Decoding dynamic affective responses to naturalistic videos with shared neural patterns.

Area of Science:

• Neuroscience
• Affective Computing
• Machine Learning

Background:

• Decoding affective states from neural activity is challenging, especially during dynamic, naturalistic experiences.
• Prior research often focuses on static stimuli, limiting applicability to real-world emotional processing.

Purpose of the Study:

• To investigate the feasibility of using shared neural patterns from brief affective picture viewing to decode dynamic emotional sequences during movie trailer watching.
• To develop and validate machine learning classifiers for predicting affective states in naturalistic settings.

Main Methods:

• Utilized functional Magnetic Resonance Imaging (fMRI) data from 28 participants viewing affective pictures and movie trailers.
• Identified responsive voxels in the bilateral occipital cortex (LOC) using General Linear Model (GLM) analysis.
• Applied between-subject hyperalignment to normalize neural responses across participants.
• Trained machine learning classifiers on affective picture data and tested their ability to decode emotional states during movie trailer viewing.

Main Results:

• Neural classifiers accurately identified valence and arousal categories of affective pictures within participants.
• Classifiers successfully tracked self-reported valence and arousal dynamics during movie trailer viewing.
• Aggregated neural decoding time series significantly correlated with dynamic ratings from a separate sample, validating the approach.

Conclusions:

• Shared neural representations from simple affective stimuli can be leveraged to decode complex, dynamic emotional experiences.
• This study demonstrates a promising method for real-time affective state decoding in naturalistic environments.
• Findings support the potential of pre-trained neural representations for understanding dynamic emotional responses.