Predicting brain age with deep learning from raw imaging data results in a reliable and heritable biomarker (1612.02572v1)

Abstract: Machine learning analysis of neuroimaging data can accurately predict chronological age in healthy people and deviations from healthy brain ageing have been associated with cognitive impairment and disease. Here we sought to further establish the credentials of "brain-predicted age" as a biomarker of individual differences in the brain ageing process, using a predictive modelling approach based on deep learning, and specifically convolutional neural networks (CNN), and applied to both pre-processed and raw T1-weighted MRI data. Firstly, we aimed to demonstrate the accuracy of CNN brain-predicted age using a large dataset of healthy adults (N = 2001). Next, we sought to establish the heritability of brain-predicted age using a sample of monozygotic and dizygotic female twins (N = 62). Thirdly, we examined the test-retest and multi-centre reliability of brain-predicted age using two samples (within-scanner N = 20; between-scanner N = 11). CNN brain-predicted ages were generated and compared to a Gaussian Process Regression (GPR) approach, on all datasets. Input data were grey matter (GM) or white matter (WM) volumetric maps generated by Statistical Parametric Mapping (SPM) or raw data. Brain-predicted age represents an accurate, highly reliable and genetically-valid phenotype, that has potential to be used as a biomarker of brain ageing. Moreover, age predictions can be accurately generated on raw T1-MRI data, substantially reducing computation time for novel data, bringing the process closer to giving real-time information on brain health in clinical settings.

• The paper shows that the CNN model accurately predicts brain age with a mean absolute error of 4.16 years and a correlation of 0.96 compared to traditional methods.
• It demonstrates that brain-predicted age has high heritability with an estimate of 0.84, underscoring the genetic influence on brain aging.
• The method offers improved clinical utility by reducing pre-processing steps and achieving high test-retest reliability across scanners.

Deep Learning-Based Brain Age Prediction as a Biomarker

The paper presents a paper on predicting brain age using deep learning, specifically employing convolutional neural networks (CNNs), applied to raw and pre-processed T1-weighted MRI data. The research aims to establish brain-predicted age as a reliable biomarker of brain ageing, demonstrating its potential heritability and its use for assessing cognitive decline and neurodegenerative diseases.

Methodology and Results Overview

The research utilized a CNN approach to predict brain age from neuroimaging data and compared it to Gaussian Processes Regression (GPR), a traditional method previously utilized for similar tasks. The paper involved comprehensive datasets:

1. Accuracy Evaluation: Using the Brain-Age Normative Control (BANC) dataset with 2,001 healthy individuals, the CNN method achieved a correlation (r) of 0.96 with a mean absolute error (MAE) of 4.16 years using grey matter (GM) data, showcasing comparable accuracy to GPR.
2. Heritability Assessment: The paper analyzed a sample of 62 twins to evaluate the heritability of brain-predicted age. The CNN model, using combined GM and white matter (WM) data, yielded a heritability estimate of 0.84, suggesting a substantial genetic influence.
3. Reliability Testing: High test-retest reliability was observed within a single scanner (ICC = 0.90-0.98) and variable multi-centre reliability (ICC = 0.51-0.96), indicating robustness, albeit with reduced reliability for WM and raw data between different scanners.

Implications and Future Prospects

The paper demonstrates the feasibility of using CNNs to predict brain age directly from raw MRI data, thereby streamlining the workflow by eliminating extensive pre-processing. This reduction in computational time facilitates potential real-time clinical applications, offering timely insights into brain health.

The strong correlation between genetic factors and brain-predicted age underlines the biomarker's relevance, encouraging further investigation into the genetic determinants of brain ageing. This could lead to identifying potential targets for interventions aimed at mitigating age-associated cognitive decline.

The suggestions for future research include expanding the heritability sample to include diverse populations and exploring different scanner technologies to improve between-center reliability. These steps are crucial for the broader clinical translation of brain-predicted age as a biomarker.

Conclusion

The findings support CNN-based brain age predictions as an accurate and reliable biomarker for brain ageing, with promising applications in clinical settings. Its high heritability emphasizes the need for genetic research, while ensuring reliability across different technological settings remains a key challenge for future endeavors.