J Affect Disord. 2025 Aug 14:120068. doi: 10.1016/j.jad.2025.120068. Online ahead of print.
ABSTRACT
BACKGROUND: This study aims to apply a semi-supervised machine learning approach for classifying major depressive disorder (MDD) patients into more homogeneous cognitive subtypes based on multidimensional cognitive profiles, and to perform multimodal neuroimaging to identify subtype-specific neural signatures.
METHODS: A total of 147 MDD patients and 222 healthy controls (HCs) completed the Cambridge Neuropsychological Test Automated Battery (CANTAB) and magnetic resonance imaging (MRI) scans. Cognitive subtypes were derived based on neurocognitive profiles using heterogeneity through discriminative analysis (HYDRA). General linear models (GLMs) were employed to assess differences across groups in neurocognitive indexes and neuroimaging data followed by Tukey’s post-hoc test for pairwise comparisons between the groups.
RESULTS: Based on cognitive profiles, MDD patients were classified into cognitive deficit (CD, N = 75) and cognitive preservation (CP, N = 72) subtypes. Voxel-based morphometry (VBM) revealed reduced grey matter volume (GMV) in the left fusiform gyrus and left cerebellum in MDD patients when compared to HCs, with CD patients showing greater atrophy than patients in CP subtype. Meanwhile, the amplitude of low-frequency fluctuations (ALFF) in the temporal lobe of both MDD subtypes was decreased when compared to that of HCs, showing no inter-subtype differences.
CONCLUSIONS: A subtype of MDD characterized by comprehensive cognitive deficits is associated with structural atrophy in the left fusiform gyrus and cerebellum, suggesting these regions as potential biomarkers for the cognitive deficit subtype of MDD. However, no significant differences in ALFF were observed between the two cognitive subgroups.
PMID:40818498 | DOI:10.1016/j.jad.2025.120068
Recent Comments