Predicting biological brain age using deep learning methods

Background The interplay of genetic and environmental factors can trigger a cascade neuropathological changes leading to individual differences in brain aging. apparent age the (brain age) differ from chronological age, potentially reflecting different resilience mechanisms. Brain has also potential as biomarker predict future cognitive impairments dementia. In this study, we developed biological measure based on between predicted (PBA) (CA) using deep learning model (DLM). Method sample included 16734 T1-w MRI multiple time-points 15115 healthy individuals, aged 32-96 yrs., from: 1) ADNI (n = 1489); 2) AIBL 957); GENIC 406); 4) UK Biobank 13882). Healthy status was defined absence dementia/cognitive impairment, neuro-psychiatric disorders, and/or self-report good health. Medical diagnoses were collected through physician-/self-reports or medical records following International Classification Diseases. A DLM convolutional neural networks trained develop an algorithm raw images registered MNI space training (90%) hold-out test (10%) sets. To assess generalizability, applied Gothenburg H70-Birth Cohort 1944 792 septuagenarians). Result method achieved mean absolute error (MAE) error, PBA CA, 3.04 yrs. H70 cohort 73.9±1.49 [64.7-86.1] years. Conclusion shows accurate predictions, comparable with those previous studies other computation methods. As next steps, will evaluate hyperparameter optimization implementation cross-validation methodology. We seek include further cohorts individuals heterogeneous age-span increase reliability generalizability. This enable address e.g., normal pathological aging well exploring