From a deep learning model back to the brain—Identifying regional predictors and their relation to aging

Description: Deep convolutional neural networks (CNN) enabled a major leap in image processing tasks including brain imaging analysis. In this work we utilized CNNs for predicting individual’s chronological age based on their T1 MRI. Whereas CNNs provide high predictive power, it is often difficult to identify the features that underlie a given prediction. Discovering which brain structures, contribute most to the prediction of a given model have significant theoretical and translational value. Previous work examined methods to attribute pixel/voxel-wise contribution to the prediction in a single image, resulting in ‘explanation maps’ (EM) that were found noisy and unreliable. Here, we developed a novel inference framework for combining these maps across subjects, thus creating a population-based rather than subject specific maps. We apply this method on a CNNs ensemble trained on predicting subjects’ chronological age from raw anatomical T1 brain images of 10,176 healthy subjects, obtained from various open-source datasets.

Communities: developmental

Related article:

View ID Name Type
View ID Name Type
Field Value
Compact Identifier
Add DateJune 26, 2019, 3:43 p.m.
Uploaded bygidonle
Related article DOI10.1002/hbm.25011
Related article authorsGidon Levakov, Gideon Rosenthal, Ilan Shelef, Tammy Riklin Raviv and Galia Avidan
Citation guidelines

If you use the data from this collection please include the following persistent identifier in the text of your manuscript:

This will help to track the use of this data in the literature. In addition, consider also citing the paper related to this collection.