Abstract:Normative modelling is an emerging method for understanding the underlying heterogeneity within brain disorders like Alzheimer Disease (AD) by quantifying how each patient deviates from the expected normative pattern that has been learned from a healthy control distribution. Since AD is a multifactorial disease with more than one biological pathways, multimodal magnetic resonance imaging (MRI) neuroimaging data can provide complementary information about the disease heterogeneity. However, existing deep learning based normative models on multimodal MRI data use unimodal autoencoders with a single encoder and decoder that may fail to capture the relationship between brain measurements extracted from different MRI modalities. In this work, we propose multi-modal variational autoencoder (mmVAE) based normative modelling framework that can capture the joint distribution between different modalities to identify abnormal brain structural patterns in AD. Our multi-modal framework takes as input Freesurfer processed brain region volumes from T1-weighted (cortical and subcortical) and T2-weighed (hippocampal) scans of cognitively normal participants to learn the morphological characteristics of the healthy brain. The estimated normative model is then applied on Alzheimer Disease (AD) patients to quantify the deviation in brain volumes and identify the abnormal brain structural patterns due to the effect of the different AD stages. Our experimental results show that modeling joint distribution between the multiple MRI modalities generates deviation maps that are more sensitive to disease staging within AD, have a better correlation with patient cognition and result in higher number of brain regions with statistically significant deviations compared to a unimodal baseline model with all modalities concatenated as a single input.

What problem does this paper attempt to address?

This paper aims to solve the problem of abnormal brain structure pattern recognition in patients with Alzheimer's disease (AD). Specifically, the paper proposes a normative modeling framework based on multi - modal variational auto - encoder (mmVAE) to capture the joint distribution between different MRI modalities, so as to more accurately identify the abnormal brain structure patterns in Alzheimer's disease patients. ### Specific problems that the paper attempts to solve include: 1. **Identifying abnormal brain structure patterns in AD patients**: By constructing a multi - modal variational auto - encoder that can learn the brain region volumes of the healthy control group, the paper hopes to quantify the deviation of brain volume in AD patients and identify the abnormal brain structure patterns caused by the influence of different stages of AD. 2. **Improving the sensitivity to disease staging**: Existing unimodal auto - encoders may not be able to fully capture the relationship between different MRI modalities, resulting in insufficient sensitivity to different stages of AD. The method proposed in the paper is more sensitive to disease staging by jointly modeling the data of multiple MRI modalities to generate deviation maps. 3. **Enhancing the correlation with patients' cognitive functions**: The paper hopes that through the multi - modal modeling method, the generated deviation maps can better reflect the changes in patients' cognitive functions, thereby providing more valuable information for clinical diagnosis. 4. **Increasing the number of brain regions with statistically significant deviations**: Compared with the unimodal baseline model, the method proposed in the paper can identify more brain regions with statistically significant deviations, thus more comprehensively revealing the brain structure abnormalities in AD patients. ### Method overview: - **Multi - modal variational auto - encoder (mmVAE)**: The paper designs a framework that contains two modality - specific encoders and decoders, which respectively process T1 - weighted and T2 - weighted MRI data. The latent space parameters of different modalities are combined through the Product of Experts (PoE) layer to form a shared latent space, and then reconstructed by the modality - specific decoders. - **Normative modeling**: First, train the mmVAE on the healthy control group to learn the morphological features of the healthy brain. Then apply the model to AD patients, quantify the deviation of their brain volumes, and standardize these deviations by Z - score. - **Deviation calculation and statistical analysis**: Calculate the brain region volume deviation of each AD patient and compare it with the healthy control group. Control the false positive rate through FDR correction and identify the brain regions with statistically significant deviations. ### Experimental results: - **Sensitivity to disease staging**: The experimental results show that the deviation maps generated by mmVAE are more sensitive to different stages of AD, and the deviation values increase with the increase of the severity of the disease. - **Correlation with patients' cognitive functions**: The deviation maps generated by mmVAE have a higher correlation with ADAS13 and RAVLT cognitive assessment scores. - **Number of brain regions with statistically significant deviations**: The number of brain regions with statistically significant deviations identified by mmVAE is larger. ### Conclusion: The multi - modal variational auto - encoder (mmVAE) normative modeling framework proposed in the paper performs well in identifying abnormal brain structure patterns in AD patients, can more sensitively reflect disease staging, better correlate with patients' cognitive functions, and identify more brain regions with statistically significant deviations. Future work will further verify the generalization ability of the model and explore its application in AD subtype clustering and classification.

Normative Modeling using Multimodal Variational Autoencoders to Identify Abnormal Brain Structural Patterns in Alzheimer Disease

Multi-modal Variational Autoencoders for normative modelling across multiple imaging modalities

Normative Modeling via Conditional Variational Autoencoder and Adversarial Learning to Identify Brain Dysfunction in Alzheimer's Disease

Improving Normative Modeling for Multi-modal Neuroimaging Data using mixture-of-product-of-experts variational autoencoders

Analyzing heterogeneity in Alzheimer Disease using multimodal normative modeling on imaging-based ATN biomarkers

Anatomy-Guided Surface Diffusion Model for Alzheimer's Disease Normative Modeling

MC-RVAE: Multi-channel recurrent variational autoencoder for multimodal Alzheimer's disease progression modelling

Normative Modeling for AD Diagnosis and Biomarker Identification

Personalizing progressive changes to brain structure in Alzheimer's disease using normative modeling

Individualized multi-horizon MRI trajectory prediction for Alzheimer's Disease

Uncovering neural substrates across Alzheimer's disease stages using contrastive variational autoencoder

Multimodal deep learning models for early detection of Alzheimer’s disease stage

An interpretable generative multimodal neuroimaging-genomics framework for decoding Alzheimer's disease

Normative brain mapping of 3-dimensional morphometry imaging data using skewed functional data analysis

Incomplete multi-modal representation learning for Alzheimer's disease diagnosis

Transformer-based normative modelling for anomaly detection of early schizophrenia

Self-supervised multimodal neuroimaging yields predictive representations for a spectrum of Alzheimer's phenotypes

Disease‐driven domain generalization for neuroimaging‐based assessment of Alzheimer's disease

Automated detection of Alzheimer's disease: a multi-modal approach with 3D MRI and amyloid PET

Multimodal 3D Deep Learning for Early Diagnosis of Alzheimer's Disease

DeepAD: A Robust Deep Learning Model of Alzheimer's Disease Progression for Real-World Clinical Applications