search

james-cole / brainageR

Software for generating a brain-predicted age value, using Gaussian Processes regression, implemented in R
GNU Lesser General Public License v3.0
74 stars 26 forks source link

How to correct predicted brain age? #13

Open jojowang-17 opened 7 months ago

jojowang-17 commented 7 months ago

Hi,

Brain age model are often impacted by the regression to the mean, resulting an overestimation of the age of younger subjects and an underestimation of the age of older subjects. I tested this model on PPMI dataset, but found that most subjects are with much younger brain age.

I think some correction should be conducted here. For example, one way is to rely on the slope (alpha) and intercept (beta) of a linear regression model of BrainAGE against chronological age in the training set. This way an offset is calculated (as alpha * omega + beta) and then subtracted from the estimated brain-age to yield a bias-free BrainAGE.

Could you offer the the alpha, beta value of this BrainAgeR model? or any suggestions on correcting brain age when using BrainAgeR model?

Thanks!

james-cole commented 7 months ago

Hi there,

Thanks for your interest in brainageR. I totally agree re: the regression to the mean effect being an issue, though my preference is not to adjust the model. The effect is more pronounced when the fit is bad, and I think it is important to detect generalisability issues.

My view is that it is more transparent to covary for chronological age in subsequent analyses, and will be statistically equivalent in most cases. See https://www.sciencedirect.com/science/article/pii/S2213158220300668

Hope that helps, James

From: jojowang-17 @.> Date: Wednesday, 28 February 2024 at 6:47 pm To: james-cole/brainageR @.> Cc: Subscribed @.***> Subject: [james-cole/brainageR] How to correct predicted brain age? (Issue #13)

⚠ Caution: External sender

Hi,

Brain age model are often impacted by the regression to the mean, resulting an overestimation of the age of younger subjects and an underestimation of the age of older subjects. I tested this model on PPMI dataset, but found that most subjects are with much younger brain age.

I think some correction should be conducted here. For example, one way is to rely on the slope (alpha) and intercept (beta) of a linear regression model of BrainAGE against chronological age in the training set. This way an offset is calculated (as alpha * omega + beta) and then subtracted from the estimated brain-age to yield a bias-free BrainAGE.

Could you offer the the alpha, beta value of this BrainAgeR model? or any suggestions on correcting brain age when using BrainAgeR model?

Thanks!

— Reply to this email directly, view it on GitHubhttps://github.com/james-cole/brainageR/issues/13, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACCUV6BFBVF3EYVTCNQE5GTYV53UJAVCNFSM6AAAAABD6SGZ4SVHI2DSMVQWIX3LMV43ASLTON2WKOZSGE2TSNRSGAZDSMI. You are receiving this because you are subscribed to this thread.Message ID: @.***>