The PredicAlert project aims to develop a new methodology to predict the state of alertness from cardiorespiratory variables usable in environments where EEG and polysomnography are not feasible. As part of this project, the scope of application is the MRI and in particular the rest functional MRI studies. The main objectives of this project are the filing of one or more patents and the improvement of the diagnostic potential of rest functional imaging. As such, it fully fulfils the socio-economic objective of this call for projects.In our project, the evaluation of vigilance in an MRI environment is mandatory. The degree of awakening on the sleep-watch axis is mainly measured using EEG-derived techniques, such as polysomnography (Oken et al., 2006), which are based on records of low electrical potentials. They are highly disturbed in the electromagnetically MRI environment. Currently, there is no system to easily and accurately monitor vigilance in MRI studies (Wang, Ong, Patanaik, Zhou, & Chee, 2016). Work on the level of vigilance and the evaluation of physiological parameters involved alternative devices to polysomnography, including cardiac or respiratory signals (R. Bartsch, Kantelhardt, Penzel, & Havlin, 2007; R. P. Bartsch, Schumann, Kantelhardt, Penzel, & Ivanov, 2012; Muzet et al., 2016; Viola et al., 2003). Regardless of the device used, all these studies showed that the different time and frequency parameters varied according to the drowsiness profile. Thus cardiac variability increases during the light slow sleep transition, decreases in deep slow sleep and then increases again in paradoxical sleep. Other studies have also found strong synchronisation between cardiac and respiratory cycles during the slow (light and deep) stages of sleep, when parasympathic tone predominates (R. Bartsch et al., 2007; Ehrhart et al., 2000). During paradoxical sleep, when sympathetic tone predominates, the authors found a decrease in the synchronisation of heart and respiratory cycles. It is therefore possible to predict the level of alertness of a subject elongated in an MRI from cardiac and respiratory signals that are easily recorded in these contions. Despite the need, there is currently no validated device for monitoring vigilance in rest MRI studies (Wang et al., 2016). A few studies published in high impact journals have carried out indirect measurements based on spontaneous closure of the eyelids or heartbeat and, very recently, two studies have been based on EEG data (Chen et al., 2018; Haimovici, Tagliazucchi, Balenzuela, & Laufs, 2017; Stevner et al., 2019). All confirm the crucial interest in controlling this factor. However, the technical constraints with the GET are such in an MRI scanner that it is illusory to try to extend these studies to cohorts or populations of fragile patients.