Purpose
The brain is constantly changing throughout our lifespan. For normal aging, some progression of substance loss and change in a variety of brain structures are expected, e.g. volumetric, morphometric and signal changes. The biological age of the brain thus reflects an individual’s chronological age. In individuals with neurodegenerative disorders, such as Alzheimer’s disease, one expects biological and cognitive deviations from normal aging.
These deviations can partly be detected as a gap between “brain age” and chronological age. Having a method that is able to estimate...
Methods and materials
To train our model we used a total of 7462 T1-weighted images from 4355 healthy subjects sourced from eight publicly available data sets: ADNI [2], CC-359 [3], HCP Young Adult [4], IXI [5], OASIS [6], PPMI [7], SALD [8] and YALE [9] (ages 20-90 yrs, mean 61.08, m/f = 1992/2363, Fig. 2). To ensure generalization and robustness of our approach, we evaluated our model on completely independent data: 473 T1-weighted MR images from 267 healthy elderly subjects from the AIBL data set [10] (ages 60-85...
Results
The mean absolute error of the predicted brain age versus the chronological age for subjects in the test set were 4.14 yrs (SD 5.78). For each input volume, age prediction was made in less than 15s, including all the aforementioned preprocessing steps (Fig. 4).
Conclusion
As the trained model is able to accurately predict the chronological age very fast, it could potentially be incorporated into clinical routine through PACS systems and established postprocessing tools.
We are currently working to incorporate techniques from “explainable AI” into our setup. As CNNs extract image-features automatically, model interpretation is challenging. However, by using techniques like regression activation maps we aim to provide a better understanding of model predictions, and also potentially uncover features (brain MRI-derived biomarkers) related to abnormal aging.
Acknowledgments:Our work was supported...
Personal information and conflict of interest
S. Kaliyugarasan; Bergen/NO - nothing to discloseA. Lundervold; Bergen/NO - nothing to discloseA. S. Lundervold; Bergen/NO - nothing to disclose
References
Cole, James H., et al. "Predicting brain age with deep learning from raw imaging data results in a reliable and heritable biomarker." NeuroImage 163 (2017): 115-124.
Alzheimer's Disease Neuroimaging Initiative (ADNI). http://adni.loni.usc.edu.
Calgary-Campinas-359 (CC-359). https://sites.google.com/view/calgary-campinas-dataset/home/download.
The Human Connectome Project (HCP). https://www.humanconnectome.org/study/hcp-young-adult/data-releases.
IXI Dataset. https://brain-development.org/ixi-dataset.
Open Access Series of Imaging Studies (OASIS). https://www.oasis-brains.org.
Parkinson’s Progression Markers Initiative (PPMI). http://www.ppmi-info.org.
Southwest University Adult Lifespan Dataset (SALD). http://fcon_1000.projects.nitrc.org/indi/retro/sald.html.
Yale Low-Resolution Controls Dataset (YALE). http://fcon_1000.projects.nitrc.org/indi/retro/yale_lowres.html.
The Australian Imaging, Biomarkers and Lifestyle (AIBL). https://aibl.csiro.au.