Machine-learning based brain age estimation in major depression showing no evidence of accelerated aging. (30th August 2019)

• Record Type: Journal Article
• Authors: Besteher, Bianca; Gaser, Christian; Nenadić, Igor
• Is Part Of: Psychiatry research. Volume 290(2019)
• Page Start: 1
• Page End: 4
• Publication Date: 2019-08-30
• Subjects: Ageing -- Machine learning -- Magnetic resonance imaging (MRI) -- Major depression; Psychiatry -- Periodicals; Brain -- Imaging -- Periodicals; Psychiatry -- Periodicals; Diagnostic Imaging -- Periodicals; Psychiatrie -- Périodiques; Cerveau -- Imagerie pour le diagnostic -- Périodiques; 616.890754
• DOI: 10.1016/j.pscychresns.2019.06.001
• Languages: English
• ISSNs: 0925-4927

Abstract

Highlights: Use of a novel brain structural parameter, the BrainAGE score (brain age estimation gap), based on a machine-learning approach to estimate subjects' age based on T1 MRIs. No evidence for accelerated brain aging in MDD patients. Pilot cohort as a reference for further studies on brain-ageing in MDD. Abstract: Molecular biological findings indicate that affective disorders are associated with processes akin to accelerated aging of the brain. The use of the BrainAGE (brain age estimation gap) framework allows machine-learning based detection of a gap between age estimated from high-resolution MRI scans an chronological age, and thus an indicator of systems-level accelerated aging. We analysed 3T high-resolution structural MRI scans in 38 major depression patients (without co-morbid axis I or II disorders) and 40 healthy controls using the BrainAGE method to test the hypothesis of accelerated aging in (non-psychotic) major depression. We found no significant difference (or trend) for elevated BrainAGE in this pilot sample. Unlike previous findings in schizophrenia (and partially bipolar disorder), unipolar depression per se does not seem to be associated with accelerated aging patterns across the brain. However, given the limitations of the sample, further study is needed to test for effects in subgroups with comorbidities, as well as longitudinal designs.