Pattern classification based on the amygdala does not predict an individual's response to emotional stimuli

Area of Science:

• Neuroscience
• Cognitive Neuroscience
• Neuroimaging

Background:

• Functional magnetic resonance imaging (fMRI) studies commonly report significant group-level amygdala activation in response to emotional stimuli.
• The extent to which these findings generalize to individual participants using multi-voxel pattern analysis (MVPA) remains less clear.
• Understanding individual-level brain activity patterns is crucial for targeted interventions like neurofeedback.

Purpose of the Study:

• To investigate whether emotional valence decoding using MVPA is reliable at the individual participant level.
• To compare the efficacy of amygdala-only versus whole-brain analysis for decoding emotional valence.
• To determine the spatial distribution of brain activity patterns encoding emotional information.

Main Methods:

• Combined fMRI data from 112 participants across two prior studies.
• Employed a linear support vector machine (SVM) to decode emotional picture valence (negative, neutral, positive).
• Performed region-of-interest analysis focused on the amygdala and a whole-brain exploratory analysis.

Main Results:

• Amygdala-based MVPA yielded statistically significant valence decoding in only 4.5% of participants (mean accuracy: 37% ± 5%).
• Whole-brain MVPA achieved statistically significant decoding in 50.9% of participants (mean accuracy: 49% ± 6%).
• Individual participant decoding accuracy was significantly higher with whole-brain analysis compared to amygdala-only analysis.

Conclusions:

• Emotional picture valence is encoded by spatially distributed brain activity patterns, not solely by the amygdala.
• Whole-brain MVPA is more effective than amygdala-focused analysis for individual-level decoding of emotional valence.
• Findings have implications for emotion regulation research and treatments utilizing real-time fMRI neurofeedback targeting the amygdala.