Rossana Castaldo

Acute mental stress assessment via short term HRV analysis in healthy adults: A systematic review with meta-analysis

Mental stress reduces performances, on the work place and in daily life, and is one of the first causes of cognitive dysfunctions, cardiovascular disorders and depression. This study systematically reviewed existing literature investigating, in healthy subjects, the associations between acute mental stress and short term Heart Rate Variability (HRV) measures in time, frequency and non-linear domain. The goal of this study was to provide reliable information about the trends and the pivot values of HRV measures during mental stress. A systematic review and meta-analysis of the evidence was conducted, performing an exhaustive research of electronic repositories and linear researching references of papers responding to the inclusion criteria. After removing duplicates and not pertinent papers, journal papers describing well-designed studies that analyzed rigorously HRV were included if analyzed the same population of healthy subjects at rest and during mental stress. 12 papers were shortlisted, enrolling overall 758 volunteers and investigating 22 different HRV measures, 9 of which reported by at least 2 studies and therefore meta-analyzed in this review. Four measures in time and non-linear domains, associated with a normal degree of HRV variations resulted significantly depressed during stress. The power of HRV fluctuations at high frequencies was significantly depressed during stress, while the ratio between low and high frequency resulted significantly increased, suggesting a sympathetic activation and a parasympathetic withdrawal during acute mental stress. Finally, among the 15 non-linear measures extracted, only 2 were reported by at least 2 studies, therefore pooled, and only one resulted significantly depressed, suggesting a reduced chaotic behaviour during mental stress. HRV resulted significantly depressed during mental stress, showing a reduced variability and less chaotic behaviour. The pooled frequency domain measures demonstrated a significant autonomic balance shift during acute mental stress towards the sympathetic activation and the parasympathetic withdrawal. Pivot values for the pooled mean differences of HRV measures are provided. Further studies investigating HRV non-linear measures during mental stress are still required. However, the method proposed to transform and then meta-analyze the HRV measures can be applied to other fields where HRV proved to be clinically significant.

Wearable technology and ECG processing for fall risk assessment, prevention and detection

Falls represent one of the most common causes of injury-related morbidity and mortality in later life. Subjects with cardiovascular disorders (e.g., related to autonomic dysfunctions and postural hypotension) are at higher risk of falling. Autonomic dysfunctions increasing the risk of falling in the short and mid-term could be assessed by Heart Rate Variability (HRV) extracted by electrocardiograph (ECG). We developed three trials for assessing the usefulness of ECG monitoring using wearable devices for: risk assessment of falling in the next few weeks; prevention of imminent falls due to standing hypotension; and fall detection. Statistical and data-mining methods are adopted to develop classification and regression models, validated with the cross-validation approach. The first classifier based on HRV features enabled to identify future fallers among hypertensive patients with an accuracy of 72% (sensitivity: 51.1%, specificity: 80.2%). The regression model to predict falls due to orthostatic dropdown from HRV recorded before standing achieved an overall accuracy of 80% (sensitivity: 92%, specificity: 90%). Finally, the classifier to detect simulated falls using ECG achieved an accuracy of 77.3% (sensitivity: 81.8%, specificity: 72.7%). The evidence from these three studies showed that ECG monitoring and processing could achieve satisfactory performances compared to other system for risk assessment, fall prevention and detection. This is interesting as differently from other technologies actually employed to prevent falls, ECG is recommended for many other pathologies of later life and is more accepted by senior citizens.

Detection of mental stress due to oral academic examination via ultra-short-term HRV analysis

Mental stress may cause cognitive dysfunctions, cardiovascular disorders and depression. Mental stress detection via short-term Heart Rate Variability (HRV) analysis has been widely explored in the last years, while ultra-short term (less than 5 minutes) HRV has been not. This study aims to detect mental stress using linear and non-linear HRV features extracted from 3 minutes ECG excerpts recorded from 42 university students, during oral examination (stress) and at rest after a vacation. HRV features were then extracted and analyzed according to the literature using validated software tools. Statistical and data mining analysis were then performed on the extracted HRV features. The best performing machine learning method was the C4.5 tree algorithm, which discriminated between stress and rest with sensitivity, specificity and accuracy rate of 78%, 80% and 79% respectively.

Acute mental stress detection via ultra-short term HRV analysis

Acute mental stress reduces working performanc-es and is one of the first causes of cognitive dysfunctions, car-diovascular disorders and depression. Stress detection via short term (5 minutes) Heart Rate Variability (HRV) has been widely investigated in the last years. Recent improvements in wearable sensing devices and mobile computing raised a new research question: is ultra-short (2 minutes) HRV as effective as the short term one to detect mental stress? This study aimed to answer this research question. Short and ultra-short HRV was compared in 42 healthy subjects (age 25-38 years) under-taking the widely adopted and highly-effective the Stroop Color Word Test (CWT). ECG signals were recorded during rest and stress session using a chest wearable monitoring de-vice, the BioHarness M3 (ZephyrTech, NZ). HRV measures were then extracted and analyzed according to the literature and using validated software tools. Variations between short and ultra-short HRV measures in rest and stress sessions were analysed with the statistical Wilcoxon significance test. The results of the current study suggested that 6 HRV measures are effective in detecting acute mental stress both using short and ultra-short term analysis: Mean RR, Low Frequency power, Sample Entropy, Detrended fluctuation analysis: Short term and Long term fluctuation slope and Mean line length of Recurrence plot analysis.

To What Extent Can We Shorten HRV Analysis in Wearable Sensing? A Case Study on Mental Stress Detection.

Mental stress is one of the first causes of cognitive dysfunctions, cardiovascular disorders and depression. In addition, it reduces performances, on the work place and in daily life. The diffusion of wearablesensors (embedded in smart-watches, phones, etc.) has opened up the potential to assess mental stress detection through ultra-shortterm Heart Rate Variability (HRV) analysis (i.e., less than 5 min).Although informative analyses of features coming from short HRV (i.e., 5 min) have already been performed, the reliability of ultra-short HRVremains unclear. This study aims to tackle this gap by departing from a systematic review of the existing literature and investigating, in healthy subjects, the associations between acute mental stress and short/ultra-short term HRV features in time, frequency, and non-linear domains. Building on these findings, three experiments were carried outto empirically assess the usefulness of HRV for mental stress detection using ultra-short term analysis and wearable devices. Experiment 1 detected mental stress in a real life situation by exploring to which extent HRV excerpts can be shortened without losing their ability to detect mental stress. This allowed us to advance a method to explore to what extentultra-short HRV features can be consideredas good surrogates of 5 min HRV features. Experiment 2 and 3 sought todevelop automatic classifiers to detect mental stress through 2 min HRV excerpts, by usinga Stroop Color Word Test(CWT) and a highly pacedvideo game, which are two common laboratory-based stressors.

Heart Rate Variability Analysis and Performance during a Repeated Mental Workload Task

We designed and conducted an experiment using a repetitive task to investigate associations between mental workload, performance, and Heart Rate Variability (HRV) features across repetitions. According to the literature, we define mental workload as the interaction between a person and a task that causes task demands to exceed the person’s capacity to deliver. Mental workload was triggered by the use of a highly-paced video game repeated over time. Before engaging with the task, each subject was assessed in controlled condition (i.e., relaxing period) for a short time. Short term HRV features variations between the baseline (i.e., control situation) and each repetitive gaming session (i.e., mental task) were explored. The results show that HRV dynamics diminish with repetitions, while performance increases. Importantly, this suggests that HRV features can be well correlated with performance. Overall, this study advances the use of HRV analysis in the behavioral sciences at large, allowing the design of flexible neurophysiological lab-based experiments. Thus, it also opens the way to future autonomic behavioral neuroscience research.

Preliminary Results from a Proof of Concept Study for Fall Detection via ECG Morphology

Falls are a major problem in later life. Early fall detection systems are increased over the years as undetected falls can have severe consequences for the fallers. Fall detection systems are based mainly on posture detection using accelerometers and gyroscopes. Alternatively, this study aims to understand if it is possible to detect posture changes using only electrocardiogram (ECG) morphology, which is significantly associated with moving from one position to another. This paper presents preliminary results of a feasibility study aiming to investigate at what extend it was possible to detect lying and standing position, using wearable devices to observe ECG morphology alterations. According to the literature, 29 ECG features were extracted. 11 healthy subjects (aged 19-36 years) were monitoring while lying down and standing up. ECG and accelerometer signals were recorded continually using a chest wearable monitoring device, the BioHarness M3 (ZephyrTech, NZ).Variations in the ECG features while the subjects lay down and stood up were analysed with the parametric statistical paired T-test. The results of the current study suggested that 4 ECG features were effective in detecting changes while lying down or standing up. Linear Discriminant Analysis (LDA) was used to generate a classifier based on these ECG features to detect automatically the changes while lying down or standing up with total classification accuracy, sensitivity and specificity rates of 77.3%, 81.8%, and 72.7% respectively. The results obtained from the current study support a preliminary proof of concept and pave the way to more complex studies aiming to detecting real falls using ECG variations.

Selection of Entropy-Measure Parameters for Force Plate-Based Human Balance Evaluation

Human balance is commonly evaluated through the center of pressure (COP) displacement measured with a force plate, producing 2D time-series that represent COP trajectories in the anteroposterior and mediolateral directions. Entropy measures have been previously used to quantify the regularity of those time-series in different groups and/or experimental conditions. However, these measures are computed using multiple input parameters, the selection of which has been scarcely investigated within this context. This study aimed to investigate the behavior of COP time-series entropy measures using different parameters values, in order to inform their selection. Specifically, we investigated Approximate Entropy (ApEn) and Sample Entropy (SampEn), which are very sensitive to their input parameters: m (embedding dimension), r (tolerance) and N (length of data). A dataset containing COP time-series for 159 subjects with no physical disabilities was used. As a case study, subjects were grouped in young adults (age < 60, n = 85), and older adults (age ≥ 60) with (n = 18) and without (n = 56) history of falls. ApEn and SampEn were computed for m = {2, 3} and r = {0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5} with a fixed data length (N = 1200 points). ApEn and SampEn values were compared between groups using one-way ANOVA. Our results suggest that ApEn and SampEn are able to discriminate with ease between young and older adults for a wide range of m and r values. However, the selection becomes critical for the discrimination between non-fallers and fallers. An m = 2 and r = {0.4, 0.45} are suggested in this case.

Estimation of the Heart Rate Variability Features via Recurrent Neural Networks

Heart rate variability (HRV) analysis has increasingly become a promising marker for the assessment of the autonomic nervous system. The easy derivation of the HRV has determined its popularity, being successfully used in many research and clinical studies. However, the conventional HRV analysis is performed on 5 min ECG recordings which in e-health monitoring might be unsuitable, due to real-time requirements. Thus, the aim of this study is to evaluate the association between the raw ECG heartbeats and the HRV features to further reduce the number of heart beats required for the HRV estimation enabling real time monitoring. We propose a deep learning based system, specifically a recurrent neural network for the inference of two time domain HRV features: AVNN (the average of all the NN intervals) and IHR (instantaneous heart rate). The obtained results suggest that both AVNN and IHR can be accurately inferred from a shorter ECG interval of about 1 min, with a mean error of <5% of the computed HRV features.

Ultra-Short Entropy for Mental Stress Detection

Approximate Entropy (ApEn) and Sample Entropy (SampEn) are measures of signals’ complexity and are widely used in Heart Rate Variability (HRV) analysis. In particular, recent studies proved that almost all the features measuring complexity of RR series statistically decreased during the stress and therefore, thus showing ability to detect stress. However, the choice of the similarity threshold r and minimum data length N required for their computation are still controversial. In fact, most entropy measures are considered not reliable for recordings shorter than 5 min and different threshold values r have shown to affect the analysis thus leading to incorrect conclusions. Therefore, the aim of this study was to understand the impact of changing parameters r and N for the computation of ApEn and SampEn and to select the optimal parameters to detect stress in healthy subjects. To accomplish it, 84 RR series, extracted from electrocardiography signals acquired during real-life stress, were analyzed. ApEn and SampEn were estimated for two different values of r computed using previously published methods and for N = {100, 200, 300, 400, 500} data points. The statistical significance for the differences in mean ApEn and SampEn values was assessed by non-parametric tests. The two methods used to compute r produced entropy values significantly different over different N values. In contrast, ApEn and SampEn showed consistency in differentiating rest and stress conditions for different input parameters. More specifically, ApEnChon and SampEnChon showed to have a better discrimination power between stressed subjects and resting subjects on ultra-short recordings (N < 500).