Abstract:Purpose: Brain aging is a complex and heterogeneous process characterized by both structural and functional decline. This study aimed to establish a novel deep learning (DL) method for predicting brain age by utilizing structural and metabolic imaging data. Methods: The dataset comprised participants from both the Universal Medical Imaging Diagnostic Center (UMIDC) and the Alzheimer's Disease Neuroimaging Initiative (ADNI). The former recruited 395 normal control (NC) subjects, while the latter included 438 NC subjects, 51 mild cognitive impairment (MCI) subjects, and 56 Alzheimer's disease (AD) subjects. We developed a novel dual-pathway, 3D simple fully convolutional network (Dual-SFCNeXt) to estimate brain age using [18F]fluorodeoxyglucose positron emission tomography ([18F]FDG PET) and structural magnetic resonance imaging (sMRI) images of NC subjects as input. Several prevailing DL models were trained and tested using either MRI or PET data for comparison. Model accuracies were evaluated using mean absolute error (MAE) and Pearson's correlation coefficient (r). Brain age gap (BAG), deviations of brain age from chronologic age, was correlated with cognitive assessments in MCI and AD subjects. Results: Both PET- and MRI-based models achieved high prediction accuracy. The leading model was the SFCNeXt (the single-pathway version) for PET (MAE = 2.92, r = 0.96) and MRI (MAE = 3.23, r = 0.95) on all samples. By integrating both PET and MRI images, the Dual-SFCNeXt demonstrated significantly improved accuracy (MAE = 2.37, r = 0.97) compared to all single-modality models. Significantly higher BAG was observed in both the AD (P < 0.0001) and MCI (P < 0.0001) groups compared to the NC group. BAG correlated significantly with Mini-Mental State Examination (MMSE) scores (r=-0.390 for AD, r=-0.436 for MCI) and the Clinical Dementia Rating Scale Sum of Boxes (CDR-SB) scores (r = 0.333 for AD, r = 0.372 for MCI). Conclusion: The integration of [18F]FDG PET with structural MRI enhances the accuracy of brain age prediction, potentially introducing a new avenue for related multimodal brain age prediction studies.

Brain Age Prediction Based on Resting-State Functional MRI Using Similarity Metric Convolutional Neural Network.

Brain Age Prediction Based on Resting-State Functional Connectivity Patterns Using Convolutional Neural Networks

Predicting Brain Age of Healthy Adults Based on Structural MRI Parcellation Using Convolutional Neural Networks

Parallel Pathway Convolutional Neural Network with Low-rank Fusion for Brain Age Prediction.

Brain Age Estimation From MRI Using Cascade Networks with Ranking Loss

Brain Age Estimation by Convolutional Neural Network Based on T1-Weighted MRI and Gender Feature

Connectome-based Predictive Modelling of Ageing, Overall Cognitive Functioning and Memory Performance.

Age Prediction Using Resting-State Functional MRI

Brain Age Prediction in Developing Childhood with Multimodal Magnetic Resonance Images.

Optimising a Simple Fully Convolutional Network for Accurate Brain Age Prediction in the PAC 2019 Challenge

Accurate Brain Age Prediction with Lightweight Deep Neural Networks.

MRI-based brain age prediction model for children under 3 years old using deep residual network

Group Similarity Constraint Functional Brain Network Estimation for Mild Cognitive Impairment Classification

Age Estimation From Mr Images Via 3d Convolutional Neural Network And Densely Connect

OTFPF: Optimal Transport-Based Feature Pyramid Fusion Network for Brain Age Estimation with 3D Overlapped ConvNeXt

Brain Age Estimation Using Multi-Feature-based Networks

SFCNeXt: a simple fully convolutional network for effective brain age estimation with small sample size

Brain Age Prediction of Children Using Routine Brain MR Images via Deep Learning

Enhancing Brain Age Estimation with a Multimodal 3D CNN Approach Combining Structural MRI and AI-Synthesized Cerebral Blood Volume Data

[18F]FDG PET integrated with structural MRI for accurate brain age prediction

Brain age prediction using combined deep convolutional neural network and multi-layer perceptron algorithms