This page was translated automatically from Turkish. Read the original

Bu sayfa, 26. Dünya Psikiyatri Kongresi'nde sunulan çalışmanın yapay zekâ desteğiyle hazırlanmış Türkçe özetidir. Çalışmanın sahibi sunan yazarlardır; atıf için orijinal kaynağa başvurun.

EV0927

Artificial intelligence in obsessive-compulsive disorder: current applications and future directions

ARTIFICIAL INTELLIGENCE IN OBSESSIVE–COMPULSIVE DISORDER: CURRENT APPLICATIONS AND FUTURE DIRECTIONS

Although artificial intelligence holds promise for early diagnosis, neurobiological subtyping, and personalized treatment in obsessive-compulsive disorder, current evidence is limited by small, heterogeneous samples.

The aim of this study is to examine the role of artificial intelligence (AI) in improving diagnosis and enhancing treatment interventions in obsessive-compulsive disorder (OCD). A narrative review was conducted using the PubMed and Scopus databases with the keywords "obsessive-compulsive disorder" and "artificial intelligence." The poster notes that, despite the early onset of OCD, there is an average delay of 17 years from symptom onset to diagnosis and treatment, which is associated with a poor clinical course. Six areas of AI application are outlined: a large language model diagnosing OCD from clinical vignettes more accurately than physicians, with this advantage also persisting in zero-shot identification before fine-tuning; the identification of neurobiologically distinct subtypes through unsupervised machine learning using structural and functional magnetic resonance imaging data, although findings are inconsistent across imaging modalities; closed-loop deep brain stimulation that monitors signals in real time and adjusts stimulation in treatment-resistant OCD; AI-generated exposure hierarchies found to be appropriate, specific, and clinically useful; machine learning models for predicting response to clomipramine and SSRIs; and continuous symptom monitoring using wearable devices that combine heart rate, electrodermal activity, and skin temperature. The limitations section reports that structural MRI classifiers in the multicenter ENIGMA-OCD consortium (more than 2,000 patients) could not reliably distinguish patients with OCD from healthy controls, particularly among unmedicated patients. An OCD-specific risk is highlighted: patient-facing chatbots may reinforce reassurance-seeking and ritualized behavior, making clinician oversight necessary. Frameworks such as FAITA-Mental Health are recommended for evaluating AI tools across six domains: reliability, user experience, user autonomy, equity and inclusion, transparency, and safety/crisis management. In conclusion, AI offers a path toward precision psychiatry in OCD through early detection, personalized interventions, and objective outcome measures, but prospective studies, multimodal models, and ethical and regulatory standards remain priorities.

Tüm WCP 2026 özetleri