*library_books*

## Similar protocols

## Protocol publication

[…] An overview of the brain-predicted age calculation procedure is presented in . All structural images were preprocessed using **SPM**12. Images were bias corrected and segmented into GM, WM, and CSF using SPM Segment. Visual quality control was carried out at this stage to ensure the accuracy of image segmentation; all images were included for both groups. Segmented images for GM and WM were then nonlinearly registered to a custom template, based on the training dataset, using SPM DARTEL (). Finally, images were affine registered to MNI152 space (voxel size = 1.5 mm3) and resampled using modulation to retain volumetric information and smoothed with a 4-mm full-width half-maximum Gaussian kernel. Summary measures of brain volumes were also generated for GM, WM, CSF, and intracranial volume (ICV).Fig. 1Brain-predicted ages were generated as previously outlined (), using the **Pattern** Recognition for Neuroimaging Toolbox (PRoNTo v2.0, www.mlnl.cs.ucl.ac.uk/pronto) software package. First, a model of healthy brain aging was defined using brain volumetric maps from the training dataset (N = 2001) as follows: Spatially normalized images were converted to vectors and the resulting GM and WM vectors were concatenated for each individual. A linear kernel representation of these data was derived by calculating an N × N similarity matrix, where each point in the matrix was the dot product of 2 participants' image vectors. A Gaussian Processes regression model was then defined, with chronological age as the dependent variable and 3-dimensional brain volumetric image data (in similarity matrix form) as the independent variables.Predictions for all training participants were generated using 10-fold cross-validation, whereby the data were randomly divided into 10 folds, each comprising 10% of the participants. The model was then retrained using 9 folds of the data and age predictions were made for data in the “left-out” fold. This procedure was iterated so that all folds were left out in turn, resulting in unbiased (i.e., independent) age predictions for each participant. Model accuracy was expressed as the correlation between age and brain-predicted age (Pearson's r), total variance explained (R2), mean absolute error (MAE), and root mean squared error (RMSE). Statistical significance of this model was assessed using permutation testing (n = 1000).Next, the coefficients from the full training model (N = 2001) were applied to the test data (i.e., DS participants and controls), to generate unbiased brain-predicted ages. Finally, brain-predicted age difference (brain-PAD) scores were calculated for each individual in the DS and control groups by subtracting chronological age from brain-predicted age. Hence, a positive brain-PAD score indicates that the individual's brain is predicted to be “older” than their chronological age. Brain-PAD scores were subsequently used for further analysis to index relative structural brain aging. […]

## Pipeline specifications

Software tools | SPM, PRoNTo |
---|---|

Application | Neuroimaging analysis |

Organisms | Homo sapiens |

Diseases | Alzheimer Disease |