Abstract:Brain aging involves structural and functional changes and therefore serves as a key biomarker for brain health. Combining structural magnetic resonance imaging (sMRI) and functional magnetic resonance imaging (fMRI) has the potential to improve brain age estimation by leveraging complementary data. However, fMRI data, being noisier than sMRI, complicates multimodal fusion. Traditional fusion methods often introduce more noise than useful information, which can reduce accuracy compared to using sMRI alone. In this paper, we propose a novel multimodal framework for biological brain age estimation, utilizing a sex-aware adversarial variational autoencoder (SA-AVAE). Our framework integrates adversarial and variational learning to effectively disentangle the latent features from both modalities. Specifically, we decompose the latent space into modality-specific codes and shared codes to represent complementary and common information across modalities, respectively. To enhance the disentanglement, we introduce cross-reconstruction and shared-distinct distance ratio loss as regularization terms. Importantly, we incorporate sex information into the learned latent code, enabling the model to capture sex-specific aging patterns for brain age estimation via an integrated regressor module. We evaluate our model using the publicly available OpenBHB dataset, a comprehensive multi-site dataset for brain age estimation. The results from ablation studies and comparisons with state-of-the-art methods demonstrate that our framework outperforms existing approaches and shows significant robustness across various age groups, highlighting its potential for real-time clinical applications in the early detection of neurodegenerative diseases.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to solve the following key problems: 1. **Challenges in multimodal data fusion**: - Structural magnetic resonance imaging (sMRI) and functional magnetic resonance imaging (fMRI) provide complementary information, but fMRI data is usually noisier than sMRI. Traditional fusion methods may introduce more interfering information when dealing with this noise, thus reducing the accuracy of the model. - Formula representation: \[ \text{fMRI noise} > \text{sMRI noise} \] 2. **The influence of gender differences on brain - age estimation**: - Existing studies rarely consider the anatomical and functional differences between male and female brains, and these differences will affect the estimation of brain age. Therefore, how to integrate gender information into the model to improve prediction accuracy is an important problem. 3. **Feature decoupling and representation learning**: - In order to make better use of the complementary information in sMRI and fMRI data, a method that can effectively decouple shared features and modality - specific features is required. This helps to reduce the influence of noise and improve the quality of feature representation. - Formula representation: \[ z_i = [\text{Shared}(z_i), \text{Dist}(z_i)] \] - Where \(\text{Shared}(z_i)\) represents shared features and \(\text{Dist}(z_i)\) represents modality - specific features. 4. **Application of single - modality data**: - In practical applications, the situation of having both sMRI and fMRI data simultaneously is relatively limited, and in most cases, only sMRI data is available. Therefore, how to make the model maintain high performance when only sMRI data is available is a practical problem. To solve these problems, the authors propose a new multimodal framework - **Sex - Aware Adversarial Variational Autoencoder (SA - A V AE)**. This framework effectively decouples the features in sMRI and fMRI data by combining adversarial learning and variational auto - encoder techniques, and introduces gender information to capture gender - specific aging patterns. In addition, the framework also designs a single - modality adaptation strategy, enabling it to perform well even when only sMRI data is available. ### Summary This paper mainly solves problems such as multimodal neuroimaging data fusion, the influence of gender differences on brain - age estimation, and feature decoupling, and proposes an innovative framework to improve the accuracy and robustness of biological brain - age estimation.

Biological Brain Age Estimation using Sex-Aware Adversarial Variational Autoencoder with Multimodal Neuroimages

Multi-Task Adversarial Variational Autoencoder for Estimating Biological Brain Age with Multimodal Neuroimaging

Robust Brain Age Estimation via Regression Models and MRI-derived Features

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

Multimodal biological brain age prediction using magnetic resonance imaging and angiography with the identification of predictive regions

Uncertainty-Based Biological Age Estimation of Brain MRI Scans

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

Graph Transformer Geometric Learning of Brain Networks Using Multimodal MR Images for Brain Age Estimation

MRI Volume-Based Robust Brain Age Estimation Using Weight-Shared Spatial Attention in 3D CNNs

Brain Age Estimation Using Universum Learning-Based Kernel Random Vector Functional Link Regression Network

Distinguishing Healthy Ageing from Dementia: a Biomechanical Simulation of Brain Atrophy using Deep Networks

Brain structure ages—A new biomarker for multi‐disease classification

Brain age prediction across the human lifespan using multimodal MRI data

Explainable Brain Age Prediction using coVariance Neural Networks

Variational AutoEncoder For Regression: Application to Brain Aging Analysis

Studying the Effects of Sex-related Differences on Brain Age Prediction using brain MR Imaging

Biological Brain Age Prediction Using Cortical Thickness Data: A Large Scale Cohort Study

Disentangled-Multimodal Adversarial Autoencoder: Application to Infant Age Prediction With Incomplete Multimodal Neuroimages

SynthBA: Reliable Brain Age Estimation Across Multiple MRI Sequences and Resolutions

Dual Graph Attention based Disentanglement Multiple Instance Learning for Brain Age Estimation

Deep Neural Networks Learn General and Clinically Relevant Representations of the Ageing Brain