Accommodating site variation in neuroimaging data using normative and hierarchical Bayesian models

Bayer, Johanna; Dinga, Richard; Kia, Seyed Mostafa; Kottaram, Akhil; Wolfers, Thomas; Lv, Jinglei; Zalesky, Andrew; Schmaal, Lianne; Marquand, André F.

doi:10.1016/j.neuroimage.2022.119699

Cited by 37 publications

(42 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The most common methods for minimizing site effects involve either site harmonization using ComBat-GAM (Pomponio et al, 2020) prior to normative modeling or the inclusion of site as an explanatory variable in the normative models. A recent publication that used a smaller sample (569 healthy participants) and a narrower age range (6 to 40 years) suggested the HBR with site as an explanatory variable may be superior to ComBat-based site harmonization for the normative modeling of brain morphometry (Bayer et al, 2022). We found no support for this assertion, as in this study MFPR-derived models using Combat-GAM for site harmonization outperformed HBR-derived models with site as a random effect.…”

Section: Discussioncontrasting

confidence: 88%

“…Prior studies have shown that sex accounts for a significant amount of variance in brain morphology (Potvin et al, 2016;Potvin et al, 2017), both cross-sectionally (Ge et al, 2021;Kaczkurkin et al, 2019) and with regard to age-related changes (Ching et al, 2022;Forde et al, 2020;Wierenga et al, 2022). In view of that, we developed sex-specific models for each brain morphometric measure thus extending prior normative studies that considered males and females together (Bayer et al, 2022;Rutherford et al, 2022a). Additionally, we provide normative models for regional cortical surface area measures; changes in the cortical surface area have important functional implications for cognition during development (Cafiero et al, 2019;Wierenga et al, 2014;Vijayakumar et al, 2016) and aging (Lemaitre et al, 2012) but have been overlooked in prior normative studies (Bethlehem et al, 2022;Rutherford et al, 2022a).…”

Section: Discussionmentioning

confidence: 99%

“…(7) Warped Bayesian Linear Regression (WBLR) (Fraza et al, 2021): this framework is based on Bayesian linear regression with likelihood warping and was implemented using the "PCNtoolkit" in Python with default settings (Bayer et al, 2022);…”

Section: Ii) Algorithm Selectionmentioning

confidence: 99%

“…For the HBR, WBLR, GPR, and GAMLSS, site effects were modeled as random effects as recommended by prior literature (Dinga et al, 2021;Bayer et al, 2022;Rutherford et al, 2022b).…”

Section: Ii) Algorithm Selectionmentioning

confidence: 99%

“…At present, there is no systematic comparative evaluation of the performance of these algorithms and no empirical determination of the minimum sample size necessary for reliable normative estimates of brain morphometric measures. With few exceptions (Bethlehem et al, 2022;Dima et al, 2022;Frangou et al, 2022;Lawrence et al, 2021;Rutherford et al, 2022a), the size of the sample used for the reference population is small to modest (range: 145-870) (Wolfers et al, 2018;Marquand et al, 2019;Zabihi et al, 2019;Bethlehem et al, 2020;Wolfers et al, 2020;Zabihi et al, 2020;Lv et al, 2021;Bayer et al, 2022).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Normative Modeling of Brain Morphometry Across the Lifespan Using CentileBrain: Algorithm Benchmarking and Model Optimization

et al. 2023

Preprint

View full text Add to dashboard Cite

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. 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 (53.33% female; aged 3 to 90 years) collated from 86 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), Multivariable Fractional Polynomial Regression (MFPR), Warped Bayesian Linear Regression (WBLG), and Hierarchical Bayesian Regression (HBR). Model optimization involved testing 9 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). Statistical comparisons across models at PFDR<0.05 indicated that MFPR-derived sex and region-specific models with nonlinear polynomials for age and linear effect of global measures had superior predictive accuracy; for models of regional subcortical volume the MAE range was 70-80 mm3 and the corresponding range for regional cortical thickness and regional cortical surface area was 0.09-0.26 mm; and 24-660 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. MFPR-derived models were robust across distinct age groups when sample sizes exceeded 3,000. These results support CentileBrain (https://centilebrain.org/) an empirically benchmarked framework for normative modeling that is freely available to the scientific community.

show abstract

Section: Discussioncontrasting

confidence: 88%

Section: Discussionmentioning

confidence: 99%

Section: Ii) Algorithm Selectionmentioning

confidence: 99%

“…For the HBR, WBLR, GPR, and GAMLSS, site effects were modeled as random effects as recommended by prior literature (Dinga et al, 2021;Bayer et al, 2022;Rutherford et al, 2022b).…”

Section: Ii) Algorithm Selectionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Normative Modeling of Brain Morphometry Across the Lifespan Using CentileBrain: Algorithm Benchmarking and Model Optimization

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

Estimating cortical thickness trajectories in children across different scanners using transfer learning from normative models

Gaiser,

Berthet,

Kia

et al. 2024

Human Brain Mapping

View full text Add to dashboard Cite

This work illustrates the use of normative models in a longitudinal neuroimaging study of children aged 6–17 years and demonstrates how such models can be used to make meaningful comparisons in longitudinal studies, even when individuals are scanned with different scanners across successive study waves. More specifically, we first estimated a large‐scale reference normative model using Hierarchical Bayesian Regression from N = 42,993 individuals across the lifespan and from dozens of sites. We then transfer these models to a longitudinal developmental cohort (N = 6285) with three measurement waves acquired on two different scanners that were unseen during estimation of the reference models. We show that the use of normative models provides individual deviation scores that are independent of scanner effects and efficiently accommodate inter‐site variations. Moreover, we provide empirical evidence to guide the optimization of sample size for the transfer of prior knowledge about the distribution of regional cortical thicknesses. We show that a transfer set containing as few as 25 samples per site can lead to good performance metrics on the test set. Finally, we demonstrate the clinical utility of this approach by showing that deviation scores obtained from the transferred normative models are able to detect and chart morphological heterogeneity in individuals born preterm.

show abstract

ESPA: An Unsupervised Harmonization Framework via Enhanced Structure Preserving Augmentation

Eshaghzadeh Torbati,

Minhas,

Tafti

et al. 2024

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Accommodating site variation in neuroimaging data using normative and hierarchical Bayesian models

Cited by 37 publications

References 42 publications

Normative Modeling of Brain Morphometry Across the Lifespan Using CentileBrain: Algorithm Benchmarking and Model Optimization

Normative Modeling of Brain Morphometry Across the Lifespan Using CentileBrain: Algorithm Benchmarking and Model Optimization

Estimating cortical thickness trajectories in children across different scanners using transfer learning from normative models

ESPA: An Unsupervised Harmonization Framework via Enhanced Structure Preserving Augmentation

Contact Info

Product

Resources

About