Predicting Brain Age at Slice Level: Convolutional Neural Networks and Consequences for Interpretability

Abstract

Problem: Chronological aging in later life is associated with brain degeneration processes and increased risk for disease such as stroke and dementia. With a worldwide tendency of aging populations and increased longevity, mental health, and psychiatric research have paid increasing attention to understanding brain-related changes of aging. Recent findings suggest there is a brain age gap (a difference between chronological age and brain age predicted by brain imaging indices); the magnitude of the gap may indicate early onset of brain aging processes and disease. Artificial intelligence has allowed for a narrowing of the gap in chronological and predicted brain age. However, the factors that drive model predictions of brain age are still unknown, and there is not much about these factors that can be gleaned from the black-box nature of machine learning models. The goal of the present study was to test a brain age regression approach that is more amenable to interpretation by researchers and clinicians. Methods: Using convolutional neural networks we trained multiple regressor models to predict brain age based on single slices of magnetic resonance imaging, which included gray matter- or white matter-segmented inputs. We evaluated the trained models in all brain image slices to generate a final prediction of brain age. Unlike whole-brain approaches to classification, the slice-level predictions allows for the identification of which brain slices and associated regions have the largest difference between chronological and neuroimaging-derived brain age. We also evaluated how model predictions were influenced by slice index and plane, participant age and sex, and MRI data collection site. Results: The results show, first, that the specific slice used for prediction affects prediction error (i.e., difference between chronological age and neuroimaging-derived brain age); second, the MRI site-stratified separation of training and test sets removed site effects and also minimized sex effects; third, the choice of MRI slice plane influences the overall error of the model. Conclusion: Compared to whole brain-based predictive models of neuroimaging-derived brain age, slice-based approach improves the interpretability and therefore the reliability of the prediction of brain age using MRI data.

Keywords: brain age; convolutional neural networks; deep learning; model interpretability; neuroimaging.

Figure 1

Depiction of the brain age prediction framework. Each view has an independent CNN model and an independently-trained linear regression model. S is the number of slices and N and M are the dimensions of the slice (e.g., if evaluating the axial slice, the N and M are the dimensions for the sagittal and coronal views).

Figure 2

Examples of the augmentation procedure. First row are gray matter segmented images before augmentation; the second row are their augmented counterparts.

Figure 3

Change in Mean Absolute Error (MAE) with respect to changing the slice that is evaluated by the network. Each slice index value are an average of either all training set or all validation set. The shade represents the 0.95 confidence interval for those points. The slices in the image are examples of the index that best or worst predicts brain age.

Figure 4

Regression curves for the validation set. Every point represents a person (each person is presented three times, one for each view). Dashed red orthogonal to the x-axis is the age average of the dataset, while the horizontal dashed line is aligned to 0 error as a reference.

Figure 5

Site effects for axial, coronal, and sagittal views. For each orientation, the chronological age and predicted age are shown side-by-side by site.

Figure 6

Lightbox view of axial slices age predictions following the voxel-level approach. The value for σ indicates the amount of gaussian spatial smoothing applied to the predictions. Images on a range from 20 (red) to 60 (yellow).

Figure 7

Change in Mean Absolute Error (MAE) with respect to changing the slice that is evaluated by the network. Two independent models are evaluated, one for gray matter (solid lines) and another one for white matter (dashed lines). Each slice index value is an average of the entire training set or the entire validation set. The shade represents the 0.95 confidence interval for those points.