMS were also used as the ground-truth motor hotspot locations in ANN, and they were compared with those estimated by ANN. A minimum distance between the motor hotspots identified by TMS and EEG features was only 0.24 cm, demonstrating the feasibility of our proposed novel motor hotspot identification method based on EEG features.

Paper Nr: 25
Title:

Learning Human Behaviour Patterns by Trajectory and Activity Recognition

Authors:

Letícia Fernandes, Marília Barandas and Hugo Gamboa

Abstract: The world’s population is ageing, increasing the awareness of neurological and behavioural impairments that may arise from the human ageing. These impairments can be manifested by cognitive conditions or mobility reduction. These conditions are difficult to be detected on time, there is a lack of routine screening which demands the development of solutions to better assist and monitor human behaviour. This study investigates the question of what we can learn about human behaviour patterns from the rich and pervasive mobile sensing data. Data was collected over 6 months, measuring two different human routines through human trajectory analysis and activity recognition comprising indoor and outdoor environment. A framework for modelling human behaviour was developed using human motion features, extracted with and without previous knowledge of the user’s behaviour. The human patterns were modelled through probability density functions and clustering approaches. Using the learned patterns, inferences about the current human behaviour were continuously quantified by an anomaly detection algorithm where distance measurements were used to detect significant changes in behaviour. Experimental results demonstrate the effectiveness of the proposed framework that revealed an increased potential to learn behavioural patterns and detect anomalies.

Paper Nr: 26
Title:

A Novel Approach for Modelling the Relationship between Blood Pressure and ECG by using Time-series Feature Extraction

Authors:

Stefan Kochev, Neven Stevchev, Svetlana Kocheva, Tome Eftimov and Monika Simjanoska

Abstract: This paper addresses the ECG-blood pressure relationship - a fact that physicians have discussed for years. The hypothesis set in the paper is that blood pressure is related to electrocardiogram (ECG) and that the systolic blood pressure (SBP) and diastolic blood pressure (DBP) values can be predicted by using information only from a given ECG signal. Therefore, we established a protocol for creating a database considering measurements from real patients in ambulance environment, and consequently developed methodology for analysing the collected measurements. The proposed methodology follows two steps: i) first the signals are considered as time series data, and ii) a time series feature extraction method is applied to extract the important features from the ECG signals. Hereafter, a novel Machine learning method is applied (CLUS) that produced best results among the traditionally-used Machine learning methods. The best results obtained are 12.81 ± 2.66 mmHg for SBP and 8.12 ± 1.80 mmHg for DBP. After introducing calibration method the obtained mean absolute errors (MAEs) reduced to 6.93 ± 4.70 mmHg for SBP, and 7.13 ± 4.48 mmHg for DBP. Given the latest literature, the results are appropriately compared and confirm the relation between the ECG signal and the blood pressure.

Paper Nr: 28
Title:

Time-frequency Features for sEMG Signals Classification

Authors:

Somar Karheily, Ali Moukadem, Jean-Baptiste Courbot and Djaffar O. Abdeslam

Abstract: This paper proposes a new approach for the identification of hand movements in order to control prosthetic hand. sEMG signals were used to identify movements by using two time frequency transforms: Short Time Fourier Transform and Stockwell transform. Then, we apply Singular Value Decomposition (SVD) to decrease the features dimension and to form the final features’ vector. These extracted features were used by two kinds of classifiers: K nearest neighbours and linear discriminant analysis. Finally, we numerically study these methods on a database of 10 subjects and 17 hand gestures.

Paper Nr: 35
Title:

Indirect Posture Correction System without Additional Equipment using Display Content Rotation

Authors:

Akira Takahashi, Masahiro Inazawa, Yuki Ban and Shin’ichi Warisawa

Abstract: The poor posture of office workers who engage in PC work is a problem. Poor posture may cause musculoskeletal disorders. In previous studies on the posture correction system, there are some problems. One of them is that the posture correction system may interfere with the tasks of office workers. There is a posture correction system which does not interfere with the tasks; however, it requires large equipment. In order to solve these problems, we proposed a system that corrects the posture of office workers by rotating the content in the display. This is a method that rotates content in the opposite direction of head movement. We expect that users unconsciously move their head to look at content. We evaluated our porposed method with 2 user studies. User study 1 was conducted to verify whether the angle of the spine changed by rotating the display. It suggested that rotating the dislay induce the head to adjust laterally, not longitudinally. In study 2, we succeeded in moving the direction of the angle of spine of experimental participants to the right by an average of 1 deg by rotating the content right. Thus, we showed the possibility of posture correction without large-scale equipment.

Paper Nr: 37
Title:

Impact of Task-evoked Mental Workloads on Oculo-motor Indices during a Manipulation Task

Authors:

Minoru Nakayama and Yoshiya Hayakawa

Abstract: Oculo-motor metrics which included metrics of microsaccades were analysed in response to the level of cognitive mental workload during a manipulation task. While some oculo-motor metrics correlate with the estimated scores of the mental workload, these metrics mutually correlate with each other. A model of causal relationship was created using all metrics, including subjective measurements. Metrics of microsaccades perform the function of intermediating behaviour between participant’s subjective assessments and conventional ocular measurements, such as saccades and pupil responses.

Paper Nr: 42
Title:

Deep-learning in Identification of Vocal Pathologies

Authors:

Felipe L. Teixeira and João P. Teixeira

Abstract: The work consists in a classification problem of four classes of vocal pathologies using one Deep Neural Network. Three groups of features extracted from speech of subjects with Dysphonia, Vocal Fold Paralysis, Laryngitis Chronica and controls were experimented. The best group of features are related with the source: relative jitter, relative shimmer, and HNR. A Deep Neural Network architecture with two levels were experimented. The first level consists in 7 estimators and second level a decision maker. In second level of the Deep Neural Network an accuracy of 39,5% is reached for a diagnosis among the 4 classes under analysis.

Paper Nr: 46
Title:

Visualizing and Modifying Difficult Pixels in Cell Image Segmentation

Authors:

Daisuke Matsuzuki and Kazuhiro Hotta

Abstract: In this paper, we visualize and modify difficult pixels to recognize for deep learning. In general, an image includes pixels that are easy or difficult to recognize. At the final layer, many deep learning methods use a softmax function to convert the outputs of network to probabilities. Pixels with small maximum probability are often difficult to recognize. We visualize those difficult pixels in a test image using the relationship between confidence and pixel-wise difficulty. By visualizing difficult pixels, we confirm the connection of cell membrane that could not be recognized by conventional method. We can connect the cell membrane by modifying difficult pixels. In experiments, we use cell image of mouse liver dataset including three classes; “cell membrane”, “cell nucleus” and “cytoplasm”. Our proposed method shows high recall score for “cell membrane”. We also confirmed the connection of cell membrane in qualitative evaluation.

Paper Nr: 49
Title:

ECG based Human Identification using Short Time Fourier Transform and Histograms of Fiducial QRS Features

Authors:

Abdullah Biran and Aleksandar Jeremic

Abstract: Human identification from the biological signal the Electrocardiogram (ECG) has been demonstrated in several studies. In this paper, we present a new technique for personal identification using short time Fourier transform (STFT) and histograms of four fiducial QRS features. We examined the applicability of our methodology on 162 ECG records of 81 subjects from the publicly available ECG ID data base. Our experiments show that the normalized Euclidean STFT distance can identify individuals with 95 % accuracy. Hence, with fusing six histogram distances computed from the QRS fiducial features and applying majority voting, the identification accuracy increases up to 100 %. These findings indicate that ECG is sufficiently unique signal and can be useful as biometric identifier.

Paper Nr: 53
Title:

Detecting Neonatal Seizures using Short Time Fourier Transform and Frechet Distance

Authors:

Aleksandar Jeremic and Dejan Nikolic

Abstract: Recently there has been an increase in the number of long-term cot-bed EEG systems being implemented in clinical practice in order to monitor neurological development of neonatal patients. Consequently a significant research effort has been made in the development of automatic EEG data analysis tools including but not limited to seizure detection as seizure frequency and/or intensity are one of the most important indicators of brain development. In this paper we propose to evaluate time dependent power spectral density using short time Fourier transform and using Frechet distance measure to detect presence and/or absence of seizures. We propose to use three different distance measures as they capture different properties of the corresponding PSD matrices. We evaluate the performance of the proposed algorithms using real data set obtained in the NICU of the McMaster University Hospital. In order to benchmark performance of our proposed techniques we trained and tested a support vector machine (SVM) classifier.

Paper Nr: 55
Title:

Possibilities of Predicting Arterial Pressure by Means of Heart Rate Variability

Authors:

Anton Y. Dolganov

Abstract: The paper shows results of the study which aims to predict values of arterial pressure by means of heart rate variability features. A total list of 64 features was tested, which included features in time and frequency domain, as well as non-linear features. As a means of feature selection, the genetic programming was used. In particular binary encoding was used for generation of features in combinations as well as degree of the polynomial. Data of 50 students-volunteers recorded in sitting position was used. Results of the study suggests that certain heart rate variability features can be used for prediction of the change of arterial pressure. Perspectives and future plans for results improvement were described.