Decoding Emotional Dynamics in Bipolar Disorder

This neuroimaging study investigates the neural mechanisms underlying emotional dynamics and mood instability in individuals with bipolar disorder (BD). Bipolar disorder is characterized by rapid and intense mood fluctuations, yet the neurobiological basis of these transitions, how the brain shifts between emotional states in real time, remains poorly understood. The study aims to identify the moment-to-moment brain processes that drive emotional lability and to explore whether positive emotion amplification can stabilize emotional and neural states in BD.

Study Design

This is a study conducted at the Laureate Institute for Brain Research (LIBR) in Tulsa, Oklahoma. The study includes 72 participants total: 36 adults diagnosed with bipolar disorder type I or II (currently in a depressive or mixed state) and 36 healthy control participants without psychiatric history. Participants will complete two visits:

• A preparation session for consent, clinical interviews, and questionnaire completion, and
• A MRI scanning session that includes both resting-state and task-based fMRI.

Data will be collected using multimodal methods, including functional magnetic resonance imaging (fMRI), diffusion weighted imaging (DWI), structural MRI, and physiological monitoring (heart rate, respiration). Behavioral and emotional measures will be recorded throughout the study to align neural data with subjective emotional experience.

Scientific Rationale Mood instability is a defining and impairing feature of bipolar disorder, associated with deficits in emotion regulation and cognitive control. Prior neuroimaging work has identified alterations in prefrontal-limbic circuitry, including decreased activation in regulatory regions such as the anterior cingulate cortex (ACC) and prefrontal cortex (PFC), and increased activation in emotion-responsive regions such as the amygdala. However, most studies examine static mood states rather than dynamic fluctuations in emotional experience.

The present study applies machine learning, complexity science, and network control theory to quantify and model emotional state dynamics. By decoding brain activity during emotion regulation tasks, the research aims to characterize how emotional states evolve over time, how this differs in BD compared to healthy controls, and whether targeted regulation strategies, specifically positive emotion amplification, can modulate these dynamics.

Specific Aims and Hypotheses Aim 1: Decode momentary emotional states from whole-brain fMRI data using machine learning approaches.

Hypothesis 1: A machine learning classifier can accurately distinguish distinct emotional states (e.g., rumination vs. positive reflection) from fMRI activation patterns. BD participants will exhibit more unstable, fluctuating state trajectories than healthy controls.

Aim 2: Quantify emotional dynamics using metrics from complexity science and network control theory.

Hypothesis 2: Individuals with BD will show higher emotional metastability and lower fractal scaling-indicators of greater temporal irregularity in brain activity-relative to healthy controls. Network control theory analysis will identify the brain regions that contribute to state transitions.

Aim 3: Examine the effects of positive emotion amplification on emotional stability and brain network dynamics.

Hypothesis 3: The regulation of positive affect will engage cognitive control regions (e.g., dorsolateral PFC, ACC) and promote more stable emotional trajectories in BD participants.

Experimental Tasks and Procedures

• Visit 1 (Preparation Session):

Participants will undergo informed consent, psychiatric screening (using the MINI), and a series of standardized questionnaires assessing mood, emotion regulation, anxiety, rumination, and hedonic capacity (e.g., MADRS, YMRS, PANAS-X, DERS, ERQ, STAI, PROMIS scales).

Participants will also recall eight autobiographical events-four positive (reminiscence) and four negative (rumination)-and write brief keyword descriptions of each. These personalized cues will be used later in the MRI task to elicit emotional states without revealing personal content.

• Visit 2 (MRI Scanning Session):

Participants will complete both resting-state and task-based MRI scans lasting up to two hours. Physiological signals (heart rate and respiration) will be recorded concurrently to remove physiological artifacts and examine autonomic correlates of emotion.

MRI sequences include:

• High-resolution T1-weighted structural scans
• Diffusion-weighted scans for white-matter connectivity
• Resting-state fMRI (12-minute duration)
• Task fMRI: Think and Regulate Affective States Task (TReAT)

TReAT Task Overview

The Think and Regulate Affective States Task (TReAT) is a novel paradigm designed to model real-world emotional processing. Participants are presented with brief cue words corresponding to their personal autobiographical events and alternate between several types of blocks:

• Think Blocks: Participants think about the cued event, immersing themselves in the associated emotional experience.
• Rating Blocks: Immediately after, they rate emotional valence (positive-negative) and arousal using the Affective Slider.
• Regulation Blocks: Participants attempt to amplify positive mood while focusing on the same cue.
• Attention Blocks: Participants perform a brief arrow-direction attention task (6 trials, 2 seconds each) to clear cognitive load between emotional blocks.
• Rest Blocks: Participants fixate on a cross, instructed to relax and clear their thoughts.

These blocks are repeated across four fMRI runs, each lasting approximately 12-15 minutes. The design allows modeling of both spontaneous and regulated emotional states, enabling fine-grained temporal decoding of emotional dynamics.

After each run, participants rate fatigue, sleepiness, and emotional engagement. Post-scan questionnaires (e.g., PANAS-X, STAI-S, Feedback Questionnaire) assess emotional and physical comfort.

Data Analysis Plan Functional MRI data will be preprocessed using standard pipelines and analyzed with multivariate pattern analysis (MVPA) to classify emotional states. State-space trajectory analyses will examine how decoded brain states fluctuate over time within and between subjects. Measures of metastability, fractal scaling, and network controllability will quantify the temporal complexity and flexibility of brain networks.

Between-group comparisons (BD vs. HC) will assess whether BD participants exhibit greater temporal irregularity or reduced control energy in emotion-related circuits. The modulation of these parameters by positive emotion regulation will be tested using within-subject contrasts of Regulation vs. Think blocks.

Scientific and Clinical Significance This study integrates cutting-edge computational methods: machine learning, complexity metrics, and network control theory to decode the temporal structure of emotion regulation in bipolar disorder. By identifying neurobiological signatures of instability and testing whether positive affect regulation stabilizes these dynamics, this work aims to bridge the gap between affective neuroscience and personalized psychiatry.

The resulting dataset will contribute to the National Institute of Mental Health (NIMH) Data Archive and inform future large-scale studies targeting biomarkers of emotional dysregulation. Ultimately, this research will lay the groundwork for adaptive, brain-state-driven treatments that dynamically respond to patients' emotional states, offering new strategies for mood stabilization in bipolar disorder.