Sibylle Fallet

A multimodal approach to reduce false arrhythmia alarms in the intensive care unit

As part of the 2015 PhysioNet/CinC Challenge, this work aims at lowering the number of false alarms, which are a persistent concern in the intensive care unit. The multimodal database consists of 1250 life-threatening alarm recordings, each categorized as a bradycardia, tachycardia, asystole, ventricular tachycardia or ventricular flutter/fibrillation arrhythmia. Based on the quality of available signals, heart rate was either estimated from pulsatile waveforms (photoplethysmogram and/or arterial blood pressure) using an adaptive frequency tracking algorithm or computed from ECGs using an adaptive mathematical morphology approach. Furthermore, we introduced a supplementary measure based on the spectral purity of the ECGs to determine if a ventricular tachycardia or flutter/fibrillation arrhythmia has taken place. Finally, alarm veracity was determined based on a set of decision rules on heart rate and spectral purity values. Our method achieved overall scores of 76.11 and 85.04 on the real-time and retrospective subsets, respectively.

False arrhythmia alarms reduction in the intensive care unit: a multimodal approach.

The purpose of this study was to develop algorithms to lower the incidence of false arrhythmia alarms in the ICU using information from independent sources, namely electrocardiogram (ECG), arterial blood pressure (ABP) and photoplethysmogram (PPG). Our approach relies on robust adaptive signal processing techniques in order to extract accurate heart rate (HR) values from the different waveforms. Based on the quality of available signals, heart rate was either estimated from pulsatile waveforms using an adaptive frequency tracking algorithm or computed from ECGs using an adaptive mathematical morphology approach. Furthermore, we developed a supplementary measure based on the spectral purity of the ECGs to determine whether a ventricular tachycardia or flutter/fibrillation arrhythmia has taken place. Finally, alarm veracity was determined based on a set of decision rules on HR and spectral purity values. The proposed method was evaluated on the PhysioNet/CinC Challenge 2015 database, which is composed of 1250 life-threatening alarm recordings, each categorized into either bradycardia, tachycardia, asystole, ventricular tachycardia or ventricular flutter/fibrillation arrhythmia. This resulted in overall true positive rates of 95%/99% and overall true negative rates of 76%/80% on the real-time and retrospective subsets of the test dataset, respectively.

Adaptive frequency tracking for robust heart rate estimation using wrist-type photoplethysmographic signals during physical exercise

In recent years, wearable photoplethysmographic (PPG) biosensors have emerged as promising tools to monitor heart rate (HR) during physical exercise. However, PPG waveforms are easily corrupted by motion artifacts, rendering HR estimation difficult. In this study, HR was estimated using wrist-type PPG signals. A normalized least-mean-squares (NLMS) algorithm was first used to attenuate motion artifacts and reconstruct multiple PPG waveforms from different combinations of corrupted PPG waveforms and accelerometer (ACC) data. An adaptive band-pass filter was then used to track the common instantaneous frequency component (i.e. HR) of the reconstructed PPG waveforms. Our proposed HR estimation method, which is almost real time, resulted in an average absolute error of 1.71 ± 0.49 beats-per-minute and a Pearson correlation coefficient of 0.994 between the true and the estimated HR values. Importantly, as all ACC-PPG combinations were used for motion artifacts cancellation, no assumption about individual ACC axis contribution was required.

Sleep Apnea Detection using Features from the Respiration and the ECG recorded with Smart-Shirts

The automatic detection of sleep apnea episodes, without the need of polysomnography and outside a clinical facility, could help facilitate the diagnosis of this disorder. In this work, features to detect sleep apnea events were computed from respiration and electrocardiogram recordings acquired with a wearable smart-shirt. First, a classical scheme exploiting the amplitude decrease of the respiration during apnea episodes was presented. Second, a novel measure of the phase coupling between the respiration and the respiratory sinus arrhythmia from the ECG was introduced. It was shown that these features were significantly different during sleep apnea episodes than for normal breathing.

Robust heart rate estimation using wrist-type photoplethysmographic signals during physical exercise: an approach based on adaptive filtering.

Photoplethysmographic (PPG) signals are easily corrupted by motion artifacts when the subjects perform physical exercise. This paper introduces a two-step processing scheme to estimate heart rate (HR) from wrist-type PPG signals strongly corrupted by motion artifacts. Adaptive noise cancellation, using normalized least-mean-square algorithm, is first performed to attenuate motion artifacts and reconstruct multiple PPG waveforms from different combinations of corrupted PPG waveforms and accelerometer data. An adaptive band-pass filter is then used to track the common instantaneous frequency component (i.e. HR) of the reconstructed PPG waveforms. The proposed HR estimation scheme was evaluated on two datasets, composed of records from running subjects and subjects performing different kinds of arm/forearm movements and resulted in average absolute errors of 1.40  ±  0.60 and 4.28  ±  3.16 beats-per-minute for these two datasets, respectively. Importantly, the proposed method is fully automatic, induces an average estimation delay of 0.93 s, and is therefore suitable for real-time monitoring applications.

Can one detect atrial fibrillation using a wrist-type photoplethysmographic device?

AbstractThis study aims at evaluating the potential of a wrist-type photoplethysmographic (PPG) device to discriminate between atrial fibrillation (AF) and other types of rhythm. Data from 17 patients undergoing catheter ablation of various arrhythmias were processed. ECGs were used as ground truth and annotated for the following types of rhythm: sinus rhythm (SR), AF, and ventricular arrhythmias (VA). A total of 381/1370/415 10-s epochs were obtained for the three categories, respectively. After pre-processing and removal of segments corresponding to motion artifacts, two different types of feature were derived from the PPG signals: the interbeat interval-based features and the wave-based features, consisting of complexity/organization measures that were computed either from the PPG waveform itself or from its power spectral density. Decision trees were used to assess the discriminative capacity of the proposed features. Three classification schemes were investigated: AF against SR, AF against VA, and AF against (SR&VA). The best results were achieved by combining all features. Accuracies of 98.1/95.9/95.0 %, specificities of 92.4/88.7/92.8 %, and sensitivities of 99.7/98.1/96.2 % were obtained for the three aforementioned classification schemes, respectively. Graphical AbstractAtrial fibrillation detection using PPG signals

A Novel Short-Term Event Extraction Algorithm for Biomedical Signals

In this paper, we propose a fast novel nonlinear filtering method named Relative-Energy (Rel-En), for robust short-term event extraction from biomedical signals. We developed an algorithm that extracts short- and long-term energies in a signal and provides a coefficient vector with which the signal is multiplied, heightening events of interest. This algorithm is thoroughly assessed on benchmark datasets in three different biomedical applications, namely ECG QRS-complex detection, EEG K-complex detection, and imaging photoplethysmography (iPPG) peak detection. Rel-En successfully identified the events in these settings. Compared to the state-of-the-art, better or comparable results were obtained on QRS-complex and K-complex detection. For iPPG peak detection, the proposed method was used as a preprocessing step to a fixed threshold algorithm that lead to a significant improvement in overall results. While easily defined and computed, Rel-En robustly extracted short-term events of interest. The proposed algorithm can be implemented by two filters and its parameters can be selected easily and intuitively. Furthermore, Rel-En algorithm can be used in other biomedical signal processing applications where a need of short-term event extraction is present.