Depressive Sympotoms
Conditions
Keywords
Brain-Computer Interface, Support Vector Machine, Brain Self-Regulation, Neuromodulation, EEG-Neurofeedback, neurofeedback
Brief summary
This completed feasibility study examined whether right-handed women aged 18-35 years with mild-to-moderate depressive symptoms could use real-time electroencephalography (EEG) neurofeedback based on a subject-independent classifier to regulate brain activity toward a classifier-defined reference pattern. Participants were randomly assigned to receive either classifier-contingent visual feedback with performance-contingent monetary rewards or noncontingent sham visual feedback with yoked monetary rewards. All participants completed five training sessions over approximately 3 weeks. The primary purpose was to determine whether overall classifier accuracy during training exceeded the theoretical 50% chance level. Changes in clinician-rated and self-reported depressive symptoms, associations between classifier performance and symptom change, EEG spatial patterns, and procedural tolerability were exploratory.
Detailed description
The subject-independent classifier had been developed from EEG recordings obtained from an independent reference sample of healthy right-handed women during positive autobiographical memory recall and motor imagery. It used a Filter Bank Common Spatial Pattern feature-extraction pipeline and a linear support vector machine to classify the two mental states from 28 EEG channels. After eligibility and baseline assessments, participants were randomly assigned in a 1:1 ratio to an experimental or control arm using a sequence generated before the intervention. Participants and the clinical psychologists administering the Hamilton Depression Rating Scale were masked to allocation. The experimental arm was implemented before the control arm because the control rewards were yoked to a reward sequence generated in the experimental arm; this operational sequence did not determine assignment. Each participant completed five EEG neurofeedback sessions on separate days over approximately 3 weeks. Each session contained four 300-second runs. In each run, participants completed positive autobiographical memory recall and motor imagery blocks while viewing a thermometer-like display. In the experimental arm, the display was updated approximately once per second according to the classifier output and monetary rewards depended on run-level classifier accuracy. In the control arm, display movements were random and independent of EEG activity, and rewards were yoked to the run-by-run reward sequence of the experimental participant with the highest total reward. This record was created retrospectively after study completion. The protocol information reflects the completed study procedures and the outcome hierarchy reported in the final manuscript; exploratory analyses have not been promoted to primary outcomes.
Interventions
BEHAVIORALContingent Subject-Independent Classifier-Based EEG Neurofeedback
Five sessions were delivered on separate days over approximately 3 weeks. Each session comprised four 300-second runs separated by 1-3-minute rest intervals. Each run contained four 20-second positive autobiographical memory recall blocks and four 20-second motor imagery blocks. A pretrained Filter Bank Common Spatial Pattern-linear support vector machine classifier processed 28 EEG channels in real time. The visual thermometer rose or fell by one level according to whether each classifier output matched the instructed mental state. At the end of each run, participants viewed classifier accuracy and received a prespecified performance-contingent monetary reward.
BEHAVIORALNoncontingent Sham EEG Feedback With Yoked Rewards
Participants received the same five-session schedule, task structure, visual display, and scripted instructions as the experimental arm. Each thermometer update was a randomly determined one-level increase or decrease and was independent of EEG activity. Rewards followed the run-by-run sequence obtained by the experimental-arm participant with the highest total reward and were not contingent on the control participant's classifier performance.
Sponsors
Pontificia Universidad Catolica de Chile
Study design
Allocation
RANDOMIZED
Intervention model
PARALLEL
Primary purpose
OTHER
Masking
SINGLE (Outcomes Assessor)
Masking description
Participants and the trained clinical psychologists who administered the Hamilton Depression Rating Scale were masked to group allocation throughout the study. The personnel operating the neurofeedback system were not reported as masked.
Intervention model description
Two parallel arms with 1:1 randomized allocation. The experimental arm was implemented before the control arm only to generate the yoked reward schedule; the implementation order did not determine assignment.
Eligibility
Sex/Gender
FEMALE
Age
18 Years to 35 Years
Healthy volunteers
No
Inclusion criteria
* Self-identification as a woman. * Female sex assigned at birth, reported during the intake interview * Self-reported right-handedness * Age 18-35 years * Beck Depression Inventory-II score of 14-29 * Hamilton Depression Rating Scale score below 24 * Score below 2 on the suicide-related item of both the Beck Depression * Inventory-II and Hamilton Depression Rating Scale
Exclusion criteria
* Current psychotherapy * Self-reported history of bipolar disorder or schizophrenia * Prior suicide attempt * Self-reported history of post-traumatic stress disorder, obsessive-compulsive disorder, substance use disorder, or anorexia nervosa * Neurological disorder, including dementia, brain tumor, or epilepsy * Intellectual disability * Regular or intermittent use of benzodiazepines, anticonvulsants, or psychostimulants * Use of another psychotropic medication unless the medication and dose had remained unchanged for at least 3 months before enrollment
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Mean Overall Classifier Accuracy Across Five EEG Neurofeedback Sessions | Across five training sessions over approximately 3 weeks | Classifier accuracy was the percentage of classifier outputs during regulation blocks that matched the instructed mental state. Each run contained 160 classifier outputs during regulation periods. Run-level accuracy was averaged across four runs to obtain a session-level value, and session-level accuracy was averaged across five sessions for each participant. Overall accuracy was compared with the theoretical 50% chance level within each arm and between arms. |
Countries
Chile
Contacts
PRINCIPAL_INVESTIGATORSergio M Ruiz, MD, PhD
Department of Psychiatry, School of Medicine, Pontificia Universidad Católica de Chile, Santiago, Chile
STUDY_CHAIRRanganatha Sitaram, ME, PhD
Department of Radiology, St. Jude Children's Research Hospital, Memphis (TN), USA.
STUDY_DIRECTORJaime A Pereira, MD, PhD
Department of Psychiatry, School of Medicine, Pontificia Universidad Católica de Chile, Santiago, Chile
Outcome results
None listed