Functional parcellation of mouse visual cortex using statistical techniques reveals response-dependent clustering of...

Area of Science:

• Neuroscience
• Computational Neuroscience

Background:

• The mouse visual cortex comprises multiple areas with retinotopic maps.
• These areas are traditionally viewed as discrete processing units.
• Conventional methods focus on input-output neuronal responses.

Purpose of the Study:

• To determine if six core visual areas exhibit functionally distinct responses.
• To apply machine learning to classify areas based on neural activity patterns.
• To investigate if these distinctions persist during resting-state activity.

Main Methods:

• Supervised machine learning classifiers were applied to neural responses.
• Two datasets (wide-field and two-photon imaging) were utilized.
• Resting-state cortical responses were analyzed for area boundary modeling.

Main Results:

• Classifiers accurately predicted visual area labels, consistent with retinotopy.
• Area boundaries were successfully modeled using resting-state activity.
• Classification accuracy was higher with visual stimuli compared to resting state.

Conclusions:

• Neural responses from visual cortical areas are effectively classifiable using data-driven models.
• Distinct intra-areal circuits likely underlie these unique activity patterns.
• Machine learning provides a powerful tool for dissecting functional organization in the visual cortex.