Brain Aging Differences across 10 Brain Disorders by Brain Age Prediction
Poster No:
1138
Submission Type:
Abstract Submission
Authors:
Chuang Liang1, Godfrey Pearlson2, Juan R. Bustillo3, Peter Kochunov4, Jessica A. Turner5, Wei Shao1, Rongtao Jiang6, Jing Sui7, Daoqiang Zhang1, Zening Fu5, Shile Qi1, Vince Calhoun5
Institutions:
1Nanjing University of Aeronautics and Astronautics, Nanjing, China, 2Yale School of Medicine, New Haven, CT, 3University of New Mexico, Albuquerque, NM, 4University of Maryland School of Medicine, Baltimore, MD, 5Georgia State University, Atlanta, GA, 6Yale University, New Haven, CT, 7Beijing Normal University, Beijing, China
First Author:
Co-Author(s):
Introduction:
Brain age can be used as a brain health index to quantify individuals' deviation from a normative brain aging trajectory[1-3]. The difference between neuroimaging-predicted brain age and the chronological age, predicted age difference (PAD), is a potential biomarker that reflects individual differences in brain developmental trajectories[4, 5]. Risk factors for different brain disorders may converge on neurological aspects associated with accelerated brain aging processes[6, 7]. However, most existing studies have been conducted on a limited sample-size that focused on a specific disorder without identifying interpretable neuroimaging features contributing to PAD prediction. Further understanding of brain aging trajectories and their spatial patterns among various brain disorders is important to disentangle disorder common and unique pathophysiological processes.
Methods:
A total of 2752 independent patients, including attention-deficit/hyperactivity disorders (ADHD, n=344), autism spectrum disorder (ASD, n=484), schizophrenia (SZ, n=152), bipolar (BP, n=143), major depressive disorder (MDD, n=258), drinking (DRK, n=155), smoking (SMK, n=144), drinking and smoking (D&S, n=54), Alzheimer's disease (AD, n=361), mild cognitive impairment (MCI, n=657) and 2752 age-gender-quantity matched healthy controls (HCs) were collected as testing sets. Age-range matched HCs (n=53149) for each patient group were selected from 4 consortiums (including ABCD, HCP, GSP and UKB) as the corresponding training sets. Averaged gray matter volume (GMV) for each region of interest (ROI) based on Schaefer (1000 ROIs) and subcortex (16 ROIs) brain atlas[8] were used as features (Fig. 1a) in brain age prediction. Extreme gradient boosting (XGBoost) was trained on the selected age-range matched HCs, which was applied to the corresponding patient group to generate the individual brain age predictions. The accuracy of the XGBoost model[6] on training set was estimated by 10-fold cross-validation (Fig. 1b). The PAD differences (covariates: age, age2, sex and site were regressed) between HC and each patient group were conducted (Fig. 1c). Shapley additive explanations (SHAP)[9] was used to identify the interpretable brain features for PAD prediction in different diagnostic groups (Fig. 1d).
Results:
(a) The high degree of correlations between chronological age and the predicted brain age in both training (r=0.67~0.94) and independent testing sets (r=0.73~0.91) confirmed the excellent prediction performance (Fig. 2a). (b) The PAD was higher in all the diagnostic groups than HCs, in which strong difference effects of PAD were observed in AD and D&S (d=0.84~0.97), moderate effects in SMK, DRK, SZ, BP and MCI (d=0.45~0.72), small effects in MDD (d=0.28), and negligible effects in ASD and ADHD (d=0.02~0.06, Fig. 2b). (3) The brain aging biomarkers that contribute the most in the PAD prediction for each diagnostic groups were: medial prefrontal cortex (mPFC)-lentiform nucleus (LN)-thalamus-amygdala in psychiatric disorders (SZ, BP and MDD), mPFC-postcentral cortex (PC)-thalamus in addiction disorders (DRK, SMK and D&S) and mPFC-superior temporal cortex-striatum-hippocampus in neurodegenerative (AD and MCI) disorders (Fig. 2c).
Conclusions:
This is the first attempt to estimate brain age and its biological interpretation across 10 brain disorders in large scale cohorts. Results suggest that different brain disorders showed different brain aging, the highest aging in dementia, followed by addiction and psychiatry, (AD>D&S>SMK> DRK>SZ>BP>MCI>MDD). PAD differences were not observed in ASD and ADHD. Furthermore, the PAD in each patient group was driven by a unique spatial brain pattern, but also with some shared regions within the psychiatry (mPFC), addiction (mPFC and PC) and dementia (striatum). In summary, the observed different brain aging, each with unique interpretable patterns, may serve as neuroimaging biomarkers for understanding the aging neural mechanisms of various brain disorders.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s)
Neurodevelopmental/ Early Life (eg. ADHD, autism)
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 2
Lifespan Development:
Aging 1
Keywords:
Aging
STRUCTURAL MRI
Other - Brain age prediction; brain aging; brain disorders; predicted age difference; interpretable brain patterns.
1|2Indicates the priority used for review
·Figure 1. Flowchart of the study design.
·Figure 2. (a) The Pearson correlation between chronological age and the predicted brain age. (b) The PAD differences between HC and specific patient group and (c) its interpretable brain regions.
Provide references using author date format
2. Cole, J.H., et al., Brain age predicts mortality. Molecular psychiatry, 2018. 23(5): p. 1385-1392.
3. Bethlehem, R.A., et al., Brain charts for the human lifespan. Nature, 2022. 604(7906): p. 525-533.
4. Cole, J.H., et al., Predicting brain age with deep learning from raw imaging data results in a reliable and heritable biomarker. NeuroImage, 2017. 163: p. 115-124.
5. Cole, J.H. and K. Franke, Predicting age using neuroimaging: innovative brain ageing biomarkers. Trends in neurosciences, 2017. 40(12): p. 681-690.
6. Kaufmann, T., et al., Common brain disorders are associated with heritable patterns of apparent aging of the brain. Nature neuroscience, 2019. 22(10): p. 1617-1623.
7. Wang, J., et al., Gray matter age prediction as a biomarker for risk of dementia. Proceedings of the National Academy of Sciences, 2019. 116(42): p. 21213-21218.
8. Lawrence, R.M., et al., Standardizing human brain parcellations. Scientific data, 2021. 8(1): p. 78.
9. Ballester, P.L., et al., Gray matter volume drives the brain age gap in schizophrenia: a SHAP study. Schizophrenia, 2023. 9(1): p. 3.