| 1. |
|
|
| 2. |
|
|
| 3. |
- Ge, R, et al.
(författare)
-
Normative Modeling of Brain Morphometry Across the Lifespan Using CentileBrain: Algorithm Benchmarking and Model Optimization
- 2023
-
Ingår i: bioRxiv : the preprint server for biology. - : Cold Spring Harbor Laboratory.
-
Tidskriftsartikel (övrigt vetenskapligt/konstnärligt)abstract
- Background: Normative modeling is a statistical approach to quantify the degree to which a particular individual-level measure deviates from the pattern observed in a normative reference population. When applied to human brain morphometric measures it has the potential to inform about the significance of normative deviations for health and disease. Normative models can be implemented using a variety of algorithms that have not been systematically appraised. Methods: To address this gap, eight algorithms were compared in terms of performance and computational efficiency using brain regional morphometric data from 37,407 healthy individuals (53% female; aged 3-90 years) collated from 87 international MRI datasets. Performance was assessed with the mean absolute error (MAE) and computational efficiency was inferred from central processing unit (CPU) time. The algorithms evaluated were Ordinary Least Squares Regression (OLSR), Bayesian Linear Regression (BLR), Generalized Additive Models for Location, Scale, and Shape (GAMLSS), Parametric Lambda, Mu, Sigma (LMS), Gaussian Process Regression (GPR), Warped Bayesian Linear Regression (WBLG), Hierarchical Bayesian Regression (HBR), and Multivariable Fractional Polynomial Regression (MFPR). Model optimization involved testing nine covariate combinations pertaining to acquisition features, parcellation software versions, and global neuroimaging measures (i.e., total intracranial volume, mean cortical thickness, and mean cortical surface area). Findings: Statistical comparisons across models at PFDR<0.05 indicated that the MFPR-derived sex- and region-specific models with nonlinear polynomials for age and linear effects of global measures had superior predictive accuracy; the range of the MAE of the models of regional subcortical volumes was 70-520 mm3 and the corresponding ranges for regional cortical thickness and regional cortical surface area were 0.09-0.26 mm and 24-560 mm2, respectively. The MFPR-derived models were also computationally more efficient with a CPU time below one second compared to a range of 2 seconds to 60 minutes for the other algorithms. The performance of all sex- and region-specific MFPR models plateaued at sample sizes exceeding 3,000 and showed comparable MAEs across distinct 10-year age-bins covering the human lifespan. Interpretation: These results provide an empirically benchmarked framework for normative modeling of brain morphometry that is useful for interpreting prior literature and supporting future study designs. The model and tools described here are freely available through CentileBrain (https://centilebrain.org/), a user-friendly web platform.
|
|
| 4. |
|
|
| 5. |
|
|
| 6. |
- Nunes, A, et al.
(författare)
-
Using structural MRI to identify bipolar disorders - 13 site machine learning study in 3020 individuals from the ENIGMA Bipolar Disorders Working Group
- 2020
-
Ingår i: Molecular psychiatry. - : Springer Science and Business Media LLC. - 1476-5578 .- 1359-4184. ; 25:9, s. 2130-2143
-
Tidskriftsartikel (refereegranskat)abstract
- Bipolar disorders (BDs) are among the leading causes of morbidity and disability. Objective biological markers, such as those based on brain imaging, could aid in clinical management of BD. Machine learning (ML) brings neuroimaging analyses to individual subject level and may potentially allow for their diagnostic use. However, fair and optimal application of ML requires large, multi-site datasets. We applied ML (support vector machines) to MRI data (regional cortical thickness, surface area, subcortical volumes) from 853 BD and 2167 control participants from 13 cohorts in the ENIGMA consortium. We attempted to differentiate BD from control participants, investigated different data handling strategies and studied the neuroimaging/clinical features most important for classification. Individual site accuracies ranged from 45.23% to 81.07%. Aggregate subject-level analyses yielded the highest accuracy (65.23%, 95% CI = 63.47–67.00, ROC-AUC = 71.49%, 95% CI = 69.39–73.59), followed by leave-one-site-out cross-validation (accuracy = 58.67%, 95% CI = 56.70–60.63). Meta-analysis of individual site accuracies did not provide above chance results. There was substantial agreement between the regions that contributed to identification of BD participants in the best performing site and in the aggregate dataset (Cohen’s Kappa = 0.83, 95% CI = 0.829–0.831). Treatment with anticonvulsants and age were associated with greater odds of correct classification. Although short of the 80% clinically relevant accuracy threshold, the results are promising and provide a fair and realistic estimate of classification performance, which can be achieved in a large, ecologically valid, multi-site sample of BD participants based on regional neurostructural measures. Furthermore, the significant classification in different samples was based on plausible and similar neuroanatomical features. Future multi-site studies should move towards sharing of raw/voxelwise neuroimaging data.
|
|
