Papers by Rui Pedro Paiva
2015 Computing in Cardiology Conference (CinC), 2015
Neurally mediated syncope (NMS) is a transient and self-limited loss of consciousness that affect... more Neurally mediated syncope (NMS) is a transient and self-limited loss of consciousness that affects all ages and is associated with high rates of falls and hospitalizations. In this study we propose a new algorithm for real-time prediction of NMS that integrates indexes of autonomic modulation among other parameters, which is based on the analysis of the electrocardiogram (ECG) and photoplethysmogram (PPG) alone. ECG and PPG signals were acquired from 43 patients with suspected NMS, during scheduled diagnostic headup tilt table (HUTT) tests. Heart rate variability (HRV) indexes were integrated in a NMS prediction algorithm comprising surrogates of chronotropic, inotropic, blood pressure and vascular tone changes. The proposed algorithm was validated using a threeway data split approach. HRV indexes improved the algorithm performance in both the train/validation phase and the test phase, showing the importance of autonomic modulation indexes in real-time prediction of NMS.
Physiological Measurement, 2015
Monitoring of cardiovascular function on a beat-to-beat basis is fundamental for protecting patie... more Monitoring of cardiovascular function on a beat-to-beat basis is fundamental for protecting patients in different settings including emergency medicine and interventional cardiology, but still faces technical challenges and several limitations. In the present study, we propose a new method for the extraction of cardiovascular performance surrogates from analysis of the photoplethysmographic (PPG) signal alone. We propose using a multi-Gaussian (MG) model consisting of five Gaussian functions to decompose the PPG pulses into its main physiological components. From the analysis of these components, we aim to extract estimators of the left ventricular ejection time, blood pressure and vascular tone changes. Using a multi-derivative analysis of the components related
Traditional Artificial Neural Networks (ANN) have been investigated in the past for skin lesion c... more Traditional Artificial Neural Networks (ANN) have been investigated in the past for skin lesion classification and nowadays their performance is already quite useful to assist in medical diagnosis and decision processes. In the field of visual object recognition, recent developments of such networks (Deep Convolutional Neural Networks) are currently the winners of the ImageNet competition. This work extends the use of CNN for classification of pigmented skin lesions, by investigating a training methodology based on transfer learning on pre-trained networks.
A Prototype for Classification of Classical Music using Neural Networks
International Conference on Biomedical and Health Informatics, 2018
During the acquisition of lung sounds, several sources of noise can interfere with the recordings... more During the acquisition of lung sounds, several sources of noise can interfere with the recordings. Therefore, the detection of noise present in lung sounds plays an important role in the correct diagnosis of several pulmonary disorders such as in chronic obstructive pulmonary diseases. Denoising tools reported so far focus mainly in the detection of abnormal lung sounds from the background noise (usually vesicular background) or even just in the discrimination of normal from abnormal lung sounds. Algorithms for heart sound cancellation have also been proposed. However, it can be noticed that there is a lack of signal processing methods to efficiently detected and/or remove artifacts introduced in the acquisition environment or produced by the subject (e.g., speech). The present study focuses in the analysis of lungs sounds recorded in two different populations containing events of cough, speech and other artifacts from the surrounding environment. Feature extraction and binary classification were performed achieving, on average, values of a sensitivity and specificity ranging from 76% to 97% for the classification of cough, speech and other artifacts and from 83% to 90% for the specific detection of cough events. The detection of artifacts achieved sensitivity and specificity values of 84% and 61%, respectively for one population and 88% and 52% for another population.
2017 39th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2017
We present a new method for the discrimination of explosive cough events, which is based on a com... more We present a new method for the discrimination of explosive cough events, which is based on a combination of spectral content descriptors and pitch-related features. After the removal of near-silent segments, a vector of event boundaries is obtained and a proposed set of 9 features is extracted for each event. Two data sets, recorded using electronic stethoscopes and comprising a total of 46 healthy subjects and 13 patients, were employed to evaluate the method. The proposed feature set is compared to three other sets of descriptors: a baseline, a combination of both sets, and an automatic selection of the best 10 features from both sets. The combined feature set yields good results on the cross-validated database, attaining a sensitivity of 92.32.3% and a specificity of 84.73.3%. Besides, this feature set seems to generalize well when it is trained on a small data set of patients, with a variety of respiratory and cardiovascular diseases, and tested on a bigger data set of mostly healthy subjects: a sensitivity of 93.4% and a specificity of 83.4% are achieved in those conditions. These results demonstrate that complementing the proposed feature set with a baseline set is a promising approach.
Precision Medicine Powered by pHealth and Connected Health, 2017
The automatic analysis of respiratory sounds has been a field of great research interest during t... more The automatic analysis of respiratory sounds has been a field of great research interest during the last decades. Automated classification of respiratory sounds has the potential to detect abnormalities in the early stages of a respiratory dysfunction and thus enhance the effectiveness of decision making. However, the existence of a publically available large database, in which new algorithms can be implemented, evaluated, and compared, is still lacking and is vital for further developments in the field. In the context of the International Conference on Biomedical and Health Informatics (ICBHI), the first scientific challenge was organized with the main goal of developing algorithms able to characterize respiratory sound recordings derived from clinical and non-clinical environments. The database was created by two research teams in Portugal and in Greece, and it includes 920 recordings acquired from 126 subjects. A total of 6898 respiration cycles were recorded. The cycles were annotated by respiratory experts as including crackles, wheezes, a combination of them, or no adventitious respiratory sounds. The recordings were collected using heterogeneous equipment and their duration ranged from 10s to 90s. The chest locations from which the recordings were acquired was also provided. Noise levels in some respiration cycles were high, which simulated real life conditions and made the classification process more challenging.
2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2016
The global inhomogeneity (GI) index is a electrical impedance tomography (EIT) parameter that qua... more The global inhomogeneity (GI) index is a electrical impedance tomography (EIT) parameter that quantifies the tidal volume distribution within the lung. In this work the global inhomogeneity index was computed for twenty subjects in order to evaluate his potential use in the detection and follow up of chronic obstructive pulmonary disease (COPD) patients. EIT data of 17 subjects were acquired: 14 patients with the main diagnoses of COPD and 3 healthy subjects which served as a control group. Two or three datasets of around 30 seconds were acquired at 33 scans/s and analysed for each subject. After reconstruction, a tidal EIT image was computed for each breathing cycle and a GI index calculated from it. Results have shown significant differences in GI values between the two groups (0.745 ± 0.007 for COPD and 0.668 ± 0.006 for lung-healthy subject, p < 0.005). The GI values obtained for each subject have shown small variance between them, which is a good indication of stability. The results suggested that the GI may be useful for the identification and follow up of ventilation problems in patients with COPD.
2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2016
The automatic detection of adventitious lung sounds is a valuable tool to monitor respiratory dis... more The automatic detection of adventitious lung sounds is a valuable tool to monitor respiratory diseases like chronic obstructive pulmonary disease. Crackles are adventitious and explosive respiratory sounds that are usually associated with the inflammation or infection of the small bronchi, bronchioles and alveoli. In this study a multi-feature approach is proposed for the detection of events, in the frame space, that contain one or more crackles. The performance of thirty-five features was tested. These features include thirty-one features usually used in the context of Music Information Retrieval, a wavelet based feature as well as the Teager energy and the entropy. The classification was done using a logistic regression classifier. Data from seventeen patients with manifestations of adventitious sounds and three healthy volunteers were used to evaluate the performance of the proposed method. The dataset includes crackles, wheezes and normal lung sounds. The optimal detection parameters, such as the number of features, were chosen based on a grid search. The performance of the detection was studied taking into account the sensitivity and the positive predictive value. For the conditions tested, the best results were obtained for the frame size equal to 128 ms and twenty-seven features.
Heart murmur classification with feature selection
2010 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10, 2010
Noise detection during heart sound recording using periodicity signatures
Physiological Measurement, 2011
Identification and analysis of stable breathing periods in electrical impedance tomography recordings
Physiological Measurement, 2021
Objective. In this paper, an automated stable tidal breathing period (STBP) identification method... more Objective. In this paper, an automated stable tidal breathing period (STBP) identification method based on processing electrical impedance tomography (EIT) waveforms is proposed and the possibility of detecting and identifying such periods using EIT waveforms is analyzed. In wearable chest EIT, patients breathe spontaneously, and therefore, their breathing pattern might not be stable. Since most of the EIT feature extraction methods are applied to STBPs, this renders their automatic identification of central importance. Approach. The EIT frame sequence is reconstructed from the raw EIT recordings and the raw global impedance waveform (GIW) is computed. Next, the respiratory component of the raw GIW is extracted and processed for the automatic respiratory cycle (breath) extraction and their subsequent grouping into STBPs. Main results. We suggest three criteria for the identification of STBPs, namely, the coefficient of variation of (i) breath tidal volume, (ii) breath duration and (iii) end-expiratory impedance. The total number of true STBPs identified by the proposed method was 294 out of 318 identified by the expert corresponding to accuracy over 90%. Specific activities such as speaking, eating and arm elevation are identified as sources of false positives and their discrimination is discussed. Significance. Simple and computationally efficient STBP detection and identification is a highly desirable component in the EIT processing pipeline. Our study implies that it is feasible, however, the determination of its limits is necessary in order to consider the implementation of more advanced and computationally demanding approaches such as deep learning and fusion with data from other wearable sensors such as accelerometers and microphones.
2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2021
Patients suffering from pulmonary diseases typically exhibit pathological lung ventilation in ter... more Patients suffering from pulmonary diseases typically exhibit pathological lung ventilation in terms of homogeneity. Electrical Impedance Tomography (EIT) is a noninvasive imaging method that allows to analyze and quantify the distribution of ventilation in the lungs. In this article, we present a new approach to promote the use of EIT data and the implementation of new clinical applications for differential diagnosis, with the development of several machine learning models to discriminate between EIT data from healthy and nonhealthy subjects. EIT data from 16 subjects were acquired: 5 healthy and 11 non-healthy subjects (with multiple pulmonary conditions). Preliminary results have shown accuracy percentages of 66% in challenging evaluation scenarios. The results suggest that the pairing of EIT feature engineering methods with machine learning methods could be further explored and applied in the diagnostic and monitoring of patients suffering from lung diseases. Also, we introduce the use of a new feature in the context of EIT data analysis (Impedance Curve Correlation).
Presently, the demands for good paper quality are growing higher and higher. Since one important ... more Presently, the demands for good paper quality are growing higher and higher. Since one important variable to assess paper quality is paper brightness, pulp bleaching is a most important concern. Therefore, it is extremely important to have a thorough understanding of the bleaching plant, in order to achieve those high standards. In this paper a neuro-fuzzy approach is proposed for modelling of the pulp bleaching plant at Companhia de Celulose do Caima, S.A. (Portugal). This strategy is conducted in two phase: in the first one, subtractive clustering is applied in order to extract a set of fuzzy rules; then, in the second stage, the centres and widths of the membership functions are tuned by means of a fuzzy neural network trained with backpropagation. This technique seems promising since it permits good results with large nonlinear plants. Furthermore, it describes the plant using a set of linguistic rules, which have the advantage of being closer to natural human language, so, more...
International Conference on Biomedical and Health Informatics, 2018
Atrial Fibrillation (AF) is the most common arrhythmia and is associated with an increased risk o... more Atrial Fibrillation (AF) is the most common arrhythmia and is associated with an increased risk of heart-related deaths and the development of conditions such as heart failure, dementia, and stroke. Affecting mostly elderly people, AF is associated with high comorbidity, increased mortality and is a major socio-economic impact in our society. Therefore, the detection of AF episodes in personalized health (p-Health) environments can be decisive in the prevention of major cardiac threats and in the reduction of health care costs. In this paper we present a new algorithm for detection of AF based on the assessment of the three main physiological characteristics of AF: (1) the irregularity of the heart rate; (2) the absence of the P-wave and (3) the presence of fibrillatory waves. Several features were extracted from the analysis of 12-lead electrocardiogram (ECG) signals, the best features were selected and a support vector machine classification model was adopted to discriminate AF an...
This research addresses the role of audio and lyri cs in the music emotion recognition. Each dime... more This research addresses the role of audio and lyri cs in the music emotion recognition. Each dimension (e.g., audio) was separately studied, as well as in a context of bimodal analysis. We perform classi fication by quadrant categories (4 classes). Our approach is based on several audio and lyrics state-of-the-art features, as well as novel lyric features. To evalu ate our approach we create a ground-truth dataset. The main conclusions show tha t unlike most of the similar works, lyrics performed better than audio. This sug gests the importance of the new proposed lyric features and that bimodal analys is is always better than each
Precision Medicine Powered by pHealth and Connected Health, 2017
We present a multi-feature approach to the detection of cough and adventitious respiratory sounds... more We present a multi-feature approach to the detection of cough and adventitious respiratory sounds. After the removal of near-silent segments, a vector of event boundaries is obtained and a proposed set of 126 features is extracted for each event. Evaluation was performed on a data set comprised of internal audio recordings from 18 patients. The best performance (F-measure = 0.69 0.03; specificity = 0.90 0.01) was achieved when merging wheezes and crackles into a single class of adventitious respiratory sounds.
Skin Lesion Classification using Bag-of-3D-Features
2021 Telecoms Conference (ConfTELE), 2021
Computer-aided diagnostic has become a thriving research area in recent years, namely on the iden... more Computer-aided diagnostic has become a thriving research area in recent years, namely on the identification of skin lesions such as melanoma. This work presents a novel approach to this field by exploiting the 3D characteristics of the skin lesion surface, advancing beyond common features such as, shape, colour, and texture, extracted from dermoscopic RGB images. To this end, a relevant set of features was investigated to obtain 3D skin lesion characteristics from images with depth information. These features were used to train a Bag-of-Features model to distinguish between malignant and benign lesions, also discriminating melanoma from all other lesion types. Despite the large class imbalance, often present in medical image datasets, the feature set achieved a top accuracy of 73.08%, comprising 75.00% sensitivity and 66.67% specificity when classifying between malignant and benign lesions, and 88.46% accuracy (100.00% sensitivity and 86.96% specificity) when discriminating melanoma from all other lesion images, using only depth information. The achieved experimental results indicate the existence of relevant discriminative characteristics in the 3D surface of skin lesions which allow the improvement of existing classification methods based on 2D image characteristics only.
2015 3rd Experiment International Conference (exp.at'15), 2015
We present a Matlab framework for heart sound processing and analysis. This framework includes al... more We present a Matlab framework for heart sound processing and analysis. This framework includes algorithms developed for segmentation of the main heart sound components capable of handling situations with high-grade murmur, and for measuring systolic time intervals (STI). Methods for cardiac function parameter extraction based on STI are also included. Currently, the proposed algorithms are being extended for multichannel applications. The algorithms outlined in the paper have been extensively evaluated using data collected from patients with several types of cardiovascular diseases under real-life conditions.
Physiological Measurement, 2016
Electrical impedance tomography (EIT) is increasingly used in patients suffering from respiratory... more Electrical impedance tomography (EIT) is increasingly used in patients suffering from respiratory disorders during pulmonary function testing (PFT). The EIT chest examinations often take place simultaneously to conventional PFT during which the patients involuntarily move in order to facilitate their breathing. Since the influence of torso and arm movements on EIT chest examinations is unknown, we studied this effect in 13 healthy subjects (37 ± 4 years, mean age ± SD) and 15 patients with obstructive lung diseases (72 ± 8 years) during stable tidal breathing. We carried out the examinations in an upright sitting position with both arms adducted, in a leaning forward position and in an upright sitting position with consecutive right and left arm elevations. We analysed the differences in EIT-derived regional end-expiratory impedance values, tidal impedance variations and their spatial distributions during all successive study phases. Both the torso and the arm movements had a highly significant influence on the end-expiratory impedance values in the healthy subjects (p = 0.0054 and p < 0.0001, respectively) and the patients (p < 0.0001 in both cases). The global tidal impedance variation was affected by the torso, but not the arm movements in both study groups (p = 0.0447 and p = 0.0418, respectively). The spatial heterogeneity of the tidal ventilation distribution was slightly influenced by the alteration of the
Uploads
Papers by Rui Pedro Paiva