Brain age prediction via cross-stratified ensemble learning

• Published in: NeuroImage (Orlando, Fla.) Vol. 299; p. 120825
• Main Authors: Li, Xinlin, Hao, Zezhou, Li, Di, Jin, Qiuye, Tang, Zhixian, Yao, Xufeng, Wu, Tao
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.10.2024; Elsevier Limited; Elsevier
• Subjects: Age; Aging; Algorithms; Alzheimer's disease; Alzheimer's disease (AD); Biomarkers; Brain; Brain age prediction; Brain aging; Consortia; Datasets; Deep learning (DL); Disease; Ensemble learning; Learning; Magnetic resonance imaging; Neural networks; Neurodegenerative diseases; Neuroimaging; Parkinson's disease; Brain age prediction; Brain aging; Alzheimer's disease (AD); Deep learning (DL); Ensemble learning
• ISSN: 1053-8119; 1095-9572; 1095-9572
• DOI: 10.1016/j.neuroimage.2024.120825

•The proposed cross-stratified ensemble learning algorithm via three varied deep learning base learners can be used to improve the accuracy of brain age prediction, respectively.•It was demonstrated that the cross-stratified ensemble learning algorithm adapted to the age variation of the subjects in different age groups with the capability of balancing the learn differences of base learners.•The valuable biomarker of predicted age difference (PAD) presented the increased trend across the normal control (NC), mild cognitive impairment (MCI) and Alzheimer's disease (AD). As an important biomarker of neural aging, the brain age reflects the integrity and health of the human brain. Accurate prediction of brain age could help to understand the underlying mechanism of neural aging. In this study, a cross-stratified ensemble learning algorithm with staking strategy was proposed to obtain brain age and the derived predicted age difference (PAD) using T1-weighted magnetic resonance imaging (MRI) data. The approach was characterized as by implementing two modules: one was three base learners of 3D-DenseNet, 3D-ResNeXt, 3D-Inception-v4; another was 14 secondary learners of liner regressions. To evaluate performance, our method was compared with single base learners, regular ensemble learning algorithms, and state-of-the-art (SOTA) methods. The results demonstrated that our proposed model outperformed others models, with three metrics of mean absolute error (MAE), root mean-squared error (RMSE), and coefficient of determination (R2) of 2.9405 years, 3.9458 years, and 0.9597, respectively. Furthermore, there existed significant differences in PAD among the three groups of normal control (NC), mild cognitive impairment (MCI) and Alzheimer's disease (AD), with an increased trend across NC, MCI, and AD. It was concluded that the proposed algorithm could be effectively used in computing brain aging and PAD, and offering potential for early diagnosis and assessment of normal brain aging and AD.