Generative artificial intelligence-enabled dynamic detection of rat nicotine-related circuits

Nicotine addiction circuits involve integrating specific brain regions that alter to frequent smoking. Detection of these circuits via fMRI contributes to understanding addiction-related mechanisms. Identification of the functional circuits and networks altered by nicotine is essential to improve the treatment of nicotine addiction. However, analyzing fMRI data and detecting functional addiction circuits still have challenges. In this work, we developed a generative AI-enabled framework, rat addiction-related circuits detection platform (RADP), to detect nicotine-related circuits. It has an end-to-end pipeline: functional imaging data acquisition from neurobiological experiments, computational modeling for brain networks, and a novel generative model including spatiotemporal transformer auto-encoder (STA) and dynamic circuits analysis. The proposed spatiotemporal representation contrasting trains the encoder of STA to contrastively capture representations between the addictive and the control groups. Experimental results indicate that the framework can efficiently detect the verified addiction circuits and discover the unknown but significant circuits. Moreover, RADP can be served as a general tool which can be extended to other brain circuits.