Deliverable 3.3 - Unobtrusive solution for the monitoring and the assessment of psychological status in adults with CHF

The following deliverable presents the multiple physiological data sources such as HR, HRV, GSR, BR, Speech, combined with psychological questionnaires collected by the HeartMan application during a semi-structured interview between patient and relative caregiver. All these features are sent through a cloud infrastructure to the psychological support module which processes the information with a machine learning model and recognizes the psychological health status of the patient (motivated, anxious, depressed). This module represents one of the main components of psychological DSS (deliverable D4.3). In this document, we also present preliminary results obtained with a mobile platform that captures physiological and vocal responses respectively extracted from the Zephyr chest belt and the smartphone. The challenge is to combine all these parameters for automatic detection of psychological comorbidities while providing a personalized intervention based on cognitive behavioural strategies and mindfulness exercises. Even if the sample is still limited and further optimizations will be obtained at the end of the trial, the achieved mean accuracy of models generalized with the leave one subject out approach is about 88.6%. This is an encouraging achievement, since it suggests that the explored features are good candidates for psychological assessment. During next months, we will integrate further parameters such as galvanic skin response, temperature and breath rate presented in section 2.3 and 2.4, that will be collected during the trial with the final release of HeartMan wristband aiming to increase the accuracy of psychological assessment.