The representation of vocalizations in structured noise in the mouse auditory cortex

Efficient processing of complex acoustic signals within noisy environments is a fundamental challenge for the auditory system. This study investigates the cortical representation of structured noise and vocalizations embedded in it using widefield calcium imaging with the high-speed Calcium indicator jGCaMP8m in the mouse auditory cortex. The spectrotemporal statistics of the structured noise were controllable, including its temporal variance and cross-frequency correlations. We find that structured noise invokes spatially and temporally complex neural responses across the auditory cortex that reflect the spectrotemporal properties of the noise. We identified partially distinct noise response components: noise onset responses showed stronger overlap with primary than secondary areas, while later suppression components did not show a significant primary-secondary bias. Vocalization-evoked responses were significantly attenuated in the presence of noise compared to silent contexts, exhibiting reduced onset magnitude and altered adaptation dynamics. Using decoding we find the neural representation of vocalizations to become more discriminable with longer preceding noise exposure. Furthermore, structured noise cochleograms could be reconstructed from neural population activity, with performance depending on stimulus statistics, tending to be higher in primary than secondary auditory cortex while remaining substantial across both. These findings suggest that the mouse auditory cortex represents acoustic targets in noise through context-dependent and spatially distributed population activity, shaped by both the statistical structure and temporal history of the acoustic background.