Cortex-Wide Preservation of Multi-Stimulus Information across Degrees of Network Synchronization

Abstract

Sensory-evoked cortical responses vary with global network dynamics, yet the link between cortical state and stimulus processing remains unclear. Here, we introduce a paradigm that jointly decodes stimulus identity and network state from wide-field calcium imaging in mice undergoing multisensory and optogenetic stimulation across isoflurane-induced transitions from compartmentalized to synchronized activity. Effective dimensionality, a summary measure of network complexity, correlated well with anesthesia depth, while non-linear contrastive learning achieved >97% stimulus decoding accuracy across all states. Individual cortical regions maintained [≥]82.5% accuracy even during deep anesthesia with prominent slow waves. Preservation of stimulus-specific information extended throughout the cortex, demonstrating that distinct representations remain decodable within synchronized networks. Direct optogenetic cortical stimulation exhibited state-invariant decoding performance, contrasting with anesthesia-dependent decline observed for sensory stimuli. Multi-stimulus cortical representations remain decodable across varying levels of network synchronization, with implications for brain-machine interfaces and clinical tools that assess preserved cortical responsiveness under variable arousal conditions.