PSYCH-GPT: Alignment between natural language processing and large language models in psychiatric disorders

Status

Active, not recruiting

Phases

Unknown

Study type

Observational

Source

DRKS

Registry ID

DRKS00039902

Enrollment

120

Registered

2026-04-09

Start date

2026-08-01

Completion date

Unknown

Last updated

2026-08-03

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

F90 F32 F33 F20

Interventions

Group 1: All participants undergo the same examination protocol: digital screening and informed consent (approx. 7 min.), neuropsychological tests and clinical interviews (approx. 90 min.), and a func

Eligibility

Sex/Gender

All

Age

18 Years to 60 Years

Inclusion criteria

Inclusion criteria: Native-level German language proficiency Capacity to consent No contraindications for MRI For patient groups additionally: confirmed diagnosis of ADHD, depression, or schizophrenia (clinically documented); stable medication for at least 2 weeks; for depression: at least moderate severity

Exclusion criteria

Exclusion criteria: MRI-relevant metal implants (especially pacemakers, prostheses, non-MRI-compatible intrauterine devices) Fresh, large-area, or metal-containing tattoos; any tattoos in the head or neck region Acute psychotic or manic episode; acute suicidality; acute substance-related intoxication or withdrawal Severe neurological disorders Significant hearing impairment For healthy controls additionally: existing psychiatric disorder For patient groups additionally: severe psychiatric comorbidity beyond the primary diagnosis

Design outcomes

Primary

Measure	Time frame
WHAT: Alignment between fMRI activation patterns during naturalistic language processing and contextual representations of large language models (LLMs); group difference between ADHD, depression, schizophrenia, and healthy controls. WHEN: Once during the fMRI session (time point 3). HOW: Computation of similarity matrices via Representational Similarity Analysis (RSA) from fMRI data and LLM representations; group comparison via one-way ANOVA.	—

Measure

Time frame

WHAT: Alignment between fMRI activation patterns during naturalistic language processing and contextual representations of large language models (LLMs); group difference between ADHD, depression, schizophrenia, and healthy controls. WHEN: Once during the fMRI session (time point 3). HOW: Computation of similarity matrices via Representational Similarity Analysis (RSA) from fMRI data and LLM representations; group comparison via one-way ANOVA.

—

Secondary

Measure	Time frame
(1) Correlation between individual alignment scores and clinical-symptomatic parameters (e.g., symptom severity, medication); assessed via validated clinical scales (BPRS, TALD, DIVA, DISCOURSE). (2) Accuracy of responses to comprehension questions about the spoken language stimuli; assessed immediately after the fMRI session. (3) Comparison of different LLMs and model parameters regarding their predictive quality for brain-based language representations; analysed after completion of data collection.	—

Measure

Time frame

(1) Correlation between individual alignment scores and clinical-symptomatic parameters (e.g., symptom severity, medication); assessed via validated clinical scales (BPRS, TALD, DIVA, DISCOURSE). (2) Accuracy of responses to comprehension questions about the spoken language stimuli; assessed immediately after the fMRI session. (3) Comparison of different LLMs and model parameters regarding their predictive quality for brain-based language representations; analysed after completion of data collection.

—

Countries

Germany

Contacts

Public ContactHendrik Kröger

Universitätsklinikum Frankfurt, Klinik für Psychiatrie, Psychosomatik und Psychotherapie

Kroeger@med.uni-frankfurt.de+49 151 1719 2100

Outcome results

None listed

Source: DRKS (via WHO ICTRP) · Data processed: Aug 10, 2026