Brain age acceleration as biomarker of Alzheimer’s disease progression: Functional network connectivity analysis

Abstract

AbstractBackgroundThe brain age gap, the difference between an individual’s brain predicted age and chronological age, is used as a brain disease biomarker and aging. To date, although previous studies utilized mostly structural magnetic resonance imaging (MRI) data to predict brain age, less work has used functional network connectivity (FNC) from resting‐state functional MRI (rs‐fMRI) to indicate brain age and its link with Alzheimer’s disease (AD) progression. This study aims to estimate brain age from FNC and introduce it as a biomarker of AD progression.MethodThis study used 1091 rs‐fMRI data and the chronological age at the time of scanning, ranging from 42‐95, from the Open Access Series of Imaging Studies (OASIS)‐3 cohort. This dataset includes 1021 healthy subjects (HC) and 70 AD patients. We used group independent component analysis to estimate 53 components for the whole‐brain. Then, FNC feature, in total 1378, of each subject was calculated by applying the Pearson correlation. A support vector regression (SVR) was trained on 951 HC subjects in which FNC features and the chronological age was served as predictor and prediction output, respectively. We then tested the trained model on 70 HC (age mean± sd: 73.66± 7.47) and 70 AD (age mean± sd: 73.34± 7.51) subjects to predict their brain age.ResultThe correlation between predicted brain age and the chronological age of HC test data was R=0.73, (p=5.3e‐13, N=70), while the correlation between predicted brain age and AD patients' chronological age was R=0.33 (p=0.004, N=70). This result shows that the model based on HC could better predict the HC age than the AD patients. The brain age gap mean and standard deviation for the test HC and AD group was ‐2.2581± 5.0879 and 2.0814± 7.4520. This result shows that the brain age gap was increased in AD (Cohen’s d=0.68, p<0.001).ConclusionWe proved that FNC, estimated from rs‐fMRI, could predict the brain age in HC. Also, we found an acceleration in the brain predicted age of AD patients. We also found that the AD subjects' brain age gap was significantly higher than the brain age gap of the HC group.