IntervEEG-LLM: Exploring EEG-Based Multimodal Data for Customized Mental Health Interventions

Incorporating multimodal physiological signals into mental health intervention may greatly enhance its effectiveness due to the rich semantic information they contain. EEG signals are particularly common and valuable for capturing brain activity, which can provide objective insights into users' mental states. However how to interpret and leverage these signals remains under-explored. This paper initiated the first attempt by proposing IntervEEG-LLM, an end-to-end framework leveraging EEG-based multimodal signals for mental health intervention. IntervEEG-LLM first used LLMs for heterogeneous multimodal physiological understanding of signals and predicted the intermediate modeling results of users. Then it leveraged effective intervention strategies to interpret and apply the modeling results, offering concrete intervention text through natural language interfaces. Our unified framework integrates three EEG-based physiological data modalities and five mental health strategies, offering a promising approach to signal-based mental health intervention.