Deep Neural Networks Explain Spiking Activity in Auditory Cortex.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Auditory Processing

Background:

• Artificial neural networks (ANNs) excel at predicting neural responses in primate visual and auditory cortex for static stimuli and at slower time scales.
• The efficacy of ANNs in predicting fine-grained spiking activity in the auditory cortex, crucial for processing dynamic sounds like speech, remains largely unexplored.

Purpose of the Study:

• To investigate whether ANNs trained on speech audio can predict neural spiking activity in the auditory cortex at fine time scales (50 ms and below).
• To compare the predictive power of trained ANNs against traditional spectrotemporal-receptive-field models and untrained networks.

Main Methods:

• Utilized ANNs trained on speech audio datasets.
• Performed acute multi-electrode recordings from the auditory cortex of squirrel monkeys.
• Analyzed neural spike counts in response to speech and monkey vocalizations using varying time bin widths.

Main Results:

• Trained ANNs successfully predicted neuronal spike counts in the auditory cortex at time scales of 50 ms and below.
• ANNs explained significantly more explainable neural variance compared to traditional spectrotemporal-receptive-field models and untrained networks.
• Deeper ANN layers showed better prediction for non-primary neurons, indicating layer-specific processing, though significant neuron-to-neuron variability was observed.

Conclusions:

• Trained ANNs are powerful tools for modeling fine-grained neural dynamics in the auditory cortex, extending their predictive success to rapid temporal processing.
• The findings highlight the potential of ANNs to reveal complex neural coding strategies in audition that may be missed by coarser analysis methods.