Maturity of gray matter structures and white matter connectomes, and their relationship with psychiatric symptoms in youth

Abstract

Brain predicted age difference, or BrainPAD, compares chronological age to an age estimate derived by applying machine learning (ML) to MRI brain data. BrainPAD studies in youth have been relatively limited, often using only a single MRI modality or a single ML algorithm. Here, we use multimodal MRI with a stacked ensemble ML approach that iteratively applies several ML algorithms (AutoML). Eligible participants in the Healthy Brain Network (N = 489) were split into training and test sets. Morphometry estimates, white matter connectomes, or both were entered into AutoML to develop BrainPAD models. The best model was then applied to a held‐out evaluation dataset, and associations with psychometrics were estimated. Models using morphometry and connectomes together had a mean absolute error of 1.18 years, outperforming models using a single MRI modality. Lower BrainPAD values were associated with more symptoms on the CBCL (pcorr = .012) and lower functioning on the Children's Global Assessment Scale (pcorr = .012). Higher BrainPAD values were associated with better performance on the Flanker task (pcorr = .008). Brain age prediction was more accurate using ComBat‐harmonized brain data (MAE = 0.26). Associations with psychometric measures remained consistent after ComBat harmonization, though only the association with CGAS reached statistical significance in the reduced sample. Our findings suggest that BrainPAD scores derived from unharmonized multimodal MRI data using an ensemble ML approach may offer a clinically relevant indicator of psychiatric and cognitive functioning in youth.

Brain age prediction using fMRI network coupling in youths and associations with psychiatric symptoms

“…The significant association between general psychopathology and BAG in the PNC sample, where a higher symptom burden was associated with a lower BAG, is in accordance with studies showing a delay in brain maturation being linked to poorer mental health in the same sample ( Kaufmann et al, 2017 ) and also in young patients with schizophrenia ( Douaud et al, 2009 ). In addition, studies utilizing morphometry and diffusion data for BAG prediction have found that HBN participants that had a lower BAG also had more symptoms on the Child Behavior Checklist (CBCL), and decreased global functioning ( Luna et al, 2021 ). While work in youths’ has pointed to a lower BAG with increased symptoms (interpreted as a delayed development), elevated structural brain age (interpreted as apparent aging) have also been shown for frontal and subcortical regions in PNC individuals with symptoms of psychosis and obsessive–compulsive disorder ( Cropley et al, 2020 ).…”

Automated Multiclass Artifact Detection in Diffusion MRI Volumes via 3D Residual Squeeze-and-Excitation Convolutional Neural Networks

“…The 4,226 manually labeled poor-quality dMRI volumes of the HBN dataset (details of this dataset are presented in the data and pre-processing section) were derived from 66 randomly selected subjects from a larger pool of 100 participants whose age varied from 5.57 to 21.89 years with mean and median age of 10.95 and 10.05 years, respectively. The HBN dataset was selected because the age range was appropriate for the desired age-based analysis, and because many of the MRI scans were known to be corrupted with artifacts based on our prior experience with this dataset ( Luna et al, 2021 ). Prior research has demonstrated a linear relationship between participant age and FA ( Rathee et al, 2016 ; Richie-Halford et al, 2021 ).…”

Systematic Evaluation of Machine Learning Algorithms for Neuroanatomically-Based Age Prediction in Youth

“…In adults, higher brain-age relative to chronological age (i.e., higher BrainAGE) has been associated with adverse physical (Cole et al, 2018), cognitive (Anaturk et al, 2021; Boyle et al, 2021; Elliott et al, 2019) and mental health phenotypes (Kaufmann et al, 2019; Lee et al, 2021). By contrast, in children and adolescents higher BrainAGE has been associated with better cognitive test performance (Boyle et al, 2021; Erus et al, 2015; Luna et al, 2021) while associations with clinical phenotypes show a more complex pattern which may depend on the nature of the phenotype and/or the developmental stage of the sample (Chung et al, 2018; Luna et al, 2021). These findings underscore the importance of accuracy in brain-based age-prediction in youth, as childhood and adolescence are arguably periods of the most dynamic brain re-organization.…”

“…Machine learning (ML) algorithms applied to sMRI data can harness the multidimensional nature of age-related brain changes at the individual-level to predict age, as a proxy for the biological age of the brain (i.e., brain-age). The difference between brain-age and chronological age is referred to here as brain-age-gap-estimation (BrainAGE)(Franke and Gaser, 2019), which is equivalent to terms such as brain-predicted-age-difference (brainPAD) (Luna et al, 2021), brain-age-gap (BAG) (Anaturk et al, 2021) and brain-age delta(Beheshti et al, 2019) used in other studies. In adults, higher brain-age relative to chronological age (i.e., higher BrainAGE) has been associated with adverse physical (Cole et al, 2018), cognitive (Anaturk et al, 2021; Boyle et al, 2021; Elliott et al, 2019) and mental health phenotypes (Kaufmann et al, 2019; Lee et al, 2021).…”

“…We have previously shown that age prediction from sMRI data in adults is influenced by the choice of algorithm (Lee et al, 2021). Here addressed this knowledge gap in youth because with few exceptions (Ball et al, 2021; Brouwer et al, 2021; Lee et al, 2021; Luna et al, 2021), studies on brain-age prediction in this population have typically employed a single ML algorithm, most commonly relevance vector regression (RVR), Gaussian process regression (GPR), and support vector regression (SVR) (Cole et al, 2018; Franke et al, 2012; Franke et al, 2010; Gaser et al, 2013; Liem et al, 2017; Valizadeh et al, 2017). Here we systematically evaluated the performance of 21 ML algorithms applied to sMRI data from youth from five different cohorts and then tested their performance in two independent samples.…”