Poster No:
935
Submission Type:
Abstract Submission
Authors:
Richard Dinga1, Seyed Kia1, Marijn van Wingerden1
Institutions:
1Tilburg University, Tilburg, North Brabant
First Author:
Co-Author(s):
Introduction:
Age plays a prominent role in the etiology of many psychiatric and neurological disorders. Different disorders are characterized by different ages of onset (Chen, 2016). Understanding healthy and pathological aging is, therefore, critical for understanding the development of psychiatric and neurological disorders. In recent years, deviations from healthy brain aging have been connected to numerous psychiatric and neurological disorders, and many methods aimed to detect and characterize such deviations have emerged, most prominently brain age and normative modeling. Brain age (Franke, 2019) uses machine learning to predict a person's age from their brain imaging data and calculates the difference between a person's predicted and true age, and this difference is used as a biomarker of healthy aging. Normative modeling (Marquand, 2019), on the other hand, calculates the expected ranges of brain variables in a population for a given age and calculates how much a person deviates from that norm. Although both approaches are supported by a rich body of research providing theoretical and empirical evidence for their utility as biomarkers of healthy aging, the relative effectiveness of these approaches in identifying individuals with neurological or psychiatric disorders has not been established. To fill this gap, in this study, we compare brain-age models and various brain abnormality measures derived from normative models in their ability to detect subjects with a clinical diagnosis of schizophrenia (SZ), attention deficit hyperactivity disorder (ADHD), bipolar disorder (BP), mild cognitive impairment (MCI), and Alzheimer's disease (AD).
Methods:
We used region-of-interest (ROI) cortical thickness data from publicly available datasets: OASIS3 (LaMontagne, 2019), CAMCAN (Taylor, 2017), and CNP (Poldrack, 2016). The training set included all healthy subjects from scan sites without cases and half of the healthy subjects from sites with cases; the remaining healthy subjects and all cases were included in the test set. Sex-specific brain-age models were estimated using XGBoost (Chen, 2016), and age-adjusted brain-age gaps were calculated for the test set. Normative distributions for specific region of interest (ROI) given age and sex were estimated using GAMLSS (Stasinopoulos, 2012) normative models with SHASH distribution. Whole-brain abnormality measures included the number of ROIs with z-scores exceeding ±2, minimum and maximum z-scores, and mean and mean absolute z-scores. We computed AUC-ROC scores to evaluate these measures for detecting specific diagnoses. Site effects had been accounted for by data-harmonization methods (Fortin, 2018) for brain age and by explicitly including side effects in the model for normative modeling (Dinga, 2021). Statistical significance was assessed using a permutation test, which also accounted for the multiplicity caused by selecting the best abnormality score.
Results:
The dataset consists of n=2494 healthy subjects and 461 subjects with a psychiatric or neurological disorder (271 AD, 51 MCI, 49 schizophrenia, 49 bipolar disorder, 41 ADHD). For each diagnosis, we compared the best-performing abnormality measure to the brain-age gap adjusted for the optimism selection bias. Normative modeling statistically significantly outperformed brain age in the detection of schizophrenia (AUC 0.7 and 0.55, p < 0.001) but underperformed for the detection of ADHD (AUC 0.57 and 0.67, p<0.001) and bipolar disorder (AUC 0.63 and 0.72, p<0.001). The differences were not statistically significant for MCI (AUC 0.62 and 0.64, NS) and AD (AUC 0.75 and 0.72, NS).
Conclusions:
No single measure was uniformly better across all diagnoses. Brain age and normative modeling are complementary tools that capture different types of clinically relevant brain abnormalities, making neither sufficient as a universal biomarker of healthy aging. Both approaches should be considered when designing biomarkers for healthy aging.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s)
Neurodevelopmental/ Early Life (eg. ADHD, autism)
Psychiatric (eg. Depression, Anxiety, Schizophrenia)
Lifespan Development:
Aging 1
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
Keywords:
Aging
Attention Deficit Disorder
Machine Learning
Modeling
MRI
Schizophrenia
Other - normative modeling
1|2Indicates the priority used for review
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
I accept
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information.
Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
I do not want to participate in the reproducibility challenge.
Please indicate below if your study was a "resting state" or "task-activation” study.
Other
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Patients
Was this research conducted in the United States?
No
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel?
NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Not applicable
Were any animal research approved by the relevant IACUC or other animal research panel?
NOTE: Any animal studies without IACUC approval will be automatically rejected.
Not applicable
Please indicate which methods were used in your research:
Structural MRI
Computational modeling
For human MRI, what field strength scanner do you use?
3.0T
Which processing packages did you use for your study?
Free Surfer
Provide references using APA citation style.
Chen, T., (2016). XGBoost. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794. https://doi.org/10.1145/2939672.2939785
Dinga, R., (2021). Normative Modeling of Neuroimaging Data Using Generalized Additive Models of Location Scale and Shape. https://doi.org/10.1101/2021.06.14.448106
Franke, K., (2019). Ten years of brainage as a neuroimaging biomarker of brain aging: What insights have we gained? Frontiers in Neurology, 10. https://doi.org/10.3389/fneur.2019.00789
LaMontagne, P. J., (2019). Oasis-3: Longitudinal Neuroimaging, Clinical, and Cognitive Dataset for Normal Aging and Alzheimer Disease. https://doi.org/10.1101/2019.12.13.19014902
Marquand, A. F., (2019). Conceptualizing mental disorders as deviations from normative functioning. Molecular Psychiatry, 24(10), 1415–1424. https://doi.org/10.1038/s41380-019-0441-1
Poldrack, R. A., (2016). A phenome-wide examination of neural and cognitive function. Scientific Data, 3(1). https://doi.org/10.1038/sdata.2016.110
Solmi, M., (2021). Age at onset of mental disorders worldwide: Large-scale meta-analysis of 192 epidemiological studies. Molecular Psychiatry, 27(1), 281–295. https://doi.org/10.1038/s41380-021-01161-7
Stasinopoulos, M., (2012). GAMLSS: Generalized additive models for location scale and shape. CRAN: Contributed Packages. https://doi.org/10.32614/cran.package.gamlss
Taylor, J. R., (2017). The Cambridge Centre for Ageing and Neuroscience (cam-CAN) data repository: Structural and functional MRI, MEG, and cognitive data from a cross-sectional adult lifespan sample. NeuroImage, 144, 262–269. https://doi.org/10.1016/j.neuroimage.
No