BIOSIGNALS 2023 Abstracts

Full Papers Short Papers
Full Papers
Paper Nr: 2
Title:

Constrained ALS for Estimation of Human Upper Limb Synergies
Authors:

Rabya Bahadur and Saeed U. Rehman
Abstract: Human body movement is a complex task that requires the control of multiple joints via a network of muscles. The biological signals, particularly electromyography (EMG), are correlated by nature. The brain transmits these signals through the neuromuscular transmission system of the body. The combination of muscular activation for each particular movement presents a set of weights known as synergies. The traditional alternating least square (ALS) based non-negative matrix factorization (NMF) gets trapped in local minima for co-linear data. Therefore, it is not a suitable method for extracting muscle synergies. This paper advocate using l2-norm as an additional constraint for ALS-based NMF. The addition of l2-norm decorrelates the data, resulting in a better estimation of synergistic weights. For our results, we acquired EMG signals from six healthy subjects. Both plain and regularized NMF were used to extract the synergies. The synergies acquired via plain NMF have a higher cross-correlation within and indicate the triggering of the same muscles irrespective of the targeted isometric contraction. In contrast, the regularized NMF synergies targeted the correct muscular set for a particular isometric contraction. Our results show that the synergies acquired via regularized NMF are also more correlated with the physiologically inspired synergies.
Download
Paper Nr: 4
Title:

Machine Learning Algorithms for Mouse LFP Data Classification in Epilepsy
Authors:

Antonis Golfidis, Michael Vinos, Nikos Vassilopoulos, Eirini Papadaki, Irini Skaliora and Vassilis Cutsuridis
Abstract: Successful preictal, interictal and ictal activity discrimination is extremely important for accurate seizure detection and prediction in epileptology. Here, we introduce an algorithmic pipeline applied to local field potentials (LFPs) recorded from layers II/III of the primary somatosensory cortex of young mice for the classification of endogenous (preictal), interictal, and seizure-like (ictal) activity events using time series analysis and machine learning (ML) models. Using the HCTSA time series analysis toolbox, over 4000 features were extracted from the LFPs after applying over 7700 operations. Iterative application of correlation analysis and random-forest-recursive-feature-elimination with cross validation method reduced the dimensionality of the feature space to 22 features and 27 features, in endogenous-to-interictal events discrimination, and interictal-to-ictal events discrimination, respectively. Application of nine ML algorithms on these reduced feature sets showed preictal activity can be discriminated from interictal activity by a radial basis function SVM with a 0.9914 Cohen kappa score with just 22 features, whereas interictal and seizure-like (ictal) activities can be discriminated by the same classifier with a 0.9565 Cohen kappa score with just 27 features. Our preliminary results show that ML application in cortical LFP recordings may be a promising research avenue for accurate seizure detection and prediction in focal epilepsy.
Download
Paper Nr: 5
Title:

Stylohyoid and Posterior Digastric Measurement with Intramuscular EMG, Submental EMG and Swallowing Sound
Authors:

Adrien Mialland, Ihab Atallah and Agnès Bonvilain
Abstract: The stylohyoid and the posterior digastric muscles have essentially been measured through indirect imaging method because of the difficulty to measure them. They are small neck muscles, close to each other, that cannot easily be accessed independently. Yet, they showed promising results for a robust and safe indwelling detection of swallowing, both in terms of timing and recruitment. The possibility to thoroughly establish their relevance through their direct functional analysis would enable the development of an implantable active artificial larynx, that would protect the airway during swallowing detection. Therefore, we set up the first standardized procedure that allows their direct measurement through intramuscular electromyography (EMG) and that we report in this paper. We also used submental surface EMG and swallowing sound modalities to access the major time points of the swallowing process. Finally, various exercises, along with swallowing, were performed by the volunteers. 16 peoples were measured with our new procedure, and both the stylohyoid and the posterior digastric could be measured independently with no difficulty. Timings and tasks comparison are therefore ongoing.
Download
Paper Nr: 15
Title:

Impact of Body Position on Imaging Ballistocardiographic Signals
Authors:

Alexander Woyczyk and Sebastian Zaunseder
Abstract: Current works direct at the unobtrusive acquisition of vital parameters from videos. The most common approach exploits subtle color variations. The analysis of cardiovascular induced motion from videos (imaging ballistocardiography, iBCG) is another approach that can supplement the analysis of color changes. The presented study systematically investigates the impact of body position (supine vs. upright) on iBCG. Our research directs at heart rate estimation by iBCG and on the possibility to analyse ballistocardiographic waveforms from iBCG. We use own data from 30 healthy volunteers, who went through repeated orthostatic maneuvers on a tilt table. Processing is done according to common procedures for iBCG processing including feature tracking, dimensionality reduction and bandpass filtering. Our results indicate that heart rate estimation works well in supine position (root mean square error of heart rate estimation 5.68 beats per minute). The performance drastically degrades in upright (standing) position (root mean square error of heart rate estimation 21.20 beats per minute). With respect to analysis of beat waveforms, we found large intra-subject and inter-subject variations. Only in few cases, the resulting waveform closely resembles the ideal ballistocardiographic waveform. Our investigation indicates that the actual position has a large effect on iBCG and should be considered in algorithmic developments and testing.
Download
Paper Nr: 17
Title:

Comparison of the Electrophysiological Myoelectrical Activity Evolution in Induction of Labor with Pharmacological and Mechanical Methods
Authors:

Alba Diaz-Martinez, Yiyao Ye-Lin, Rogelio Monfort-Ortiz, Javier Garcia-Casado, Iria Rey-Ferreira, Felix Nieto-del-Amor, Vicente Diago-Almela, Jose L. Martinez-de-Juan and Gema Prats-Boluda
Abstract: Induction of labour (IOL) refers to triggering the contractions onset, either by pharmacological (PIOL) or mechanical methods (MIOL), and is indicated when maternal and foetal well-being is compromised. There is great uncertainty regarding the success of IOL regardless of the method. In current clinical practice, it is based on assessment of cervical status by Bishop’s score and degree of uterine activity by tocography. However, Bishop’s score has been shown to be subjective and poorly reproducible and tocography requires constant repositioning and is severely affected by obesity. Meanwhile, electrohysterography (EHG) has surpassed traditional clinical measures in monitoring PIOL progress and predicting its outcome. Although there is no evidence of uterine myoelectric activity response of MIOL. Therefore, this work aimed to identify EHG-biomarkers to help to determine possible differences in myoelectric response between PIOL and MIOL success. For this purpose, the uterine response during the first 5h after Dinoprostone (PIOL) administration and Foley catheter (MIOL) insertion was compared by EHG. For PIOL, a significantly lower time to achieve active phase of labor and delivery, together with faster myoelectric response was found: slightly higher contraction force, significantly higher Mean Frequency and lower Spectral Entropy after 2.5h. Between-group differences were especially marked in Spectral Entropy (90-150 and 210-300min). Overall, this pioneering work has demonstrated the feasibility of EHG for the characterisation of evolution also in MIOL. Furthermore, the results suggest that EHG biomarkers may be useful in the IOL method comparison, although they should be cross-checked with expanded databases and further investigations.
Download
Paper Nr: 27
Title:

Speech-Based Supervised Learning Towards the Diagnosis of Amyotrophic Lateral Sclerosis
Authors:

Ricardo Cebola, Duarte Folgado, André V. Carreiro and Hugo Gamboa
Abstract: Amyotrophic Lateral Sclerosis (ALS) diagnosis requires extensive clinical examinations, often leading to delays and a burden to patients and their caregivers. Speech has emerged in the literature as a promising biomarker for neurodegenerative diseases capable of being integrated into telemonitoring solutions. We present a comprehensive study with several phonatory tasks and speech features to evaluate the generalisation potential of models for ALS diagnosis. We use a public dataset with sustained vowels (N=64) and data with ALS and healthy volunteers being collected from ongoing research trials (N=22). Two approaches were considered: i) sample-based, where the signals were divided into fixed-length windows, and ii) patient-based, where a voting system was implemented based on the sample-based classification of each patient. We achieved a mean diagnostic performance with an F1-score over 80%. The best scores for the sample and patient-based classifications are 96% and 100% for vowels, 96% and 95% for sentences and 82% and 87% for cough. Our findings support speech as a promising digital biomarker and pave the way for remote examination at patients’ residences, increasing the data available for clinicians for better diagnosis and prognosis of ALS.
Download
Paper Nr: 32
Title:

Event-Related Desynchronization Analysis During Action Observation and Motor Imagery of Transitive Movements
Authors:

Stefania Coelli, Alessandra Calcagno, Federico Temporiti, Roberto Gatti, Manuela Galli and Anna M. Bianchi
Abstract: Rehabilitation and motor skill learning approaches based on Action Observation (AO) and Motor Imagery (MI) rely on the assumption that the sensorimotor system is stimulated by AO and MI tasks similarly to the actual execution of a movement. An advantage of AO over MI is that it is less dependent on subject’s imagination ability, and a direct comparison of their effect on cortical activations during complex upper limb movements has been rarely examined. Therefore, in this study we compare sensorimotor event related desynchronization (ERD) patterns, as a measure of cortical activation, collected from 46 healthy volunteers performing AO and MI protocols. In both mu and beta sensorimotor rhythms a stronger ERD was elicited by AO, characterized by an evident lateralization in the contralateral side of the brain with respect to the limb involved in the observed movement.
Download
Paper Nr: 35
Title:

Wearable Data Generation Using Time-Series Generative Adversarial Networks for Hydration Monitoring
Authors:

Farida Sabry, Wadha Labda, Tamer Eltaras, Fatima Hamza, Khawla Alzoubi and Qutaibah Malluhi
Abstract: Collection of biosignals data from wearable devices for machine learning tasks can sometimes be expensive and time-consuming and may violate privacy policies and regulations. Successful and accurate generation of these signals can help in many wearable devices applications as well as overcoming the privacy concerns accompanied with healthcare data. Generative adversarial networks (GANs) have been used successfully in generating images in data-limited situations. Using GANs for generating other types of data has been actively researched in the last few years. In this paper, we investigate the possibility of using a time-series GAN (TimeGAN) to generate wearable devices data for a hydration monitoring task to predict the last drinking time of a user. Challenges encountered in the case of biosignals generation and state-of-the-art methods for evaluation of the generated signals are discussed. Results have shown the applicability of using TimeGAN for this task based on quantitative and visual qualitative metrics. Limitations on the quality of the generated signals were highlighted with suggesting ways for improvement.
Download
Paper Nr: 38
Title:

Hallucinated Heartbeats: Anomaly-Aware Remote Pulse Estimation
Authors:

Jeremy Speth, Nathan Vance, Benjamin Sporrer, Lu Niu, Patrick Flynn and Adam Czajka
Abstract: Camera-based physiological monitoring, especially remote photoplethysmography (rPPG), is a promising tool for health diagnostics, and state-of-the-art pulse estimators have shown impressive performance on benchmark datasets. We argue that evaluations of modern solutions may be incomplete, as we uncover failure cases for videos without a live person, or in the presence of severe noise. We demonstrate that spatiotemporal deep learning models trained only with live samples “hallucinate” a genuine-shaped pulse on anomalous and noisy videos, which may have negative consequences when rPPG models are used by medical personnel. To address this, we offer: (a) An anomaly detection model, built on top of the predicted waveforms. We compare models trained in open-set (unknown abnormal predictions) and closed-set (abnormal predictions known when training) settings; (b) An anomaly-aware training regime that penalizes the model for predicting periodic signals from anomalous videos. Extensive experimentation with eight research datasets (rPPG-specific: DDPM, CDDPM, PURE, UBFC, ARPM; deep fakes: DFDC; face presentation attack detection: HKBU-MARs; rPPG outlier: KITTI) show better accuracy of anomaly detection for deep learning models incorporating the proposed training (75.8%), compared to models trained regularly (73.7%) and to hand-crafted rPPG methods (52-62%).
Download
Paper Nr: 42
Title:

Estimation of Fluid Intake Volume from Surface Electromyography Signals: A Comparative Study of Seven Regression Techniques
Authors:

Iman A. Ismail and Ernest N. Kamavuako
Abstract: Insufficient fluid intake in older adults, in particular, is a worrying problem and an actual concern that warrants scrutiny. Monitoring fluid intake is essential to avoid dehydration and overhydration problems. This paper presents an investigation to estimate the fluid intake volume using surface Electromyographic (sEMG) sensors. Eleven subjects participated in the experiment, and sEMG recordings of swallows from cups, bottles, and straws were collected. Four features were extracted from the EMG signals. Seven regression algorithms were implemented for quantifying the volume of swallowed fluid: Random Forest (RF), Support Vector Regressor, K-nearest neighbour (KNN), Linear Regressor (LR), Decision Tree (DT), Lasso and Ridge. The mean sip volume across subjects was 14.85 ± 5.05 ml. Results showed that using Random Forest, the root mean square (RMSE) for estimating fluid intake volume using one the Mean Absolute Value feature gave 1.37 ± 1.1 ml. These results indicate a step forward in estimating fluid intake volume based on sEMG for hydration monitoring.
Download
Paper Nr: 50
Title:

Using Sparse Representation of EEG Signal from a Shallow Sparse Autoencoder for Epileptic Seizure Prediction
Authors:

Gul H. Khan, Nadeem A. Khan, Wala Saadeh and Muhammad B. Altaf
Abstract: Patients with epilepsy are affected with unexpected seizure events, which significantly diminish their quality of life. It is crucial to evaluate whether an epileptic patient’s brain state is indicative of a possible seizure occurrence so that necessary therapy or alarm can be generated on time. If seizures could be predicted before the onset, interventions may be applied to avoid further damage during seizure attack, and patients could take medications or other treatments to prevent seizures from occurring. This research describes a patient-specific technique for predicting epileptic seizures based on a hybrid model. Single layer sparse autoencoder is trained to obtain a aparse representation of the scalp electroencephalogram (EEG) signals. SVM classifier is used to categorize the sparse signal as inter-ictal or pre-ictal. Individual EEG channel analysis for seizure prediction are presented. In addition, various hidden sizes of the autoencoder for optimal sparse representation are anlyzed.The proposed model evaluates 13 patients from the CHB-MIT dataset and obtains a sensitivity of 98% and an area under the curve (AUC) of 98%. We have evaluated the performance of our hybrid strategy to both deep learning models and conventional procedures. The proposed method outperforms current seizure prediction techniques, proving its efficacy.
Download
Paper Nr: 53
Title:

Data Augmentation, Multimodality, Subject and Activity Specificity Improve Wearable Electrocardiogram Denoising with Autoencoders
Authors:

João A. Saraiva, Mariana Abreu, Ana S. Carmo, Ana Fred and Hugo Plácido da Silva
Abstract: Event detection based on biosignals continuously acquired by wearable devices has become an emergent topic. Particularly, real-time event detection with the electrocardiogram (ECG) has been explored to monitor heart conditions and epileptic seizures in the ambulatory. However, ECG acquired in the ambulatory is much more prone to noise and artifacts, due to the dynamic nature of these environments. Therefore, real-time and robust ECG denoising methods are crucial if event detection is meant to succeed. Denoising autoencoders (DAEs) are studied as robust and fast methods to attenuate ECG noise and artifacts. ECG data augmentation techniques are shown to effectively improve the performance of such a deep learning method. Activity and subject specific models are shown to output better ECG denoised estimates, than non-specific ones. And using accelerometry (ACC) as noise reference exemplifies how biosignal multimodality improves ECG attenuation of muscle and motion artifacts. Therefore, this work establishes effective design techniques to be considered when engineering ECG deep learning models.
Download
Short Papers
Paper Nr: 3
Title:

Assessing Situation Awareness (SA) Using Single- or Dual-Location functional Near Infrared Spectroscopy (fNIRS)
Authors:

Bethany K. Bracken, Aaron Winder, Brandon Hager, Mica R. Endsley and Elena K. Festa
Abstract: To operate effectively across a variety of environments, personnel (e.g., air traffic controllers, pilots, truck drivers, emergency response crews) need to be trained to the point at which their responses are automatic. If their responses require high mental effort when carried out in emergency situations, they may be unable to perform or to establish situation awareness (SA) needed to perform and to keep themselves safe. We have been developing a software application to assess cognitive workload (i.e., mental effort) during task performance using functional near-infrared spectroscopy (fNIRS). Here we present our work toward extending this human state assessment software to include SA. We used a driving task (Crundall & Kroll, 2018; Muela et al., 2021) in which participants saw a clip of someone driving from a first person perspective followed by a Level 3 SA (prediction) question asking what hazard was about to occur. Participants were 22 Brown University undergraduate and medical students (8 females) with an average age of 22.2 (SD=4.7) and 22 Army personnel in one of the U.S. Army installations with an average age of 49 (SD=11). We were able to predict performance on the SA questions using the fNIRS data, at the group level (mean accuracy = 65% in Brown students, 71% in Army personnel, and 65% in the combined datasets). We were also able to predict SA performance of individual participants with a mean accuracy of 69% (range = .45-.88). This adds to the growing literature indicating that neurophysiological information, even when data is acquired at a single location, is useful for predicting individual SA.
Download
Paper Nr: 6
Title:

AI and IoT Enabled Sleep Stage Classification
Authors:

Dimitrios Zografakis, Panagiotis Tsakanikas, Ioanna Roussaki and Konstantina-Maria Giannakopoulou
Abstract: Sleep is a key aspect affecting health, cognitive functionality, and human psychology on all occasions. Therefore, on the one hand, sleep greatly impacts the quality of life, while on the other hand poor health and/or psychology often deteriorate the quality of sleep. Moving beyond the golden standard for sleep studies, i.e. polysomnography, and building on the current state of the art in wearables, this paper aims to propose a deep learning approach that focuses on sleep stage classification, introducing the timeseries related information input to the classification. In this respect, smartwatch sensor measurements are used and a series of methods have been tested. The proposed approach constitutes a preliminary work on sleep stage classification introducing a novel approach of feature engineering incorporating the time-related information concerning the transition of the sleep stages via a Long Short-Term Memory (LSTM) encoding of the accelerometer data from smartwaches. The obtained results are compared with the outcomes of existing related approaches on the same open dataset as previously published. The respective evaluation exhibits promising findings and shortcomings compared to previous approaches and polysomnography analysis correspondingly. In addition, the choice of appropriate evaluation metrics has emerged, since traditional classification metrics such as accuracy, are not appropriate to capture the real performance in terms of the transition of the stages sequence in the resulted hypnograms.
Download
Paper Nr: 8
Title:

Spontaneous Cardiac-Locomotor Coupling in Healthy Individuals During Daily Activities
Authors:

Aurora Rosato, Matilda Larsson, Eric Rullman and Seraina A. Dual
Abstract: .During exercise, the locomotor and the cardiovascular system work in synergy to control the blood flow through the body. In particular, the muscle contraction generates rhythmic raising and lowering of intramuscular pressure, which in synergy supports cardiovascular function. This study aims to analyze spontaneous cardiac-locomotor coupling (CLC) events during daily activities using weareable sensors. We analyze the data set PMData, containing recordings from sixteen healthy subjects during five months. The data were acquired with a smartwatch and consist of step rate (SR), heart rate (HR) and daily surveys reporting the training sessions. Coupling is defined as being present when SR and HR are within 1% of each other (strong coupling) and within the 10% of each other (weak coupling). The results show that every subject presents occurrences of CLC while performing normal daily activities. In particular, strong coupling occurs more likely for longer activities (111 ± 34 min), at moderate intensity (100 steps min < SR > 130 steps min ). The presence of CLC during daily activities rises the question whether there is a physiological mechanism controlling this phenomenon, that should be investigated in future.
Download
Paper Nr: 9
Title:

Machine Learning Algorithm Development and Metrics Extraction from PPG Signal for Improved Robustness in Wearables
Authors:

Pedro Veiga, Rui Varandas and Hugo Gamboa
Abstract: Wearable devices application in the digital measurement of health has gained attention by researchers. These devices allow for data acquisition during real-life activities, resulting in higher data availability. They often include photoplethysmography (PPG) sensors, the sensor behind pulse oximetry which is a non-invasive method for continuous oxygen saturation measurements, an essential tool for managing patients undergoing pulmonary rehabilitation and an effective method for assessing sleep-disordered breathing. However, the current market focuses on heart rate measurements and lacks the robustness of clinical applications for SpO2 assessment. The most common obstacle in PPG measurements is the signal quality. Thus, in this work a solution was developed to evaluate the signal in three distinct qualities. A Random Forest classifier achieved accuracy scores of 79%, 80% for the models capable of differentiating between usable and unusable signals, and of 74% and 80% when distinguishing between optimal and suboptimal signals. Multi-class models achieved accuracy scores of 66% and 65%. Three clinically relevant metrics were also extracted from the PPG signal. The heart rate and respiratory rate algorithms resulted in performances similar to the ones found in the literature. However, while promising, more data is needed to reach statistical significance for the SpO 2 measurement.
Download
Paper Nr: 10
Title:

Annotation-Based Evaluation of Wrist EDA Quality and Response Assessment Techniques
Authors:

E. Pattyn, E. Lutin, A. Van Kraaij, N. Thammasan, D. Tourolle, I. Kosunen, D. Tump, W. De Raedt and C. Van Hoof
Abstract: Electrodermal activity (EDA) reflects changes in electrical conductivity of the skin via activation of the sympathetic nervous system. Ambulatory EDA measurements bring multiple challenges regarding quality assessment and response detection. A signal quality indicator (SQI) is one method to overcome these. This study aimed to investigate the transferability and generalizability of several open-source state-of-the-art SQIs and response detectors regarding their performance against manually annotated EDA of participants in rest. Three annotators identified artifacts and physiological responses in wrist EDA of 45 participants (10.75 hours). The F1-score, precision, and recall of several state-of-the-art SQIs and response detectors were computed on a subset of the annotated data (n=28). The SQIs and response detectors resulted in F1 scores between 3-16% and 18-32%, respectively. These results indicated that current SQIs and response indicators are not performant enough for EDA of subjects in rest, implying similar or worse outcomes for ambulatory EDA. It is suggested that SQIs must be adjusted based on the used device and set-up.
Download
Paper Nr: 13
Title:

Prediction of Sleep Stages Based on Wearable Signals Using Machine Learning Techniques
Authors:

Rodrigo D. Braga, Daniel Osório and Hugo Gamboa
Abstract: Sleep’s impact on mood and health is widely recognized by medical researchers with such understanding disseminating among average people in recent years. The main objective of this work was the development of machine learning algorithms for automatic sleep cycles detection. The features were selected based on the AASM manual, which is considered the gold standard for human technicians. For training the models we used MESA, a database containing 2056 full overnight unattended polysomnographies. With the goal of developing an algorithm that would only require a photoplethysmography (PPG) device to be able to accurately predict sleep stages and quality, the main channels used from this dataset were peripheral oxygen saturation and PPG. Testing the performance of Random forest, Gradient Boosting, Gaussian Naive-bayes, K-Nearest Neighbours, Support Vector Machine and Multilayer Perceptron classifiers, and using features extracted from the dataset, we achieved 80.50 % accuracy, 0.7586 Cohen’s kappa, and 77.38% F1-score, for five sleep stages, using a Multilayer Perceptron. To assess its performance in a real-world scenario we acquired sleep data and compared the classifications attributed by a popular sleep stage classification android app and our algorithm, resulting in a strong level of agreement (90.96% agreement, 0.8663 Cohen’s kappa), for four sleep stages.
Download
Paper Nr: 16
Title:

Spectral Analysis of Cardiogenic Vibrations to Distinguish Between Valvular Heart Diseases
Authors:

Ecem Erin and Beren Semiz
Abstract: Cardiovascular diseases are one of the top causes of mortality, accounting for a sizeable portion of all fatalities globally. Among cardiovascular diseases, valvular heart diseases (VHDs) affect greater number of people and have higher mortality rates. Current VHD assessment methods are cost-inefficient and limited to clinical settings, therefore there is a compelling need for non-invasive and continuous VHD monitoring systems. In this work, a novel framework was proposed to distinguish between aortic stenosis (AS), aortic valve regurgitation (AR), mitral valve stenosis (MS), and mitral valve regurgitation (MR) using tri-axial seismocardiogram (SCG) signals acquired from the mid-sternum. First, seismology domain knowledge was leveraged and applied to SCG signals through ObsPy toolbox for pre-processing. From pre-processed signal segments, spectrogram, wavelet, chromagram, tempogram and zero-crossing-rate features were extracted. Following p-value analysis and variance thresholding, a multi-label/multi-class classification framework based on gradient boosting trees was developed to distinguish between AS, AR, MS and MR cases. For all four VHDs, the accuracy, precision, recall and f1-score values were above 95%, best performing axis being the dorso-ventral direction. Overall, the results showed that spectral analysis of SCG signals can provide valuable information regarding VHDs and potentially be used in the design of continuous monitoring systems.
Download
Paper Nr: 19
Title:

Description of PD Phonation in Terms of EEG-Related Frequency Bands
Authors:

Pedro Gómez, Jiří Mekyska, Luboš Brabenec, Patrik Šimko, Irena Rektorová, Andrés Gómez and Victoria Rodellar
Abstract: Parkinson’s Disease (PD) is an increasing prevalence neurodegenerative condition affecting the life quality of people suffering from its neuromotor and cognitive performance. PD symptoms include vocalization and speech alterations, known as hypokinetic dysarthria (HD). One of the manifestations of HD is unstable phonation. Repetitive Transcranial Magnetic Stimulation (rTMS) is a non-invasive method that may improve some motor and non-motor symptoms of persons with PD (PwP). The present study concentrates on analyzing and comparing the phonation behavior of two cases before (pre-stimulus) and after (post-stimulus) ten sessions of rTMS treatment, to assess the extent of changes in their vocalization. Voice recordings of a sustained vowel [a:] taken immediately before and after the treatment, and at follow-up sessions (at six, ten, and fourteen weeks after the baseline assessment) were processed by inverse filtering to estimate a biomechanical correlate of vocal fold stiffness, which band-pass filtered into EEG-related frequency bands. Log-likelihood ratios between pre- and post-stimulus amplitude distributions of each frequency band, Mann-Whitney U-tests, and normalized difference scores showed significant improvements in the actively stimulated case, which were not observed in the sham case. Early preliminary insights into the capability of phonation quality assessment on monitoring neuromechanical activity from acoustic signals are shown.
Download
Paper Nr: 21
Title:

Deep Minutiae Fingerprint Extraction Using Equivariance Priors
Authors:

Margarida Gouveia, Eduardo Castro, Ana Rebelo, Jaime S. Cardoso and Bruno Patrão
Abstract: Currently, fingerprints are one of the most explored characteristics in biometric systems. These systems typically rely on minutiae extraction, a task highly dependent on image quality, orientation, and size of the fingerprint images. In this paper, a U-Net model capable of performing minutiae extraction is proposed (position, angle, and type). Based on this model, we explore two different ways of regularizing the model based on equivariance priors. First, we adapt the model architecture so that it becomes equivariant to rotations. Second, we use a multi-task learning approach in order to extract a more comprehensive set of information from the fingerprints (binary images, segmentation, frequencies, and orientation maps). The two approaches improved accuracy and generalization capability in comparison with the baseline model. On the 16 test datasets of the Fingerprint Verification Competition, we obtained an average Equal-Error Rate (EER) of 2.26, which was better than a well-optimized commercial product.
Download
Paper Nr: 23
Title:

Study of Uncertainty Quantification Using Multi-Label ECG in Deep Learning Models
Authors:

Raquel Simão, Marília Barandas, David Belo and Hugo Gamboa
Abstract: Machine Learning (ML) models can predict diseases with noteworthy results. However, when implemented, their generalization are compromised, resulting in lower performances and render healthcare professionals more susceptible into delivering erroneous diagnostics. This study focuses on the use of uncertainty measures to abstain from classifying samples and use the rejected samples as a selection criterion for active learning. For the multi-label classification of cardiac arrhythmias different methods for uncertainty quantification were compared using three Deep Learning (DL) models: a single model and two pseudoensemble models using Monte-Carlo (MC) Dropout and Deep Ensemble (DE) techniques. When tested with an external dataset, the models’ performances dropped from a F1-Score of 96% to 70%, indicating the possibility of dataset shift. The uncertainty measures for classification with rejection resulted in an increase of the rejection rate from 10% in the training set to a range between 30% to 50% on the external dataset. For the active learning approach, 10% of the highest uncertainty samples were used to retrain the models and their performance increased by almost 5%. Although there are still challenges to the implementation of ML models, the results show that uncertainty quantification is a valuable method to employ in safety mechanisms under dataset shift conditions.
Download
Paper Nr: 25
Title:

Multimodal Stress Classification Based on Biosignals Extracted from Smart Devices and Electromyography
Authors:

Maria Justino, Phillip Probst, Daniel Zagalo, Cátia Cepeda and Hugo Gamboa
Abstract: Work-Related Stress is the second most impactful occupational health problem in Europe, behind musculoskeletal diseases. When mental health is adequately handled, a worker’s well-being, performance, and productivity can be considerably improved. This paper presents machine learning models to classify mental stress experienced by office workers using physiological signals including heart rate, acquired using a smartwatch; respiration, derived from a smartphone’s acc placed on the chest; and trapezius electromyography, using proprietary electromyography sensors. Two interactive protocols were implemented to collect data from 12 individuals. Time features were extracted from heart rate and electromyography signals, with frequency features also being extracted from the latter. Statistical and temporal features were extracted from the derived respiration signal. Different input modalities were tested for the machine learning models: one for each physiological signal and a multimodal one, combining all of them. Three algorithms: Support Vector Machine, Random Forest, and K-Nearest-Neighbor were employed for mental stress classification. Random Forest obtained the best results (67.7%) for the heart rate model whereas K-Nearest-Neighbor attained higher accuracies for the respiration (89.1%) and electromyography (95.4%) models. Both algorithms achieved 100% accuracy for the multimodal model. A possible future approach would be to validate these models in real time.
Download
Paper Nr: 26
Title:

Analysis of Postural Variability of Office Workers Using Inertial Sensors
Authors:

Fábio Mendes, Phillip Probst, Eduarda Oliosi, Luís Silva, Cátia Cepeda and Hugo Gamboa
Abstract: Musculoskeletal disorders significantly impact workers in terms of quality of life, result in low organisational productivity, and high insurance costs in society. Postural changes have been suggested as a prerequisite to prevent musculoskeletal disorders. This paper examines the differences in postural changes of forty office workers in a real working environment using a smartphone’s inertial sensors. Through these data, several variables considered to characterise postural changes while sitting were extracted. Features based on the number of changes and different postures, time spent and distance covered within a posture showed significant differences in both time of the day (morning and afternoon) and day of the week (start and end of the week). These results confirm that accumulated working time influences a person’s postural changes and could have a potential use for worker’s ergonomic occupational risk evaluation.
Download
Paper Nr: 28
Title:

Automated Detection of Decision-Making Style, Based on Users’ Online Mouse Pointer Activity
Authors:

Marcus Cheetham, Catia Cepeda, Hugo Gamboa, Christoph Hoelscher and Seyed A. Valizadeh
Abstract: Decision-making (DM) and online activity go hand in hand in many domains of everyday life (e.g., consumer behaviour, financial and investment choices, career development, health and psychological well-being). DM style refers to consistent behavioural patterns in the way different individuals approach DM situations. In this study, we explored the feasibility of inferring DM style from the trace of mouse cursor (or pointer) activity that users generated while performing an online task (the task required no explicit DM). We focussed on maximizing and satisficing DM style. Based on a set of spatial, temporal and spatial-temporal features that were extracted from mouse activity data and on measures of DM style assessed with a conventional self-report questionnaire, we modelled DM style in a supervised machine learning approach. The results show that the models of DM style have between good and high predictive accuracy. Guided by these results, we propose that this mouse-based method might play a useful role in computational recognition of DM style and merits further development. Future work will test the ability of pointer-based models to meaningfully link psychological measures of DM style to objective measures and outcomes of real-world DM situations.
Download
Paper Nr: 29
Title:

Class-wise Knowledge Distillation for Lightweight Segmentation Model
Authors:

Ryota Ikedo, Kotaro Nagata and Kazuhiro Hotta
Abstract: In recent years, we have been improving the accuracy of semantic segmentation by deepening segmentation models, but large amount of computational resources are required due to the increase in computational complexity. Therefore knowledge distillation has been studied as one of model compression methods. We propose a knowledge distillation method in which the output distribution of a teacher model learned for each class is used as a target of the student model for the purpose of memory compression and accuracy improvement. Experimental results demonstrate that the segmentation accuracy was improved without increasing the computational cost on two different datasets.
Download
Paper Nr: 31
Title:

Fall Risk Assessment Using Wearable-Based Turn Detection: Comparison of Different Algorithms During Real-World Monitoring
Authors:

Jose Albites-Sanabria, Pierpaolo Palumbo, Stefania Bandinelli, Luca Palmerini and Lorenzo Chiari
Abstract: Turning deficits have been linked to aging and movement disorders and are a common cause of falls and fractures. Despite previous works on the automatic identification of turns and on its relation to fall risk, different algorithms for turn identification have been used, but their agreement and differences have not been investigated. In this study, we compared the two most-used turn-validated algorithms (El-Gohary and Pham) using a dataset comprising real-world data from 171 community-dwelling older adults monitored for one week with a single wearable sensor. The quantity and quality of turn parameters were calculated and used as predictors of future falls. After the analysis, the El-Gohary and Pham algorithms identified 1,063,810 and 942,845 turns, respectively. The agreement of the algorithms showed a very high to moderate correlation for all turn parameters. We found that prospective fallers take longer to perform a turn, and their movements are less smooth when compared to non-fallers. A fall risk assessment model built only on turn parameters showed reasonable performance for both algorithms (AUC = 0.6). Our results show that differences between turn parameters in the algorithms, when averaged at the single-subject level, are less of a concern when looking for associations with prospective falls.
Download
Paper Nr: 33
Title:

Content Based Image Retrieval Using Depth Maps for Colonoscopy Images
Authors:

Md M. Rahman, JungHwan Oh, Wallapak Tavanapong and Piet C. de Groen
Abstract: Content Based Image Retrieval (CBIR) finds similar images given a query image. Effective CBIR has been actively studied over several decades. For measuring image similarity, low-level visual features (i.e., color, shape, texture, and spatial layout), combination of low-level features, or Convolutional Neural Network (CNN) are typically used. However, a similarity measure based on these features is not effective for some type of images, for example, colonoscopy images captured from colonoscopy procedures. This is because the low-level visual features of these images are mostly very similar. We propose a new method to compare these images and find their similarity in terms of their surface topology. First, we generate a grey-scale depth map image for each image, then extract four straight lines from it. Each point in the four lines has a grey-scale value (depth) in its depth map. The similarity of the two images is measured by comparing the depth values on the four corresponding lines from the two images. We propose a function to compute a difference (distance) between two sets of four lines using Mean Absolute Error. The experiments based on synthetic and real colonoscopy images show that the proposed method is promising.
Download
Paper Nr: 34
Title:

Correlating EEG Signals and Electrode Locations by Means of Multidimensional Scaling
Authors:

Lucía Rodríguez-Giraldo and Juan P. Ugarte
Abstract: The adoption of physiological data, such as electroencephalograms (EEG), is undergoing a growing interest in addressing the characterization of human emotions. However, the setup of recording electrodes that allows a proper study of emotions remains to be determined. This work proposes a method for processing multichannel EEG signals by means of multidimensional scaling (MDS), looking for patterns related to the electrodes spatial setup. We analyze the SEED-IV database consisting of 1080 trials, each one having 62 simultaneous EEG recorded during four different emotions induction. A low dimensional representation of each set of 62 EEG signals is obtained through the MDS algorithm. The resulting MDS maps evinced a pattern of points that is correlated with the recording electrodes sites in 68% of the trials from SEED-IV database. Among these trials, those recorded during the neutral emotion induction are slightly prevalent than the remaining emotions. Furthermore, it was determined that the electrodes spatial distribution can be successfully recovered through the MDS analysis with an EEG minimum duration of 45 s. These results suggest that the proposed analysis based on the MDS algorithm shed some light upon the information content of simultaneous EEG signals and its correlation with the underlying cerebral structures.
Download
Paper Nr: 47
Title:

Reflecting on the Status Quo of Nonverbal Human-Machine Interaction
Authors:

Fábio Barros, António Teixeira and Samuel Silva
Abstract: Among humans, speech plays a central role in providing a form of communication that is efficient and natural. But communication goes beyond the verbal component harnessing a wide range of nonverbal aspects ranging from how words are said to posture, gestures, and facial movements. These, can complement or reinforce the message increasing the adaptiveness of our communication abilities to different contexts, a desirable characteristic to also have in our interaction with machines. Nevertheless, nonverbal communication cues are still scarcely considered for human-machine interaction not only motivated by the complexity of understanding and tackling them, but also by difficulties in translating them into a broader range of scenarios. In this context, this article examines the current state of research on nonverbal interaction and reflects on the challenges that need to be addressed in a multidisciplinary effort to advance the field.
Download
Paper Nr: 52
Title:

Parkinson and REM Sleep Behaviour Disorder: HRV Difference During Polysomnography
Authors:

Parisa Sattar, Giulia Baldazzi, Nicla Mandas, Elisa Casaglia, Michela Figorilli, Monica Puligheddu and Danilo Pani
Abstract: Approximately 40% to 70% of patients affected by Parkinson’s disease (PD) suffer from autonomic dysfunction that could be related to REM sleep behavior disorder (RBD). In this work, polysomnographic recordings were analyzed to study heart rate variability (HRV) during different sleep stages in a cohort of 20 participants, ten with Parkinson Disease with RBD (RBDpd) and ten unaffected (CG). HRV analysis was performed by considering the first 5 min epoch from each stage (i.e., wake, N2, N3, and REM), including time and frequency domain indexes, and entropy measures. Statistical analysis was carried out to assess any possible significant difference between CG and RBDpd groups, but also between the wake and REM stages in each group. Significant differences of the combined effect of RBD and PD emerged in both time and frequency domains, but also when considering nonlinear parameters during REM and awake phases. Accordingly, a comparison of wake and REM phase showed significant differences in all HRV parameters for CG that was absent in the RBDpd group. Our findings reveal the potentiality of HRV as a digital biomarker for RBDpd, by indicating distinct dysfunction of both parasympathetic and sympathetic activities in the RBDpd group, partially in line with previous studies.
Download
Paper Nr: 7
Title:

Comparison of Machine Learning Algorithms for Human Activity Recognition
Authors:

Hassan Ashraf, Olivier Brüls, Cédric Schwartz and Mohamed Boutaayamou
Abstract: Human activity recognition (HAR) is utilized to automatically identify the daily-life activities of people for the effective management of age-related health conditions. Classical machine learning (ML) algorithms are used to design HAR systems, in a subject-specific or population-based configuration depending on the application. In this study, the performance of 8 classical and ensemble-learning-based ML classifiers has been studied for both HAR configurations. Inertial measurement unit (IMU) signals from 10 healthy participants, corresponding to various static, dynamic, and transitional daily-life activities, were acquired. Random forest (RF), ensemble adaptive boosting (EAB), ensemble subspace (ES), decision tree (DT), k-nearest neighbors (KNN), linear discriminant analysis (LDA), support vector machine (SVM), and artificial neural network (ANN) were used to classify these activities. The performance of the classifiers was measured in terms of mean classification accuracy (MCA). The results showed that, for a subject-specific HAR system, ES (97.78%) has achieved the highest MCA followed by RF (96.61%) and SVM (96.11%) while outperforming the DT, KNN, and LDA (P-value < 0.05). For a population-based HAR system, SVM (95.18%) achieved the highest MCA, however, no significant difference has been observed among the MCA of all the investigated classifiers (P-value > 0.05). Also, the class-wise comparison reveals that SVM outperformed the other investigated classifiers in terms of MCAs for each of the distinct activities. Based on the HAR configuration incorporating diverse static, dynamic, and transitional daily-life activities, the findings may be used to develop a customized HAR system for the effective management of movement disorders.
Download
Paper Nr: 14
Title:

Depression in Obstructive Sleep Apnea Patients: Is Using Complex Deep Learning Structures Worth It?
Authors:

Mostafa Moussa, Yahya Alzaabi and Ahsan Khandoker
Abstract: The prevalence and severity of depression make it imperative to develop a means to automatically detect it, so as to alleviate the associated mental effort and cost of seeing a dedicated professional. Depression can also co-exist with other conditions, such as Obstructive Sleep Apnea Syndrome (OSAS). In this paper, we build upon our previous work involving sleep staging, detection of OSAS, and detection of depression in OSAS patients, but focus solely on the latter of the three. We use features extracted from EEG, ECG, and breathing signals of 80 subjects suffering from OSAS and half of which also with depression, using 75 % of this 80subject dataset for training and 10-fold cross-validation and the remainder for testing. We train three models to classify depression: a random forest (RF), a three-layer artificial neural network (3-ANN), and a gated-recurrent unit long short-term memory (GRU-LSTM) recurrent neural network. Our analysis shows that, like our previous work, the 3-ANN is still the best performing model, with the GRU-LSTM following closely behind at an accuracy of 79.0 % and 78.6 %, respectively, but with a smaller F1-score at 80.0 % and 81.6 %. However, we believe that the large increase in computation time and number of learnable parameters does not justify the use of GRU-LSTM over a simple ANN.
Download
Paper Nr: 18
Title:

Arrythmia Classification Using MATLAB® Classification Learner App
Authors:

Cesar N. Silva, Fernanda F. Lopes, Jefferson A. Matos and Maria F. Castro
Abstract: Vital sign monitoring is becoming a part of our daily lives, emerging as a trend of smart wearable devices used to manage health. Cardiac arrhythmia is any variation in the normal heartbeat rhythm, causing the heart to beat improperly. This work presents a study on the classification of cardiac arrhythmias in 4 classes, Normal (N), Supraventricular Ectopic (SVE), Ventricular Ectopic (LV), and Fusion of Normal and Ventricular (F). Using the MIT-BIH Arrhythmia Database and the Classification Learner App from MATLAB® for training, it was possible to investigate 24 models, where the Subspace KNN Ensemble obtained the best accuracy (74.4%) and was later used for implementation in the suggested user interface application.
Download
Paper Nr: 20
Title:

Hybrid Improved Physarum Learner for Structure Causal Discovery
Authors:

Joao P. Soares, Vitor Barth, Alan Eckeli and Carlos D. Maciel
Abstract: Causal discovery is the problem of estimating a joint distribution from observational data. In recent years, hybrid algorithms have been proposed to overcome computational problems that lead to better results. This work presents a hybrid approach that combines PC algorithm independence tests with a bio-inspired Improved Physarum Learner algorithm. The combination indicates improvement in computational time spent and yet consistent structural results.
Download
Paper Nr: 24
Title:

The Subspace Regularization Method Improves ErrP Detection by EEGNET in BCI Experiments
Authors:

Andrea Farabbi and Luca Mainardi
Abstract: In this study, the subspace regularization method was applied on the Electroencephalographic (EEG) signal recorded during stimulation of the Error Potential (ErrP) in order to improve the detection of the latter. The ErrP is stimulated through the presentation of an erroneous event to the subject. The recorded signals were processed with the subspace regularization method to remove the background EEG not related to the erroneous event. Then, the ErrP and Non–ErrP epochs (both raw and processed with the proposed method) were classified using EEGNET, a Convolutional Neural Network considered golden standard for EEG classification. The results show that elaborating the signals with the proposed method highlight the typical characteristics of the ErrP epochs both in temporal and frequency domain. Moreover, the classification metrics evaluated, always increase if compared to not processed signal (i.e. maximum increase in accuracy, balanced accuracy and F1-score are of 7.7%, 10.1% and 11% respectively). These findings suggest that the subspace regularization method can improve the performance of ErrP-based Brain Computer Interfaces (BCI) and can be used also in real time application and for asynchronous classification of erroneous events.
Download
Paper Nr: 37
Title:

Towards Lung Cancer Staging via Μultipositional Radiomics and Machine Learning
Authors:

Dimitris Fotopoulos, Dimitris Filos, Ekaterini Xinou and Ioanna Chouvarda
Abstract: This work addresses lung cancer diagnosis, and more specifically disease staging, as a major clinical challenge, crucial for further treatment decisions. The procedure is currently performed by experts based on clinical and medical imaging data and is time consuming and costly. Within INCISIVE, an EU-funded research project which aims to develop a pan-European federated image repository for cancer and implement Artificial Intelligence (AI) tools for clinical practice, clinical challenges have been identified that can be supported by AI in medical imaging data to facilitate accurate diagnosis and support treatment planning. The support and automation of lung cancer staging has been identified as a priority among the INCISIVE clinical challenges. In this scope, we propose a method to automatically classify between the group that represents disease stages I and II (low severity), vs the group that includes stages III and IV (severe). Tumour-Node-Metastasis system is used as a reference for staging. Based on lung CT image series with tumour and lung volume segmentation, we calculate and harmonise radiomics features and we propose the combination of tumour and lung lobes radiomics features towards improving the classification performance. Having a rich feature set as a basis, several combinations of feature selection and classification methods are tested and compared. Multiple repetitions of cross-validation and external testing splits are used to reach robust manner. The proposed method is trained and tested on an integrated dataset comprised of two open datasets (the NSCLC-Radiomics and the NSCLC-Radiogenomics dataset from the Cancer Imaging Archive). It achieves average Precision and Recall of 77.5% and 78.7% respectively, which could be further improved with a more extensive training set. Therefore, this can be the basis for a prioritisation tool regarding lung cancer cases and detailed staging/treatment decisions.
Download
Paper Nr: 43
Title:

An RFID Based Localization and Mental Stress Recognition System Using Wearable Sensors
Authors:

Mhd. W. Raed, Semih Yön, Ali Güneş, Igor Kotenko, Elena Fedorchenko and Anna Polubaryeva
Abstract: A vast increase in the percentage of elderly people over the past few decades has induced a serious concern among the research fraternity worldwide. Consequently, the large increase in the number of elderly needing assistance because of chronic diseases is expected to take place. Dementia, depression and mental stress are among the most disabling diseases with dangerous consequences such as wandering into hazardous or insecure areas. This wandering, particularly in urban areas can be life threatening. Recently, with the rapid emergence of disruptive technologies like Internet of Things (IoT), Radio Frequency Identification (RFID) and wireless bio sensors, it has become feasible to build systems that combine IoT and the cloud for monitoring the elderly suffering from dementia or depression. Furthermore, mental chronic diseases, such as stress and depression, are becoming a major concern for governments around the globe. The American Psychological Association (APA) categorizes stress, anxiety and depression as main factors for diverse mental health problems. The cost for treating work-related stress, anxiety and depression, is estimated to be around 617 billion euros per year in Europe alone. Wearable devices for monitoring chronic diseases such as mental stress and depression have been considered as game-changers to the way diseases are managed, by measuring vital signs like skin conductance and changes in the levels of biological stress, and sending warnings remotely to an online server. This paper proposes a work in progress Arduino based real-time stress recognition and localization system using wearable RFID and vital sign sensors for elderly suffering from Dementia and mental stress. The current work utilizes the heart rate variability and Electro Dermal Activity wearable sensors based on the Bitalino development system for measuring mental stress and anxiety in a smart home setting for elderly living alone by exposing a number of subjects to stress and anxiety stimulating horror videos. The system was tested successfully in the university lab.
Download
Paper Nr: 45
Title:

A Rank Aggregation Algorithm for Performance Evaluation in Modern Sports Medicine with NMR-based Metabolomics
Authors:

V. Vigneron and H. Maaref
Abstract: In most research studies, much of the gathered information is qualitative in nature. This article focuses on items for which there are multiple rankings that should be optimally combined. More specifically, it describes a supervised stochastic approach, driven by a Boltzmann machine capable of ranking elements related to each other by order of importance. Unlike classic statistical ranking techniques, the algorithm does not need a voting rule for decision-making. The experimental results indicate that the proposed model outperforms two reference rank aggregation algorithms, ELECTRE IV and VIKOR, and it behaves more stable when encountering noisy data.
Download
Paper Nr: 46
Title:

Classification of HCI Datasets Using Information Fusion and Weighted Frechet Distance
Authors:

Aleksandar Jeremic and Huaying Li
Abstract: The identification and classification of human breast cancer cells (MCF7) undergoing various treatments are widely used for studies of tumour biology and drug mechanism action. The development of adequate detection/classification strategies that would meet clinical needs is currently a subject of significant research interest as the optimal techniques are application/cell/treatment dependent. In addition to commonly used machine learning techniques for classification/clustering there has been an effort to utilize deep learning techniques as well. However, due to the fact that different cancer cells and different treatments require different data sets these techniques had rather limited success. In this paper we propose an information fusion technique that utilizes Frechet distance measures by combining their decisions in an optimal way by minimizing the overall classification error. The applicability of our results is demonstrated using real data sets with ten different treatments.
Download
Paper Nr: 48
Title:

Heterogeneous Inertial Measurement Units in Motion Analysis
Authors:

Filip Malawski and Natalia Paruzel-Satława
Abstract: Inertial measurement units are commonly used in motion analysis applications, such as sports training aid, gait analysis, medical diagnosis, or rehabilitation assistance. Linear acceleration and orientation obtained from sensor fusion are employed for the detection and classification of actions, as well as for measuring relevant parameters of the motion. Typically, in multi-sensor setups, a single model of the device is used. However, considering potential end-users, it could be beneficial to allow heterogeneous setups, particularly by including everyday-use devices with built-in inertial sensors, such as smartwatches. In this work, we perform experiments with several different sensors in order to analyze agreement in their measurements. Results indicate that devices of different models are not directly interchangeable, however, in some applications, heterogeneous setups may be viable.
Download
Paper Nr: 51
Title:

Redundancy and Novelty Between ECG Leads Based on Linear Correlation
Authors:

Utkars Jain, Adam A. Butchy, Michael T. Leasure, Veronica A. Covalesky, Daniel McCormick and Gary S. Mintz
Abstract: ECGs are a common diagnostic method for diagnosing cardiac pathologies. In this study, the Pearson correlation coefficient is used to examine the latent linear correlations between the leads of a standard 12-lead ECG. We utilize both the original ECG signals from the PTB-XL database and the reconstructed signal generated by a deep learning model, ECGio. We find that leads III, aVL, V1, and V2 are, on average, the leads with the most unique information due to their low correlation with other leads.
Download