IntroductionMental health crisis response requires police officers to regulate their emotional and physiological states while making rapid decisions in socially complex situations. Virtual reality (VR) offers a promising platform for immersive de-escalation training, yet the integration of affective sensing into these environments remains limited. This study examined whether physiological measures, together with complementary experiential measures, could capture meaningful affective responses during a VR-based mental health crisis response (MHCR) training scenario.MethodsEleven trained active-duty police officers in Canada completed a co-designed VR interactive simulation involving a virtual individual experiencing symptoms of early psychosis with two endings (escalation and de-escalation) depending on user interaction with the virtual individual. Participants wore an HTC Vive Pro VR headset integrated with an Emteq Pro mask comprising multiple facial electromyography (fEMG) and an EmotiBit wearable. We performed event-related analyses that focused on key scenario moments: Start, Door, and Action/Resolution. The events were analyzed in terms of physiological metrics and affective ranges of valence and arousal using fEMG signals across seven facial muscle channels, heart rate, HRV metrics, electrodermal activity (EDA), and pre/post subjective measures of perceived usefulness and debriefing.ResultsSignificant event-related differences in facial activation, particularly for the orbicularis and zygomaticus channels, with stronger responses at the “Door” event relative to baseline. Frontalis and corrugator activity showed descriptive increases during escalation and more negative outcome pathways. Valence-arousal patterns indicated that officers who drew their weapons during the de-escalation scenario clustered more strongly in negative valence regions, whereas others showed broader, more adaptive affective trajectories. Heart rate was significantly higher during the scenario than baseline, and HRV-based comparisons suggested more adaptive autonomic profiles in officers that scored high in de-escalation skills. EDA generally increased during the scenario for most participants, indicating elevated sympathetic activation.DiscussionThese exploratory findings support the feasibility of using fEMG within immersive VR police training to capture fine-grained affective dynamics during de-escalation. The combination of event-related facial analysis, autonomic measures, and subjective evaluations provides a richer view of how officers respond to high-stakes mental health encounters.