Optimising brain age estimation through transfer learning: A suite of pre-trained foundation models for improved performance and generalisability in a clinical setting.

Estimated age from brain MRI data has emerged as a promising biomarker of neurological health. However, the absence of large, diverse, and clinically representative training datasets, along with the complexity of managing heterogeneous MRI data, presents significant barriers to the development of accurate and generalisable models appropriate for clinical use. Here, we present a deep learning framework trained on routine clinical data (N up to 18,890, age range 18-96 years). We trained five separate models for accurate brain age prediction (all with mean absolute error ≤4.0 years, R2 ≥ .86) across five different MRI sequences (T2 -weighted, T2 -FLAIR, T1 -weighted, diffusion-weighted, and gradient-recalled echo T2 *-weighted). Our trained models offer dual functionality. First, they have the potential to be directly employed on clinical data. Second, they can be used as foundation models for further refinement to accommodate a range of other MRI sequences (and therefore a range of clinical scenarios which employ such sequences). This adaptation process, enabled by transfer learning, proved effective in our study across a range of MRI sequences and scan orientations, including those which differed considerably from the original training datasets. Crucially, our findings suggest that this approach remains viable even with limited data availability (as low as N = 25 for fine-tuning), thus broadening the application of brain age estimation to more diverse clinical contexts and patient populations. By making these models publicly available, we aim to provide the scientific community with a versatile toolkit, promoting further research in brain age prediction and related areas.

Adolescent to young adult longitudinal development of subcortical volumes in two European sites with four waves.

Adolescent subcortical structural brain development might underlie psychopathological symptoms, which often emerge in adolescence. At the same time, sex differences exist in psychopathology, which might be mirrored in underlying sex differences in structural development. However, previous studies showed inconsistencies in subcortical trajectories and potential sex differences. Therefore, we aimed to investigate the subcortical structural trajectories and their sex differences across adolescence using for the first time a single cohort design, the same quality control procedure, software, and a general additive mixed modeling approach. We investigated two large European sites from ages 14 to 24 with 503 participants and 1408 total scans from France and Germany as part of the IMAGEN project including four waves of data acquisition. We found significantly larger volumes in males versus females in both sites and across all seven subcortical regions. Sex differences in age-related trajectories were observed across all regions in both sites. Our findings provide further evidence of sex differences in longitudinal adolescent brain development of subcortical regions and thus might eventually support the relationship of underlying brain development and different adolescent psychopathology in boys and girls.

Neural responses to facial emotions and subsequent clinical outcomes in difficult-to-treat depression.

BACKGROUND: Amygdala and dorsal anterior cingulate cortex responses to facial emotions have shown promise in predicting treatment response in medication-free major depressive disorder (MDD). Here, we examined their role in the pathophysiology of clinical outcomes in more chronic, difficult-to-treat forms of MDD. METHODS: Forty-five people with current MDD who had not responded to ⩾2 serotonergic antidepressants (n = 42, meeting pre-defined fMRI minimum quality thresholds) were enrolled and followed up over four months of standard primary care. Prior to medication review, subliminal facial emotion fMRI was used to extract blood-oxygen level-dependent effects for sad v. happy faces from two pre-registered a priori defined regions: bilateral amygdala and dorsal/pregenual anterior cingulate cortex. Clinical outcome was the percentage change on the self-reported Quick Inventory of Depressive Symptomatology (16-item). RESULTS: We corroborated our pre-registered hypothesis (NCT04342299) that lower bilateral amygdala activation for sad v. happy faces predicted favorable clinical outcomes (rs[38] = 0.40, p = 0.01). In contrast, there was no effect for dorsal/pregenual anterior cingulate cortex activation (rs[38] = 0.18, p = 0.29), nor when using voxel-based whole-brain analyses (voxel-based Family-Wise Error-corrected p < 0.05). Predictive effects were mainly driven by the right amygdala whose response to happy faces was reduced in patients with higher anxiety levels. CONCLUSIONS: We confirmed the prediction that a lower amygdala response to negative v. positive facial expressions might be an adaptive neural signature, which predicts subsequent symptom improvement also in difficult-to-treat MDD. Anxiety reduced adaptive amygdala responses.

Childhood adversities characterize the heterogeneity in the brain pattern of individuals during neurodevelopment.

BACKGROUND: Several factors shape the neurodevelopmental trajectory. A key area of focus in neurodevelopmental research is to estimate the factors that have maximal influence on the brain and can tip the balance from typical to atypical development. METHODS: Utilizing a dissimilarity maximization algorithm on the dynamic mode decomposition (DMD) of the resting state functional MRI data, we classified subjects from the cVEDA neurodevelopmental cohort (n = 987, aged 6-23 years) into homogeneously patterned DMD (representing typical development in 809 subjects) and heterogeneously patterned DMD (indicative of atypical development in 178 subjects). RESULTS: Significant DMD differences were primarily identified in the default mode network (DMN) regions across these groups (p < 0.05, Bonferroni corrected). While the groups were comparable in cognitive performance, the atypical group had more frequent exposure to adversities and faced higher abuses (p < 0.05, Bonferroni corrected). Upon evaluating brain-behavior correlations, we found that correlation patterns between adversity and DMN dynamic modes exhibited age-dependent variations for atypical subjects, hinting at differential utilization of the DMN due to chronic adversities. CONCLUSION: Adversities (particularly abuse) maximally influence the DMN during neurodevelopment and lead to the failure in the development of a coherent DMN system. While DMN's integrity is preserved in typical development, the age-dependent variability in atypically developing individuals is contrasting. The flexibility of DMN might be a compensatory mechanism to protect an individual in an abusive environment. However, such adaptability might deprive the neural system of the faculties of normal functioning and may incur long-term effects on the psyche.

The Role of Subgenual Resting-State Connectivity Networks in Predicting Prognosis in Major Depressive Disorder.

Background: A seminal study found higher subgenual frontal cortex resting-state connectivity with 2 left ventral frontal regions and the dorsal midbrain to predict better response to psychotherapy versus medication in individuals with treatment-naïve major depressive disorder (MDD). Here, we examined whether these subgenual networks also play a role in the pathophysiology of clinical outcomes in MDD with early treatment resistance in primary care. Methods: Forty-five people with current MDD who had not responded to ≥2 serotonergic antidepressants (n = 43, meeting predefined functional magnetic resonance imaging minimum quality thresholds) were enrolled and followed over 4 months of standard care. Functional magnetic resonance imaging resting-state connectivity between the preregistered subgenual frontal cortex seed and 3 previously identified left ventromedial, ventrolateral prefrontal/insula, and dorsal midbrain regions was extracted. The clinical outcome was the percentage change on the self-reported 16-item Quick Inventory of Depressive Symptomatology. Results: We observed a reversal of our preregistered hypothesis in that higher resting-state connectivity between the subgenual cortex and the a priori ventrolateral prefrontal/insula region predicted favorable rather than unfavorable clinical outcomes (rs39 = -0.43, p = .006). This generalized to the sample including participants with suboptimal functional magnetic resonance imaging quality (rs43 = -0.35, p = .02). In contrast, no effects (rs39 = 0.12, rs39 = -0.01) were found for connectivity with the other 2 preregistered regions or in a whole-brain analysis (voxel-based familywise error-corrected p < .05). Conclusions: Subgenual connectivity with the ventrolateral prefrontal cortex/insula is relevant for subsequent clinical outcomes in current MDD with early treatment resistance. Its positive association with favorable outcomes could be explained primarily by psychosocial rather than the expected pharmacological changes during the follow-up period.

Evidence has shown that connectivity of the subgenual cortex, a frontal midline brain region, with 3 other brain regions can predict whether people with never-treated depression benefit more from psychological or medication-based treatments. Here, using resting-state fMRI, we show that subgenual connections with one of these regions, the left ventrolateral prefrontal/insula, also predict future average depression levels in people with difficult-to-treat depression. These data suggest that functional connectivity between these regions may be linked to clinical outcomes in major depressive disorder.

The relationship between negative life events and cortical structural connectivity in adolescents.

Adolescence is a crucial period for physical and psychological development. The impact of negative life events represents a risk factor for the onset of neuropsychiatric disorders. This study aims to investigate the relationship between negative life events and structural brain connectivity, considering both graph theory and connectivity strength. A group (n = 487) of adolescents from the IMAGEN Consortium was divided into Low and High Stress groups. Brain networks were extracted at an individual level, based on morphological similarity between grey matter regions with regions defined using an atlas-based region of interest (ROI) approach. Between-group comparisons were performed with global and local graph theory measures in a range of sparsity levels. The analysis was also performed in a larger sample of adolescents (n = 976) to examine linear correlations between stress level and network measures. Connectivity strength differences were investigated with network-based statistics. Negative life events were not found to be a factor influencing global network measures at any sparsity level. At local network level, between-group differences were found in centrality measures of the left somato-motor network (a decrease of betweenness centrality was seen at sparsity 5%), of the bilateral central visual and the left dorsal attention network (increase of degree at sparsity 10% at sparsity 30% respectively). Network-based statistics analysis showed an increase in connectivity strength in the High stress group in edges connecting the dorsal attention, limbic and salience networks. This study suggests negative life events alone do not alter structural connectivity globally, but they are associated to connectivity properties in areas involved in emotion and attention.

Unraveling robust brain-behavior links of depressive complaints through granular network models for understanding heterogeneity.

BACKGROUND: Depressive symptoms are highly prevalent, present in heterogeneous symptom patterns, and share diverse neurobiological underpinnings. Understanding the links between psychopathological symptoms and biological factors is critical in elucidating its etiology and persistence. We aimed to evaluate the utility of using symptom-brain network models to parse the heterogeneity of depressive complaints in a large adolescent sample. METHODS: We used data from the third wave of the IMAGEN study, a multi-center panel cohort study involving 1317 adolescents (52.49 % female, mean ± SD age = 18.5 ± 0.7). Two network models were estimated: one including an overall depressive symptom severity sum score based on the Adolescent Depression Rating Scale (ADRS), and one incorporating individual ADRS item scores. Both networks included measures of cortical thickness in several regions (insula, cingulate, mOFC, fusiform gyrus) and hippocampal volume derived from neuroimaging. RESULTS: The network based on individual item scores revealed associations between cortical thickness measures and specific depressive complaints, obscured when using an aggregate depression severity score. Notably, the insula's cortical thickness showed negative associations with cognitive dysfunction (partial cor. = -0.15); the cingulate's cortical thickness showed negative associations with feelings of worthlessness (partial cor. = -0.10), and mOFC was negatively associated with anhedonia (partial cor. = -0.05). LIMITATIONS: This cross-sectional study relied on the self-reported assessment of depression complaints and used a non-clinical sample with predominantly healthy participants (19 % with depression or sub-threshold depression). CONCLUSIONS: This study showcases the utility of network models in parsing heterogeneity in depressive complaints, linking individual complaints to specific neural substrates. We outline the next steps to integrate neurobiological and cognitive markers to unravel MDD's phenotypic heterogeneity.

The genetic architecture of the human hypothalamus and its involvement in neuropsychiatric behaviours and disorders.

Despite its crucial role in the regulation of vital metabolic and neurological functions, the genetic architecture of the hypothalamus remains unknown. Here we conducted multivariate genome-wide association studies (GWAS) using hypothalamic imaging data from 32,956 individuals to uncover the genetic underpinnings of the hypothalamus and its involvement in neuropsychiatric traits. There were 23 significant loci associated with the whole hypothalamus and its subunits, with functional enrichment for genes involved in intracellular trafficking systems and metabolic processes of steroid-related compounds. The hypothalamus exhibited substantial genetic associations with limbic system structures and neuropsychiatric traits including chronotype, risky behaviour, cognition, satiety and sympathetic-parasympathetic activity. The strongest signal in the primary GWAS, the ADAMTS8 locus, was replicated in three independent datasets (N = 1,685-4,321) and was strengthened after meta-analysis. Exome-wide association analyses added evidence to the association for ADAMTS8, and Mendelian randomization showed lower ADAMTS8 expression with larger hypothalamic volumes. The current study advances our understanding of complex structure-function relationships of the hypothalamus and provides insights into the molecular mechanisms that underlie hypothalamic formation.

Machine learning models for diagnosis and risk prediction in eating disorders, depression, and alcohol use disorder.

This study uses machine learning models to uncover diagnostic and risk prediction markers for eating disorders (EDs), major depressive disorder (MDD), and alcohol use disorder (AUD). Utilizing case-control samples (ages 18-25 years) and a longitudinal population-based sample (n=1,851), the models, incorporating diverse data domains, achieved high accuracy in classifying EDs, MDD, and AUD from healthy controls. The area under the receiver operating characteristic curves (AUC-ROC [95% CI]) reached 0.92 [0.86-0.97] for AN and 0.91 [0.85-0.96] for BN, without relying on body mass index as a predictor. The classification accuracies for MDD (0.91 [0.88-0.94]) and AUD (0.80 [0.74-0.85]) were also high. Each data domain emerged as accurate classifiers individually, with personality distinguishing AN, BN, and their controls with AUC-ROCs ranging from 0.77 to 0.89. The models demonstrated high transdiagnostic potential, as those trained for EDs were also accurate in classifying AUD and MDD from healthy controls, and vice versa (AUC-ROCs, 0.75-0.93). Shared predictors, such as neuroticism, hopelessness, and symptoms of attention-deficit/hyperactivity disorder, were identified as reliable classifiers. For risk prediction in the longitudinal population sample, the models exhibited moderate performance (AUC-ROCs, 0.64-0.71), highlighting the potential of combining multi-domain data for precise diagnostic and risk prediction applications in psychiatry.

Adolescent maturation of cortical excitation-inhibition balance based on individualized biophysical network modeling.

The balance of excitation and inhibition is a key functional property of cortical microcircuits which changes through the lifespan. Adolescence is considered a crucial period for the maturation of excitation-inhibition balance. This has been primarily observed in animal studies, yet human in vivo evidence on adolescent maturation of the excitation-inhibition balance at the individual level is limited. Here, we developed an individualized in vivo marker of regional excitation-inhibition balance in human adolescents, estimated using large-scale simulations of biophysical network models fitted to resting-state functional magnetic resonance imaging data from two independent cross-sectional (N = 752) and longitudinal (N = 149) cohorts. We found a widespread relative increase of inhibition in association cortices paralleled by a relative age-related increase of excitation, or lack of change, in sensorimotor areas across both datasets. This developmental pattern co-aligned with multiscale markers of sensorimotor-association differentiation. The spatial pattern of excitation-inhibition development in adolescence was robust to inter-individual variability of structural connectomes and modeling configurations. Notably, we found that alternative simulation-based markers of excitation-inhibition balance show a variable sensitivity to maturational change. Taken together, our study highlights an increase of inhibition during adolescence in association areas using cross sectional and longitudinal data, and provides a robust computational framework to estimate microcircuit maturation in vivo at the individual level.

Investigating grey matter volumetric trajectories through the lifespan at the individual level.

Adolescents exhibit remarkable heterogeneity in the structural architecture of brain development. However, due to limited large-scale longitudinal neuroimaging studies, existing research has largely focused on population averages, and the neurobiological basis underlying individual heterogeneity remains poorly understood. Here we identify, using the IMAGEN adolescent cohort followed up over 9 years (14-23 y), three groups of adolescents characterized by distinct developmental patterns of whole-brain gray matter volume (GMV). Group 1 show continuously decreasing GMV associated with higher neurocognitive performances than the other two groups during adolescence. Group 2 exhibit a slower rate of GMV decrease and lower neurocognitive performances compared with Group 1, which was associated with epigenetic differences and greater environmental burden. Group 3 show increasing GMV and lower baseline neurocognitive performances due to a genetic variation. Using the UK Biobank, we show these differences may be attenuated in mid-to-late adulthood. Our study reveals clusters of adolescent neurodevelopment based on GMV and the potential long-term impact.