Identifying Neuroimaging-Based Biomarkers for Obsessive-Compulsive Disorder: A Resting-State fMRI and Machine Learning Study in Patients and Unaffected First-Degree Relatives.
Authors
Affiliations (6)
Affiliations (6)
- Department of Psychosomatic Medicine, The 1(st) Affiliated Hospital, Jiangxi Medical College, Nanchang University, Nanchang, Jiangxi, China.
- Department of Clinical Psychology, Qilu Hospital (Qingdao), Cheeloo College of Medicine, Shandong University, Qingdao, Shandong, China.
- Department of Radiology, The 1st Affiliated Hospital, Jiangxi Medical College, Nanchang University, Nanchang, China.
- Department of Nursing, The 1(st) Affiliated Hospital, Jiangxi Medical College, Nanchang University, Nanchang, China.
- Department of Radiology, The 1st Affiliated Hospital, Jiangxi Medical College, Nanchang University, Nanchang, China. Electronic address: [email protected].
- Department of Psychosomatic Medicine, The 1(st) Affiliated Hospital, Jiangxi Medical College, Nanchang University, Nanchang, Jiangxi, China. Electronic address: [email protected].
Abstract
Identifying whether brain alterations in obsessive-compulsive disorder (OCD) represent the illness itself or an underlying genetic vulnerability is challenging, primarily due to the confounding effects of medication and chronic illness. To disentangle these disease-state and genetic-trait markers, we investigated a unique cohort of first-episode, drug-naïve patients and their unaffected first-degree relatives (UFDR). Resting-state functional magnetic resonance imaging (fMRI) data were acquired from 72 first-episode, drug-naïve OCD patients, 23 UFDR, and 56 healthy controls (HC). We computed the amplitude of low-frequency fluctuations (ALFF), regional homogeneity (ReHo), and degree centrality (DC). Following covariate-controlled statistical comparisons, support vector machine (SVM) models with nested cross-validation were employed to evaluate the exploratory classification capacity of the identified functional features. No significant main effects were found for ReHo across groups. However, OCD patients exhibited significantly elevated ALFF in the left orbital middle frontal gyrus and left dorsolateral superior frontal gyrus compared to HC and UFDR, respectively. ALFF values in both regions were positively correlated with obsessive-compulsive symptom severity. Meanwhile, the UFDR group demonstrated distinct network alterations, including increased DC in the left caudate nucleus and decreased DC in the left orbital inferior frontal gyrus compared to HC. The SVM classifiers differentiated the groups with promising capacity, achieving areas under the curve of 0.834 (UFDR vs. HC), 0.791 (OCD vs. HC), and 0.801 (OCD vs. UFDR). This study helps separate state-dependent from trait-related functional features in OCD. Prefrontal ALFF hyperactivity characterizes the active disease state, whereas striatal-frontal DC alterations in unaffected relatives represent candidate endophenotypes for genetic vulnerability. While these neural signatures show exploratory discriminatory capacity, their clinical utility for early diagnosis and risk stratification requires strict validation in larger, multi-site, and longitudinal cohorts.