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

TL;DR

Biological brain age estimation benefits from multimodal MRI, but fMRI's noise complicates fusion. The authors propose SA-AVAE, a sex-aware adversarial variational autoencoder that disentangles shared and modality-specific latent features from sMRI and fMRI, while incorporating sex information into the latent space for sex-aware aging patterns. On the OpenBHB dataset, SA-AVAE achieves state-of-the-art accuracy and robustness across age groups, with ablation confirming the value of disentanglement and sex conditioning. The framework shows potential for real-time clinical deployments in early detection of neurodegenerative diseases and provides a scalable approach for multimodal biomarker estimation.

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.

Figures (6)

• Figure 1: Visualization of structural magnetic resonance imaging (sMRI) and functional magnetic resonance imaging (fMRI) in male and female patients across various age groups. sMRI reveals the anatomical details of the brain, whereas fMRI depicts brain activity by measuring changes in blood flow; this is shown on a color scale where warmer colors typically indicate higher levels of activity.
• Figure 2: Architecture of the proposed Multimodal Sex-Aware Adversarial Variational Autoencoder (SA-AVAE) for predicting biological brain age, utilizing sMRI as a compulsory modality and fMRI as an optional input to enhance prediction performance.
• Figure 3: Age distribution of male and female participants in the training and validation sets of the OpenBHB datasetdufumier2022openbhb , shown in (a) and (b), respectively.
• Figure 4: Illustration of architectural details of our proposed Sex-Aware Adversarial Variational Autoencoder.
• Figure 5: Comparison of predictions for different multi-modal models: (a) AAE, (b) VAE, (c) AVAE, (d) SA-AVAE. Each graph plots predicted brain age versus chronological brain age, with gender and confidence intervals indicated.