A Data-Driven Closed-Loop Control Approach to Drive Neural State Transitions for Mechanistic Insight

Area of Science:

• Neuroscience
• Computational Psychiatry
• Affective Science

Background:

• Altered affective state dynamics persist in remitted major depressive disorder (rMDD).
• Individuals with rMDD exhibit heightened reactivity to negative moods and slower recovery.
• Neural state transition flexibility may underlie these affective dynamic changes.

Purpose of the Study:

• To characterize dynamical mechanisms governing brain state transitions in rMDD.
• To infer optimal control policies for mood state shifts using fMRI data.
• To understand neural underpinnings of vulnerability to depression recurrence.

Main Methods:

• Developed a framework combining dynamical system reconstruction (DSR) and model-based control.
• Applied nonlinear DSR models to fMRI data from rMDD participants and healthy controls (HC).
• Inferred region-specific, state-dependent control strategies for transitions between resting and sad mood states.

Main Results:

• rMDD participants required less energy to shift into and out of sad mood states compared to HC.
• Specific brain regions (sgACC, NAcc) demonstrated higher controllability.
• rMDD showed a residual bias toward sad mood states upon return to rest, despite similar proximity to the resting state target.
• Elevated network coupling in rMDD, particularly towards the DLPFC, correlated with lower control energy.

Conclusions:

• rMDD dynamics facilitate entry into sad states but hinder complete disengagement.
• Findings suggest a residual bias towards negative affective states in rMDD.
• The study demonstrates the utility of closed-loop control applied to DSR for mechanistic insights into brain state transitions and cognitive vulnerability.