Publications

Smoldering fires are characterized by the production of early gas emissions that can include high levels of CO and Volatile Organic Compounds (VOCs) due to pyrolysis or thermal degradation. Nowadays, standalone CO sensors, smoke detectors, or a combination of these, are standard components for fire alarm systems. While gas sensor arrays together with pattern recognition techniques are a valuable alternative for early fire detection, in practice they have certain drawbacks-they can detect early gas emissions, but can show low immunity to nuisances, and sensor time drift can render calibration models obsolete. In this work, we explore the performance of a gas sensor array for detecting smoldering and plastic fires while ensuring the rejection of a set of nuisances. We conducted variety of fire and nuisance experiments in a validated standard fire room (240 m(3)). Using PLS-DA and SVM, we evaluate the performance of different multivariate calibration models for this dataset. We show that calibration models remain predictive after several months, but perfect performance is not achieved. For example, 4 months after calibration, a PLS-DA model provides 100% specificity and 85% sensitivity since the system has difficulties in detecting plastic fires, whose signatures are close to nuisance scenarios. Nevertheless, our results show that systems based on gas sensor arrays are able to provide faster fire alarm response than conventional smoke-based fire alarms. We also propose the use of small-scale fire experiments to increase the number of calibration conditions at a reduced cost. Our results show that this is an effective way to increase the performance of the model, even when evaluated on a standard fire room. Finally, the acquired datasets are made publicly available to the community (doi: 10.5281/zenodo.5643074).

JTD Keywords: Calibration, Chemical sensors, Co2, Early fire, Early fire detection, En-54, Fire alarm, Fire detection, Fire room, Fires, Gas detectors, Gas emissions, Gas sensors, Pattern recognition, Public dataset, Sensor arrays, Sensors array, Signatures, Smoke, Smoke detector, Smoke detectors, Standard fire, Standard fire room, Support vector machines, Temperature, Toxicity, Volatile organic compounds

Cardiomyopathies diseases affects a great number of the elderly population. An adequate identification of the etiology of a cardiomyopathy patient is still a challenge. The aim of this study was to classify patients by their etiology in function of indexes extracted from the characterization of the pulse transit time (PTT). This time series represents the time taken by the pulse pressure to propagate through the length of the arterial tree and corresponding to the time between R peak of ECG and the mid-point of the diastolic to systolic slope in the blood pressure signal. For each patient, the PTT time series was extracted. Thirty cardiomyopathy patients (CMP) classified as ischemic (ICM - 15 patients) and dilated (DCM - 15 patients) were analyzed. Forty-three healthy subjects (CON) were used as a reference. The PTT time series was characterized through statistical descriptive indices and the joint symbolic dynamics method. The best indices were used to build support vector machine models. The optimal model to classify ICM versus DCM patients achieved 89.6% accuracy, 78.5% sensitivity, and 100% specificity. When comparing CMP patients and CON subjects, the best model achieved 91.3% accuracy, 91.3% sensitivity, and 88.3% specificity. Our results suggests a significantly lower pulse transit time in ischemic patients.Clinical relevance - This study analyzed the suitability of the pulse transit time for the classification of ICM and DCM patients. © 2021 IEEE.

JTD Keywords: Aged, Blood pressure, Cardiomyopathies, Cardiomyopathy, Cardiomyopathy, dilated, Congestive cardiomyopathy, Human, Humans, Pulse wave, Pulse wave analysis, Support vector machine

Heart diseases are the leading cause of death in developed countries. Ascertaining the etiology of cardiomyopathies is still a challenge. The objective of this study was to classify cardiomyopathy patients through cardio, respiratory and vascular variability analysis, considering the vascular activity as the input and output of the baroreflex response. Forty-one cardiomyopathy patients (CMP) classified as ischemic (ICM, 24 patients) and dilated (DCM, 17 patients) were analyzed. Thirty-nine elderly control subjects (CON) were used as reference. From the electrocardiographic, respiratory flow, and blood pressure signals, following temporal series were extracted: beat-to-beat intervals (BBI), total respiratory cycle time series (TT), and end– systolic (SBP) and diastolic (DBP) blood pressure amplitudes, respectively. Three-dimensional representation of the cardiorespiratory and vascular activities was characterized geometrically, by fitting a polygon that contains 95% of data, and by statistical descriptive indices. The best classifiers were used to build support vector machine models. The optimal model to classify ICM versus DCM patients achieved 92.7% accuracy, 94.1% sensitivity, and 91.7% specificity. When comparing CMP patients and CON subjects, the best model achieved 86.2% accuracy, 82.9% sensitivity, and 89.7% specificity. These results suggest a limited ability of cardiac and respiratory systems response to regulate the vascular variability in these patients.

JTD Keywords: Time series analysis, Support vector machines, Blood pressure, Sensitivity, Indexes, Electrocardiography, Kernel

Low-cost gas sensors allow for large-scale spatial monitoring of air quality in the environment. However they require calibration before deployment. Methods such as multivariate regression techniques have been applied towards sensor calibration. In this work, we propose instead, the use of deep learning methods, particularly, recurrent neural networks for predicting the gas concentrations based on the outputs of these sensors. This paper presents a first study of using Gated Recurrent Unit (GRU) neural network models for gas concentration prediction. The GRU networks achieve on average, a 44.69% and a 25.17% RMSE improvement in concentration prediction on a gas dataset when compared with Support Vector Regression (SVR) and Multilayer Perceptron (MLP) models respectively. With the current advances in deep network hardware accelerators, these networks can be combined with the sensors for a compact embedded system suitable for edge applications.

JTD Keywords: Robot sensing systems, Predictive models, Logic gates, Gas detectors, Training, Temperature measurement, Support vector machines

Cardiovascular diseases are one of the most common causes of death in elderly patients. The etiology of cardiomyopathies is difficult to discern clinically. The objective of this study was to classify cardiomyopathy patients using coupling analysis, through their cardiovascular behavior and the baroreflex response. A total of thirty-eight cardiomyopathy patients (CMP) classified as ischemic (ICM, 25 patients) and dilated (DCM, 13 patients) were analyzed. Thirty elderly control subjects (CON) were used as reference. Their electrocardiographic (ECG) and blood pressure (BP) signals were studied. To characterize the cardiovascular activity, the following temporal series were extracted: beat-to-beat intervals (from the ECG signal), and end- systolic and diastolic blood pressure amplitudes (from the BP signal). Non-linear characterization techniques like high resolution joint symbolic dynamics, segmented Poincaré plot analysis, normalized shorttime partial directed coherence, and dual sequence method were used to characterize these times series. The best indices were used to build support vector machine models for classification. The optimal model for ICM versus DCM patients achieved 84.2% accuracy, 76.9% sensitivity, and 88% specificity. When CMP patients and CON subjects were compared, the best model achieved 95.5% accuracy, 97.3% sensitivity, and 93.3% specificity. These results suggest a disfunction in the baroreflex mechanism in cardiomyopathies patients.

JTD Keywords: Couplings, Time series analysis, Support vector machines, Electrocardiography, Baroreflex, Coherence, Sensitivity

Patients with Chronic Heart Failure (CHF) often develop oscillatory breathing patterns. This work proposes the characterization of respiratory pattern by Wavelet Transform (WT) technique to identify Periodic Breathing pattern (PB) and Non-Periodic Breathing pattern (nPB) through the respiratory flow signal. A total of 62 subjects were analyzed: 27 CHF patients and 35 healthy subjects. Respiratory time series were extracted, and statistical methods were applied to obtain the most relevant information to classify patients. Support Vector Machine (SVM) were applied using forward selection technique to discriminate patients, considering four kernel functions. Differences between these parameters are assessed by investigating the following four classification issues: healthy subjects versus CHF patients, PB versus nPB patients, PB patients versus healthy subjects, and nPB patients versus healthy subjects. The results are presented in terms of average accuracy for each kernel function, and comparison groups.

JTD Keywords: Chronic heart failure, Forward selection, Non-periodic breathing, Periodic breathing, Support vector machine

Cardiac death risk is still a big problem by an important part of the population, especially in elderly patients. In this study, we propose to characterize and analyze the cardiovascular and cardiorespiratory systems using the Poincaré plot. A total of 46 cardiomyopathy patients and 36 healthy subjets were analyzed. Left ventricular ejection fraction (LVEF) was used to stratify patients with low risk (LR: LVEF > 35%, 16 patients), and high risk (HR: LVEF ≤ 35%, 30 patients) of heart attack. RR, SBP and TTot time series were extracted from the ECG, blood pressure and respiratory flow signals, respectively. Parameters that describe the scatterplott of Poincaré method, related to short- and long-term variabilities, acceleration and deceleration of the dynamic system, and the complex correlation index were extracted. The linear discriminant analysis (LDA) and the support vector machines (SVM) classification methods were used to analyze the results of the extracted parameters. The results showed that cardiac parameters were the best to discriminate between HR and LR groups, especially the complex correlation index (p = 0.009). Analising the interaction, the best result was obtained with the relation between the difference of the standard deviation of the cardiac and respiratory system (p = 0.003). When comparing HR vs LR groups, the best classification was obtained applying SVM method, using an ANOVA kernel, with an accuracy of 98.12%. An accuracy of 97.01% was obtained by comparing patients versus healthy, with a SVM classifier and Laplacian kernel. The morphology of Poincaré plot introduces parameters that allow the characterization of the cardiorespiratory system dynamics.

JTD Keywords: Time series analysis, Electrocardiography, Support vector machines, Kernel, Standards, Correlation, RF signals

In this work, the cardiorespiratory pattern of patients undergoing extubation process is studied. First, the respiratory and cardiac signals were resampled, next the Symbolic Dynamics (SD) technique was implemented, followed of a dimensionality reduction applying Forward Selection (FS) and Moving Window with Variance Analysis (MWVA) methods. Finally, the Linear Discriminant Analysis (LDA) and Support Vector Machines (SVM) classifiers were used. The study analyzed 153 patients undergoing weaning process, classified into 3 groups: Successful Group (SG: 94 patients), Failed Group (FG: 39 patients), and patients who had been successful during the extubation and had to be reintubated before 48 hours, Reintubated Group (RG: 21 patients). According to the results, the best classification present an accuracy higher than 88.98 ± 0.013% in all proposed combinations.

JTD Keywords: Support vector machines, Standards, Time series analysis, Resonant frequency, Nonlinear dynamical systems, Ventilation

Differentiating normal from adventitious respiratory sounds (RS) is a major challenge in the diagnosis of pulmonary diseases. Particularly, continuous adventitious sounds (CAS) are of clinical interest because they reflect the severity of certain diseases. This study presents a new classifier that automatically distinguishes normal sounds from CAS. It is based on the multi-scale analysis of instantaneous frequency (IF) and envelope (IE) calculated after ensemble empirical mode decomposition (EEMD). These techniques have two major advantages over previous techniques: high temporal resolution is achieved by calculating IF-IE and a priori knowledge of signal characteristics is not required for EEMD. The classifier is based on the fact that the IF dispersion of RS signals markedly decreases when CAS appear in respiratory cycles. Therefore, CAS were detected by using a moving window to calculate the dispersion of IF sequences. The study dataset contained 1494 RS segments extracted from 870 inspiratory cycles recorded from 30 patients with asthma. All cycles and their RS segments were previously classified as containing normal sounds or CAS by a highly experienced physician to obtain a gold standard classification. A support vector machine classifier was trained and tested using an iterative procedure in which the dataset was randomly divided into training (65%) and testing (35%) sets inside a loop. The SVM classifier was also tested on 4592 simulated CAS cycles. High total accuracy was obtained with both recorded (94.6% ± 0.3%) and simulated (92.8% ± 3.6%) signals. We conclude that the proposed method is promising for RS analysis and classification.

JTD Keywords: Diseases, Dispersion, Empirical mode decomposition, Feature extraction, Informatics, Support vector machines

Aging population is a major concern that is reflected in the increase of chronic diseases. Heart Failure (HF) is one of the most common chronic diseases of elderly people that is punctuated with acute episodes, which result in hospitalization. The periodic modulation of the amplitude of the breathing pattern is proved to be one of the multiple symptoms of an acute episode, and thus, the features extracted from its characterization contribute in the improvement of the first diagnosis of the clinical practice. The main objective of this study is to evaluate if the features extracted from the breathing pattern along with common clinical variables are reliable enough to detect Periodic Breathing (PB). A dataset of 44 elderly patients containing clinical information and a short record of electrocardiogram and respiratory flow signal was used to train two machine learning classification methods: Support Vector Machine (SVM) and Linear Discriminant Analysis (LDA). All the available clinical parameters within the dataset along with the parameters characterizing the respiratory pattern were used to classify the observations into two groups. SVM classification was optimized and performed using a = -8 and C = 10.04 giving an accuracy of 88.2 % sensitivity of 90 % and specificity of 85.7 % Similar results were achieved with LDA classifying with an accuracy of 82.4 %, a sensitivity of 81.8% and specificity of 83.3 % PB has been accurately detected using both classifiers.

JTD Keywords: Support vector machines, Feature extraction, Training, Senior citizens, Standards, Training data, Hospitals

Cardiovascular diseases are the first cause of death in developed countries. Using electrocardiographic (ECG), blood pressure (BP) and respiratory flow signals, we obtained parameters for classifying cardiomyophaty patients. 42 patients with ischemic (ICM) and dilated (DCM) cardiomyophaties were studied. The left ventricular ejection fraction (LVEF) was used to stratify patients with low risk (LR: LVEF>35%, 14 patients) and high risk (HR: LVEF≤ 35%, 28 patients) of heart attack. RR, SBP and T_Tot time series were extracted from the ECG, BP and respiratory flow signals, respectively. The time series were transformed to a binary space and then analyzed using Joint Symbolic Dynamic with a word length of three, characterizing them by the probability of occurrence of the words. Extracted parameters were then reduced using correlation and statistical analysis. Principal component analysis and support vector machines methods were applied to characterize the cardiorespiratory and cardiovascular interactions in ICM and DCM cardiomyopaties, obtaining an accuracy of 85.7%.

JTD Keywords: Blood pressure, Electrocardiography, Joints, Kernel, Principal component analysis, Support vector machines, Time series analysis

Disconnection from mechanical ventilation, called the weaning process, is an additional difficulty in the management of patients in intensive care units (ICU). Unnecessary delays in the discontinuation process and a weaning trial that is undertaken too early are undesirable. In this study, we propose an extubation index based on the power of the respiratory flow signal (Pi). A total of 132 patients on weaning trials were studied: 94 patients with successful trials (group S) and 38 patients who failed to maintain spontaneous breathing and were reconnected (group F). The respiratory flow signals were processed considering the following three stages: a) zero crossing detection of the inspiratory phase, b) inflection point detection of the flow curve during the inspiratory phase, and c) calculation of the signal power on the time instant indicated by the inflection point. The zero crossing detection was performed using an algorithm based on thresholds. The inflection points were marked considering the zero crossing of the second derivative. Finally, the inspiratory power was calculated from the energy contained over the finite time interval (between the instant of zero crossing and the inflection point). The performance of this parameter was evaluated using the following classifiers: logistic regression, linear discriminant analysis, the classification and regression tree, Naive Bayes, and the support vector machine. The best results were obtained using the Bayesian classifier, which had an accuracy, sensitivity and specificity of 87%, 90% and 81% respectively.

JTD Keywords: Bayes methods, Bayesian classifier, Indexes, Logistics, Niobium, Regression tree analysis, Support vector machines, Ventilation

Classification algorithms with unbalanced datasets tend to produce high predictive accuracy over the majority class, but poor predictive accuracy over the minority class. This problem is very common in biomedical data mining. This paper introduces a Support Vector Machine (SVM)-based optimized feature selection method, to select the most relevant features and maintain an accurate and well-balanced sensitivity–specificity result between unbalanced groups. A new metric called the balance index (B) is defined to implement this optimization. The balance index measures the difference between the misclassified data within each class. The proposed optimized feature selection is applied to the classification of patients' weaning trials from mechanical ventilation: patients with successful trials who were able to maintain spontaneous breathing after 48 h and patients who failed to maintain spontaneous breathing and were reconnected to mechanical ventilation after 30 min. Patients are characterized through cardiac and respiratory signals, applying joint symbolic dynamic (JSD) analysis to cardiac interbeat and breath durations. First, the most suitable parameters (C+,C−,σ) are selected to define the appropriate SVM. Then, the feature selection process is carried out with this SVM, to maintain B lower than 40%. The best result is obtained using 6 features with an accuracy of 80%, a B of 18.64%, a sensitivity of 74.36% and a specificity of 82.42%.

JTD Keywords: Support vector machines, Balance index, Sensitivity-specificity balance, Cardiorespiratory interaction, Joint symbolic dynamics, Feature selection, Weaning procedure

One of the most challenging problems in intensive care is still the process of discontinuing mechanical ventilation, called weaning process. Both an unnecessary delay in the discontinuation process and a weaning trial that is undertaken too early are undesirable. In this study, we analyzed respiratory pattern variability using the respiratory volume signal of patients submitted to two different levels of pressure support ventilation (PSV), prior to withdrawal of the mechanical ventilation. In order to characterize the respiratory pattern, we analyzed the following time series: inspiratory time, expiratory time, breath duration, tidal volume, fractional inspiratory time, mean inspiratory flow and rapid shallow breathing. Several autoregressive modeling techniques were considered: autoregressive models (AR), autoregressive moving average models (ARMA), and autoregressive models with exogenous input (ARX). The following classification methods were used: logistic regression (LR), linear discriminant analysis (LDA) and support vector machines (SVM). 20 patients on weaning trials from mechanical ventilation were analyzed. The patients, submitted to two different levels of PSV, were classified as low PSV and high PSV. The variability of the respiratory patterns of these patients were analyzed. The most relevant parameters were extracted using the classifiers methods. The best results were obtained with the interquartile range and the final prediction errors of AR, ARMA and ARX models. An accuracy of 95% (93% sensitivity and 90% specificity) was obtained when the interquartile range of the expiratory time and the breath duration time series were used a LDA model. All classifiers showed a good compromise between sensitivity and specificity.

JTD Keywords: autoregressive moving average processes, feature extraction, medical signal processing, patient care, pneumodynamics, signal classification, support vector machines, time series, ARX, autoregressive modeling techniques, autoregressive models with exogenous input, autoregressive moving average model, breath duration time series, classification method, classifier method, discontinuing mechanical ventilation, expiratory time, feature extraction, final prediction errors, fractional inspiratory time, intensive care, interquartile range, linear discriminant analysis, logistic regression analysis, mean inspiratory flow, patient respiratory volume signal, pressure support level, pressure support ventilation, rapid shallow breathing, respiratory pattern variability characterization, support vector machines, tidal volume, weaning trial, Analytical models, Autoregressive processes, Biological system modeling, Estimation, Support vector machines, Time series analysis, Ventilation

Weaning trials process of patients in intensive care units is a complex clinical procedure. 153 patients under extubation process (T-tube test) were studied: 94 patients with successful trials (group S), 38 patients who failed to maintain spontaneous breathing and were reconnected (group F), and 21 patients with successful test but that had to be reintubated before 48 hours (group R). The respiratory pattern of each patient was characterized through the following time series: inspiratory time (T_I), expiratory time (T_E), breathing cycle duration (T_Tot), tidal volume (V_T), inspiratory fraction (T_I/T_Tot), half inspired flow (V_T/T_I), and rapid shallow index (f/V_T), where f is respiratory rate. Using techniques as autoregressive models (AR), autoregressive moving average models (ARMA) and autoregressive models with exogenous input (ARX), the most relevant parameters of the respiratory pattern were obtained. We proposed the evaluation of these parameters using classifiers as logistic regression (LR), linear discriminant analysis (LDA), support vector machines (SVM) and classification and regression tree (CART) to discriminate between patients from groups S, F and R. An accuracy of 93% (98% sensitivity and 82% specificity) has been obtained using CART classification.

JTD Keywords: Accuracy, Indexes, Logistics, Regression tree analysis, Support vector machines, Time series analysis, Autoregressive moving average processes, Medical signal processing, Pattern classification, Pneumodynamics, Regression analysis, Sensitivity, Signal classification, Support vector machines, Time series, SVM, T-tube testing, Autoregressive models-with-exogenous input, Autoregressive moving average models, Breathing cycle duration, Classification-and-regression tree, Expiratory time, Extubation process, Half inspired flow, Inspiratory fraction, Inspiratory time, Intensive care units, Linear discriminant analysis, Logistic regression, Rapid shallow index, Respiratory pattern parameter performance, Sensitivity, Spontaneous breathing, Support vector machines, Tidal volume, Time 48 hr, Time series, Weaning process classifiers

The process of discontinuing mechanical ventilation is called weaning and is one of the most challenging problems in intensive care. An unnecessary delay in the discontinuation process and an early weaning trial are undesirable. This study aims to characterize the respiratory pattern through features that permit the identification of patients' conditions in weaning trials. Three groups of patients have been considered: 94 patients with successful weaning trials, who could maintain spontaneous breathing after 48 h ( GSucc ); 39 patients who failed the weaning trial ( GFail ) and 21 patients who had successful weaning trials, but required reintubation in less than 48 h ( GRein ). Patients are characterized by their cardiorespiratory interactions, which are described by joint symbolic dynamics (JSD) applied to the cardiac interbeat and breath durations. The most discriminating features in the classification of the different groups of patients ( GSucc , GFail and GRein ) are identified by support vector machines (SVMs). The SVM-based feature selection algorithm has an accuracy of 81% in classifying GSucc versus the rest of the patients, 83% in classifying GRein versus GSucc patients and 81% in classifying GRein versus the rest of the patients. Moreover, a good balance between sensitivity and specificity is achieved in all classifications.

JTD Keywords: Mechanical ventilation, Weaning, Support vector machines, Joint symbolic dynamics

In this paper, we propose a biologically inspired architecture for a Hopfield-like associative memory applied to artificial olfaction. The proposed algorithm captures the projection between two neural layers of the insect olfactory system (Antennal Lobe and Mushroom Body) with a kernel based projection. We have tested its classification performance as a function of the size of the training set and the time elapsed since training and compared it with that obtained with a Support Vector Machine.

JTD Keywords: Biocomputing, Chemioception, Content-addressable storage, Hopfield neural nets, Support vector machines

The Ischemic Heart Disease (IHD) is characterized by an insufficient supply with blood of the myocardium usually caused by an artherosclerotic disease of the coronary arteries (coronary artery disease CAD). The IHD and its consequences have become a leading problem in the industrialized nations. The aim of this study was to evaluate a new diagnosing method, the cardiogoniometry, using two different classifying techniques: the method of linear discriminant function analysis (LDA) and the method of Support Vector Machines (SVM). Data of a group of 109 female subjects (62 healthy, 47 with IHD) were analyzed on the basis of extracted parameters from the three-dimensional vector loops of the heart. The LDA achieved an accuracy of 83,5% (Sensitivity 78,7%, Specificity 87,1%), whereas the SVM achieved an accuracy of 86% (Sensitivity 80,5%, Specificity 89,8%). It could be shown that cardiogoniometry, an electrophysiological diagnostic method performed at rest, detects variables that are helpful in identifying ischemic heart disease. As it is easy to apply, non-invasive, and provides an automated interpretation it may become an inexpensive addition to the cardiologic diagnostic armamentarium, possibly useful for early diagnosis of IHD or CAD, as well as in patients who do not tolerate exercise testing. It was also proven that by applying Support Vector Machines an increased diagnostic precision in comparison to the conventional discriminant function analysis can be achieved.

JTD Keywords: Cardiogoniometry, Support Vector Machines, Nonlinear classifier, Linear discriminant analysis, Vector loop