Linking Subjective Experience of Anxiety to Brain Function using Natural Language Processing

# Linking Subjective Experience of Anxiety to Brain Function using Natural Language Processing ## Overview - Corresponding author: Peter Kirk (email: [e-mail]; ORCID: https://orcid.org/my-orcid?orcid=0000-0003-0786-3039). - Please note: data is only available for participants who consented to data sharing. - Project name: "Linking Subjective Experience of Anxiety to Brain Function using Natural Language Processing", published by Social Cognitive and Affective Neuroscience in 2026. - Data collection ran between 2018 and 2021. - The procedures included evaluation of anxiety symptoms, neuroimaging, and post-scan interviews. Functional MRI consisted of three runs: 1) resting-state; 2) movie-watching; and 3) resting-state. In this study, we used a semi-naturalistic, anxiogenic stimulus (animated movie) to evoke anxiety during fMRI in a pediatric sample with and without anxiety disorders (N = 84, before exclusion). After, participants provided verbal responses to interview questions about the stimulus. The study focused on which brain circuits may instantiate subjective experience of anxiety. - This dataset includes demographics, anxiety symptoms, neuroimaging data, and natural language processing derivatives. - Variables for primary hypothesis-testing: anxiety diagnosis ("clinical_status"), anxiety symptom severity ("scared_baseline"), natural language processing of interview data ("prompt valence" and "prompt embedding"), and movie-watching fMRI ("task-francis"). A brief discussion of condition variables (sometimes called contrasts or independent variables) that were varied across the experiment. - The experiment consisted of three conditions: "fMRI task-restpre", "fMRI task-francis", and "fMRI task-restpost". In our study, we only analyzed "fMRI task-francis". ## Methods ### Subjects 84 participants were recruited and were required to be aged 8-18 years (M = 14.1 years, SD = 2.4; 54 Female, 30 Male), have a WASI IQ > 70, proficient in English, and not have MRI contraindications. Patients with anxiety were required to have a clinician-confirmed diagnosis for a primary anxiety disorder (48 healthy volunteers and 36 had anxiety disorders). Exclusion criteria included any other serious medical condition, including a pervasive neurodevelopmental disorder, substance use, suicidal ideation, major depressive disorder, obsessive-compulsive disorder, and posttraumatic stress disorder. A lifetime history of psychosis, bipolar disorder, or extreme trauma was also exclusionary. Important: see the Notes section below for additional analytic exclusions. ### Imaging Imaging data was collected on one of two GE MR750 3T MRIs. T1w structural scans were acquired with MPRAGE sequence with 1mm isotropic voxels. Multiecho fMRI EPI scans were acquired with 3.8mm istropic voxels, TR 2s and TE = [0.015s, 0.0286s, 0.0422s] for pre-movie resting-state (8.2mins), anxiogenic movie task-francis (5.8 mins), and post-movie resting-state (2.3mins). The first 4 volumes of the functional scans are discard/dummy volumes. The movie-watching functional scans begin with a fixation cross prior to the Francis movie; the first 4 volumes acquired during this fixation period are dummy volumes. Two sets of field maps with AP and PA phase-encoding directions were acquired. For distortion correction, each fMRI scan was paired with the set of field maps acquired closest in time. ### Task organization Task order was not counter balanced, as the movie fMRI run was required to be positioned in between pre-movie rest and post-movie rest. ### Additional data acquired Clinical status was established via clinician-report on the Kiddie Schedule for Affective Disorders and Schizophrenia – Present and Lifetime Version (Kaufman et al., 1997). Anxiety symptoms were also assessed with a clinician report of the Pediatric Anxiety Rating Scale (PARS; Riddle et al., 2022) and combined parent-child report on the Screen for Child Anxiety and Related Disorders (SCARED; Birmaher et al., 1997). Patients subsequently received 8-12 weeks of cognitive behavioral therapy, after which anxiety symptoms were re-assessed. Anxiety symptom measures at baseline and after treatment are denoted in data provided (_baseline vs _treatment). ### Experimental location Participants were recruited from the greater Washington, D.C., area. ### Missing data - Data is only available for participants who consented to data sharing (n=75). ### Notes Some structural scans contain artifacts, including motion and Gibbs ringing artifacts. Below, we note certain caveats and errors with some participant's data. This data was excluded from our analyses but have been included here for completeness. - "sub-s24620" data was acquired with an incorrect repetition time. - "sub-s23891" data was not successfully preprocessed due to high motion. - "sub-s24017" and "sub-s24115" were excluded due to high motion (percentage of volumes exceeding framewise displacement > 0.5mm = 20%). Thanks to the NIMH Data Science & Sharing Team for their help curating this dataset.