Early detection of exposure to a toxic chemical, e.g., in a military context, can be life-saving. We propose to use machine learning techniques and multiple continuously measured physiological signals to detect exposure, and to identify the chemical agent. Such detection and identification could be used to alert individuals to take appropriate medical counter measures in time. As a first step, we evaluated whether exposure to an opioid (fentanyl) or a nerve agent (VX) could be detected in freely moving guinea pigs using features from respiration, electrocardiography (ECG) and electroencephalography (EEG), where machine learning models were trained and tested on different sets (across subject classification). Results showed this to be possible with close to perfect accuracy, where respiratory features were most relevant. Exposure detection accuracy rose steeply to over 95% correct during the first five minutes after exposure. Additional models were trained to correctly classify an exposed state as being induced either by fentanyl or VX. This was possible with an accuracy of almost 95%, where EEG features proved to be most relevant. Exposure detection models that were trained on subsets of animals generalized to subsets of animals that were exposed to other dosages of different chemicals. While future work is required to validate the principle in other species and to assess the robustness of the approach under different, realistic circumstances, our results indicate that utilizing different continuously measured physiological signals for early detection and identification of toxic agents is promising.

No abstract

Wearable sensors offer new opportunities for the early detection and identification of toxic chemicals in situations where medical evaluation is not immediately possible. We previously found that continuously recorded physiology in guinea pigs can be used for early detection of exposure to an opioid (fentanyl) or a nerve agent (VX), as well as for differentiating between the two. Here, we investigated how exposure to these different chemicals affects the interactions between ECG and respiration parameters as determined by Granger causality (GC). Features reflecting such interactions may provide additional information and improve models differentiating between chemical agents. Traditional respiration and ECG features, as well as GC features, were extracted from data of 120 guinea pigs exposed to VX (n = 61) or fentanyl (n = 59). Data were divided in a training set (n = 99) and a test set (n = 21). Minimum Redundancy Maximum Relevance (mRMR) and Support Vector Machine (SVM) algorithms were used to, respectively, perform feature selection and train a model to discriminate between the two chemicals. We found that ECG and respiration parameters are Granger-related under healthy conditions, and that exposure to fentanyl and VX affected these relationships in different ways. SVM models discriminated between chemicals with accuracy of 95% or higher on the test set. GC features did not improve the classification compared to traditional features. Respiration features (i.e., peak inspiratory and expiratory flow) were the most important to discriminate between different chemical’s exposure. Our results indicate that it may be feasible to discriminate between chemical exposure when using traditional physiological respiration features from wearable sensors. Future research will examine whether GC features can contribute to robust detection and differentiation between chemicals when considering other factors, such as generalizing results across species.

Feedback of physiological responses have a great potential to support virtual training paradigms aimed to increase cognitive task performance under stressful threatening conditions. In the current study, we examined the sensitivity of a range of physiological indicators derived from electrodermal activity (EDA), blood pressure (BP) and heart rate (HR) to measure stress as induced by the threat of an electric shock (ES). In contrast to previous work that studied physiological stress responses compared to a rest condition, we compared conditions with high cognitive load combined with stress caused by threat of an ES, to conditions with high cognitive load without such stress. Twenty-five participants performed a cognitively demanding task in an experimental setup. At certain 10 s time intervals, indicated by a continuous tone, participants were either asked to do their best and increase cognitive task performance (non-threat condition), or they were told that they could receive an ES during this interval if cognitive task performance was not high enough (threat condition). Physiological measures, task performance and selfreported measures of stress and workload were analysed. Task performance and selfreported measures of stress and workload were roughly the same in both conditions. Especially EDA measures were affected by the threat of an ES. Threat and non-threat conditions could be distinguished with an across-participant classifier using EDA and BP features with an accuracy of 70%. These results suggest that EDA and BP can be used to evaluate stress coping training paradigms or to individually adapt the stress levels in virtual training environments.

Maintaining mental health can be quite challenging, especially when exposed to stressful situations. In many cases, mental health problems are recognized too late to effectively intervene and prevent adverse outcomes. Recent advances in the availability and reliability of wearable technologies offer opportunities for continuously monitoring mental states, which may be used to improve a person’s mental health. Previous studies attempting to detect and predict mental states with different modalities have shown only small to moderate effect sizes. This limited success may be due to the large variability between individuals regarding e.g., ways of coping with stress or behavioral patterns associated with positive or negative feelings. A study was set up for the detection of mental states based on longitudinal wearable and contextual sensing, targeted at investigating between-subjects variations in terms of predictors of mental states and variations in how predictors relate to mental states. At the end of March 2022, 16 PhD candidates from the Netherlands started to participate in the study. Over nine months, we collected data in terms of their daily mental states (valence and arousal), continuous physiological data (Oura ring) and smartphone data (AWARE framework including GPS and smartphone usage). From the raw data, we aggregated daily values for each participant in terms of sleep, physical activity, mental states, phone usage and GPS movement. First results (six months into the study at the time of writing) indicate that almost all participants show a large variability in ratings of daily mental states, which is a prerequisite for predictive modeling. Direction, strength and standard deviations of Spearman correlations between valence, arousal and the different variables suggest that several predictors of valence and arousal are more subject dependent than others. In future analyses, we will test and compare different versions of predictive modeling to highlight the potential of wearable technologies for mental state monitoring and the personalized prediction of the development of mental problems.