Simão Paredes

Prediction of acute hypotensive episodes by means of neural network multi-models

This work proposes the application of neural network multi-models to the prediction of adverse acute hypotensive episodes (AHE) occurring in intensive care units (ICU). A generic methodology consisting of two phases is considered. In the first phase, a correlation analysis between the current blood pressure time signal and a collection of historical blood pressure templates is carried out. From this procedure the most similar signals are determined and the respective prediction neural models, previously trained, selected. Then, in a second phase, the multi-model structure is employed to predict the future evolution of current blood pressure signal, enabling to detect the occurrence of an AHE. The effectiveness of the methodology was validated in the context of the 10th PhysioNet/Computers in Cardiology Challenge-Predicting Acute Hypotensive Episodes, applied to a specific set of blood pressure signals, available in MIMIC-II database. A correct prediction of 10 out of 10 AHE for event 1 and of 37 out of 40 AHE for event 2 was achieved, corresponding to the best results of all entries in the two events of the challenge. The generalization capabilities of the strategy was confirmed by applying it to an extended dataset of blood pressure signals, also collected from the MIMIC-II database. A total of 2344 examples, selected from 311 blood pressure signals were tested, enabling to obtain a global sensitivity of 82.8% and a global specificity of 78.4%.

A lead dependent ischemic episodes detection strategy using Hermite functions

Abstract In this work a new strategy for automatic detection of ischemic episodes is proposed. A new measure for ST deviation based on the time–frequency analysis of the ECG and the use of a reduced set of Hermite basis functions for T wave and QRS complex morphology characterization, are the key points of the proposed methodology. Usually, ischemia manifests itself in the ECG signal by ST segment deviation or by QRS complex and T wave changes in morphology. These effects might occur simultaneously. Time–frequency methods are especially adequate for the detection of small transient characteristics hidden in the ECG, such as ST segment alterations. A Wigner–Ville transform-based approach is proposed to estimate the ST shift. To characterize the alterations in the T wave and the QRS morphologies, each cardiac beat is described by expansions in Hermite functions. These demonstrated to be suitable to capture the most relevant morphologic characteristics of the signal. A lead dependent neural network classifier considers, as inputs, the ST segment deviation and the Hermite expansion coefficients. The ability of the proposed method in ischemia episodes detection is evaluated using the European Society of Cardiology ST–T database. A sensitivity of 96.7% and a positive predictivity of 96.2% reveal the capacity of the proposed strategy to perform ischemic episodes identification.

Wavelet based time series forecast with application to acute hypotensive episodes prediction

This paper presents a generic methodology for time series prediction, based on a wavelet decomposition/ reconstruction technique, together with a feedforward neural networks structure. The proposed methodology combines the flexibility and learning abilities of neural networks with a compact description of the signals, inherent to wavelets. In a first phase a wavelet decomposition of the signal is performed, providing a small number of coefficients that summarizes signal time evolution dynamics. The prediction problem is then effectively addressed by means of a neural networks model, previously trained using coefficients of the training dataset. The particular problem of forecasting acute hypotensive episodes (AHE) occurring in intensive care units was used to prove the effectiveness of the proposed strategy. The dataset, extracted from MIMIC-II, was made available in the context of the PhysioNet-Computers in Cardiology Challenge 2009. Results attained in this work were similar to the best ones achieved under that challenge.

Atrial Activity Detection through a Sparse Decomposition Technique

Atrial fibrillation is the most common cardiac arrhythmia, presenting significant consequences on patient health. Automatic detection of atrial fibrillation needs, ideally, the isolated study of the atrial activity registered in the electrocardiogram. Sparse decomposition techniques make possible the decomposition of a signal into their components, thus the separation between atrial and ventricular activities. However, this technique requires the a priori construction of distinct dictionaries, usually built based on atrial and ventricular activity simulation models. This work addresses the construction of the dictionaries based on real electrocardiogram signals, where P-waves, QRS-complexes and T-waves are first identified to support the creation of the dictionaries. The effectiveness of the proposed methodology is validated with real signals, obtained from MIT-BIH arrhythmia database.

An effective wavelet strategy for the trend prediction of physiological time series with application to pHealth systems

This work proposes a wavelet decomposition based scheme to estimate the evolution trend of physiological time series. The scheme does not involve the explicit development of a model and is essentially supported on the hypothesis that future evolution of a biosignal can be estimated from similar historic patterns. The strategy considers an a-trous wavelet decomposition, where the most representative trends are extracted from the historic similar patterns. Then, a set of distance-based measures able to assess the prediction likelihood of each representative trend, is introduced. From these measures and through an optimization process, a subset of these trends is selected and aggregated to derive the required time series evolution trend. The effectiveness of the methodology is validated in the prediction of blood pressure signals collected in two telemonitoring studies: TEN-HMS and MyHeart. Additionally, Friedman and Nemenyi statistics tests are implemented to rank several methods, confirming the value of the proposed strategy.

Phase space reconstruction approach for ventricular arrhythmias characterization

Ventricular arrhythmias, especially tachycardia and fibrillation are one of the main causes of sudden cardiac death. Therefore, the development of methodologies, enable to detect their occurrence and to characterize their time evolution, is of fundamental importance. This work proposes a non-linear dynamic signal processing approach to address the problem. Based on the phase space reconstruction of the electrocardiogram (ECG), some features are extracted for each ECG time window. Features from current and previous time windows are provided to a dynamic neural network classifier, enabling arrhythmias detection and evolution trends assessment. Sensitivity and specificity values, evaluated from public MIT-BIH databases, show the effectiveness of the proposed strategy.

Prediction of Heart Failure Decompensation Events by Trend Analysis of Telemonitoring Data

This paper aims to assess the predictive value of physiological data daily collected in a telemonitoring study in the early detection of heart failure (HF) decompensation events. The main hypothesis is that physiological time series with similar progression (trends) may have prognostic value in future clinical states (decompensation or normal condition). The strategy is composed of two main steps: a trend similarity analysis and a predictive procedure. The similarity scheme combines the Haar wavelet decomposition, in which signals are represented as linear combinations of a set of orthogonal bases, with the Karhunen-Loève transform, that allows the selection of the reduced set of bases that capture the fundamental behavior of the time series. The prediction process assumes that future evolution of current condition can be inferred from the progression of past physiological time series. Therefore, founded on the trend similarity measure, a set of time series presenting a progression similar to the current condition is identified in the historical dataset, which is then employed, through a nearest neighbor approach, in the current prediction. The strategy is evaluated using physiological data resulting from the myHeart telemonitoring study, namely blood pressure, respiration rate, heart rate, and body weight collected from 41 patients (15 decompensation events and 26 normal conditions). The obtained results suggest, in general, that the physiological data have predictive value, and in particular, that the proposed scheme is particularly appropriate to address the early detection of HF decompensation.

An efficient strategy for evaluating similarity between time series based on Wavelet / Karhunen-Loève transforms

The present work aims to present an innovative measure able to efficiently evaluate the similarity between two physiological time series. The proposed methodology combines the Haar wavelet decomposition, in which signals are represented as linear combinations of a set of orthogonal basis, with the Karhunen-Loève transform, that allows for the optimal reduction of that set of basis. The similarity measure is based on the Euclidean distance, but indirectly calculated through the linear combination coefficients of both time series. Moreover, an iterative scheme for computing the referred coefficients significantly decreases the computational complexity of the method that, due to its simplicity and fast execution, can be easily applicable in clinical applications, namely in computational demanding contexts such as telemonitoring environments. This strategy has been successfully implemented and validated inside HeartCycle project, applied to blood pressure signals collected by a telemonitoring platform (TEN-HMS) in the recognition of hypertension episodes.

Assessment of cardiovascular risk based on a data-driven knowledge discovery approach.

The cardioRisk project addresses the development of personalized risk assessment tools for patients who have been admitted to the hospital with acute myocardial infarction. Although there are models available that assess the short-term risk of death/new events for such patients, these models were established in circumstances that do not take into account the present clinical interventions and, in some cases, the risk factors used by such models are not easily available in clinical practice. The integration of the existing risk tools (applied in the clinicians daily practice) with data-driven knowledge discovery mechanisms based on data routinely collected during hospitalizations, will be a breakthrough in overcoming some of these difficulties. In this context, the development of simple and interpretable models (based on recent datasets), unquestionably will facilitate and will introduce confidence in this integration process. In this work, a simple and interpretable model based on a real dataset is proposed. It consists of a decision tree model structure that uses a reduced set of six binary risk factors. The validation is performed using a recent dataset provided by the Portuguese Society of Cardiology (11113 patients), which originally comprised 77 risk factors. A sensitivity, specificity and accuracy of, respectively, 80.42%, 77.25% and 78.80% were achieved showing the effectiveness of the approach.

Improvement of CVD risk assessment tools' performance through innovative patients' grouping strategies

There are available in the clinical community several practical risk tools to assess the risk of occurrence of a cardiovascular event. Although valuable, these tools typically present some lack of performance (low sensitivity/low specificity) when applied to a general (average) patient. This flaw is addressed in this work through an innovative personalization strategy that is supported on the evidence that current risk assessment tools perform differently among different populations/groups of patients. The proposed methodology is based on two main hypotheses: i) patients are grouped through a proper dimension reduction technique complemented with an unsupervised learning algorithm, ii) for each group the most suitable risk assessment tool can be selected improving the risk prediction performance. As a result, risk personalization is simply achieved by the identification of the group that patients belong to. The validation of the strategy is carried out through the combination of three current risk assessment tools (GRACE, TIMI, PURSUIT) developed to predict the risk of an event in coronary artery disease patients. The combination of these tools is validated with a real patient testing dataset: Santa Cruz Hospital, Portugal, N=460 ACS-NSTEMI1 patients. Considering the obtained results with the available dataset it is possible to state that the main objective of this work was achieved.