Finding the balance between model complexity and performance: Using ventral striatal oscillations to classify feeding...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Behavioral Neuroscience

Background:

• The ventral striatum (VS) is crucial for appetitive behaviors.
• Neuromodulation of the VS shows promise for treating appetitive disorders.
• Local field potential (LFP) oscillations in the VS offer insights into appetitive behavior.

Purpose of the Study:

• To investigate the predictive power of VS LFP oscillations for feeding behavior.
• To explore the use of machine learning for analyzing neural signals related to appetite.
• To identify neural correlates of feeding initiation and consumption in rats.

Main Methods:

• Recorded LFPs from nucleus accumbens core and shell in male rats during food access.
• Utilized logistic regression and the machine learning algorithm lasso to analyze VS LFPs.
• Varied conditions of hunger and food palatability to assess feeding behavior.

Main Results:

• Successfully predicted the amount of food eaten, increased consumption after deprivation, and food type.
• Predicted feeding initiation up to 42.5 seconds in advance.
• Classified current behavior as feeding or not-feeding with high accuracy using alpha and high gamma frequencies.

Conclusions:

• VS LFP oscillations contain predictable information about appetitive behaviors.
• Machine learning can effectively decode neural activity for behavioral prediction.
• These findings support the development of responsive neuromodulation systems for appetitive disorders.