INGENIOUS Data Fusion Module for monitoring and predicting health risks of FRs
EXUS AI Labs within the INGENIOUS project is working on implementing an alert system that will inform about the health status of first responders (FR). This system monitors the vitals of the FR and when a potential risk is identified (from their physiological data), alerts are raised to inform operators to take preventive action. Several approaches have been considered and at this stage machine learning algorithms are being utilised for the task.
Machine learning algorithms require large historical data in order to be able to be trained efficiently and classify new data accurately. Due to the lack of real first responder vitals data, in this first iteration of the INGENIOUS toolkit, an extensive research took place in order to find a dataset that could replicate the measurements of the INGENIOUS sensors.
The dataset selected contained non-EEG physiological signals and was collected at the Quality of Life Laboratory at the University of Texas at Dallas, used to infer the neurological status (including physical stress, cognitive stress, emotional stress, and relaxation) of 20 healthy subjects . The data was collected using non-invasive wrist-worn biosensors and consists of electrodermal activity (EDA), temperature, acceleration, heart rate (HR), and arterial oxygen level (SpO2).
Using this dataset, the period where the subjects were experiencing physical stress was labelled as an “alert zone”, while the relaxation period represents normal activities. This set allowed to train a deep learning algorithm to learn the patterns and the course of the vital signs of an exhausted subject and later, to be able to identify these patterns on the vital signs of First Responders. The algorithm achieved high accuracy (> 86%) within this set and provided a testbed for the INGENIOUS application.
The algorithm was later tested during a meeting with the INGENIOUS partners using simulated First Responder data and showed promising results as it achieved again high accuracy (>86%). EXUS AI LAB was able to demonstrate that models trained on different datasets can generalise well enough to allow us to apply on the field and provide an alternative when real life data are not available.
EXUS AI Labs, the R&D department of EXUS, focuses on designing and developing robust and trustworthy AI solutions that allow us to leverage the untapped potential of big data analytics across multiple verticals. EXUS AI Labs impacts the technological future by developing techniques to capture, process, analyse and visualise large datasets in timeframes not accessible to standard IT technologies.
 Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., … & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220