Abstract:Alzheimer's disease (AD) is the most common age-related dementia. Mild cognitive impairment (MCI) is the early stage of cognitive decline before AD. It is crucial to predict the MCI-to-AD conversion for precise management, which remains challenging due to the diversity of patients. Previous evidence shows that the brain network generated from diffusion MRI promises to classify dementia using deep learning. However, the limited availability of diffusion MRI challenges the model training. In this study, we develop a self-supervised contrastive learning approach to generate structural brain networks from routine anatomical MRI under the guidance of diffusion MRI. The generated brain networks are applied to train a learning framework for predicting the MCI-to-AD conversion. Instead of directly modelling the AD brain networks, we train a graph encoder and a variational autoencoder to model the healthy ageing trajectories from brain networks of healthy controls. To predict the MCI-to-AD conversion, we further design a recurrent neural networks based approach to model the longitudinal deviation of patients' brain networks from the healthy ageing trajectory. Numerical results show that the proposed methods outperform the benchmarks in the prediction task. We also visualize the model interpretation to explain the prediction and identify abnormal changes of white matter tracts.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to predict the conversion from mild cognitive impairment (MCI) to Alzheimer's disease (AD). Specifically, the researchers developed a self - supervised contrastive learning method to generate structural brain networks from conventional anatomical MRI and used these networks to train a learning framework to predict the MCI - to - AD conversion. Unlike directly modeling AD brain networks, they trained a graph encoder and a variational auto - encoder to model the healthy aging trajectory of the healthy control group. To predict the MCI - to - AD conversion, a method based on recurrent neural networks was further designed to model the longitudinal deviation of the patient's brain network relative to the healthy aging trajectory. Through this method, the researchers hope to improve the prediction accuracy and identify the abnormal changes that lead to the MCI - to - AD conversion. ### Main contributions: 1. **Cross - modal learning method**: Extract features from anatomical MRI to generate brain networks, and at the same time extract features related to white matter tracts under the guidance of diffusion MRI. 2. **Generation method**: Use graph neural networks and variational auto - encoders to predict brain network features in the healthy aging trajectory. 3. **Recurrent learning algorithm**: Predict future disease states based on the longitudinal deviation of patients. 4. **Explanation method**: By comparing the actual diseased brain network and the predicted healthy aging brain network, identify the abnormal changes introduced by the MCI - to - AD conversion. ### Method overview: - **Data preparation**: Use the longitudinal MRI data set in the ADNI database, including T1 - weighted images and diffusion MRI of AD, MCI, and healthy control group (CN). - **Brain network construction**: Combine the Desikan gray matter node atlas and the IIT white matter atlas to extract node and edge features from T1 and FA images. - **Learning framework**: - **Graph encoder**: Pre - train graph neural networks (GNN) to extract dementia - related features from brain networks. - **Variational auto - encoder (VAE)**: Model the healthy aging trajectory and predict future brain network features. - **Recurrent neural network (RNN)**: Predict the MCI - to - AD conversion based on the patient's longitudinal deviation. ### Results: - **Classification performance**: The generated brain network performs best in the AD/CN classification task, with an accuracy rate of 0.861. - **Prediction performance**: In the task of predicting the MCI - to - AD conversion, the proposed model has the highest prediction accuracy within 18 months, reaching 0.839. - **Explanation results**: By visualizing the edge features with high residuals, it is found that this method can capture the abnormal changes in the brain network during the MCI - to - AD conversion, especially the white matter hyperintensities related to cognitive decline. ### Discussion and conclusion: This study proposes a new method to predict the MCI - to - AD conversion by modeling the healthy aging trajectory. The experimental results show that this method is superior to the benchmark methods in both classification and prediction tasks, and can explain the prediction results and identify abnormal brain structural changes. Future research can further integrate different training stages and introduce quantitative model explanation methods to improve the practicality and interpretability of the model.

Predicting conversion of mild cognitive impairment to Alzheimer's disease

Prediction of Conversion from Mild Cognitive Impairment to Alzheimer's Disease Using MRI and Structural Network Features.

Predicting Conversion of Mild Cognitive Impairments to Alzheimer's Disease and Exploring Impact of Neuroimaging

Prediction Of Mild Cognitive Impairment to Alzheimer’s Disease Conversion Via Machine Learning

A Two-Stage Model for Predicting Mild Cognitive Impairment to Alzheimer's Disease Conversion.

Predictive Modeling Of Alzheimer's Disease Prognosis Using Anatomical & Diffusion MRI

Deep learning prediction of mild cognitive impairment conversion to Alzheimer’s disease at 3 years after diagnosis using longitudinal and whole-brain 3D MRI

Prediction of conversion from amnestic mild cognitive impairment to alzheimer’s disease based on brain structural connectome

Predicting conversion from MCI to AD by integration of rs-fMRI and clinical information using 3D-convolutional neural network

Deep Learning Approaches for Early Prediction of Conversion from MCI to AD using MRI and Clinical Data: A Systematic Review

Predicting MCI to AD Conversation Using Integrated sMRI and rs-fMRI: Machine Learning and Graph Theory Approach

A Novel Conversion Prediction Method of MCI to AD Based on Longitudinal Dynamic Morphological Features Using ADNI Structural MRIs.

Temporal and Spatial Analysis of Alzheimer’s Disease Based on an Improved Convolutional Neural Network and a Resting-State FMRI Brain Functional Network

Convolutional Neural Networks-Based MRI Image Analysis for the Alzheimer’s Disease Prediction From Mild Cognitive Impairment

Predicting Conversion from MCI to AD Combining Multi-Modality Data and Based on Molecular Subtype

Predicting conversion from MCI to AD using resting-state fMRI, graph theoretical approach and SVM

Predicting conversion to Alzheimer's disease in individuals with Mild Cognitive Impairment using clinically transferable features

Machine learning framework for early MRI-based Alzheimer's conversion prediction in MCI subjects

Risk Prediction of Alzheimer's Disease Conversion in Mild Cognitive Impaired Population Based on Brain Age Estimation

Deep Generative Transfer Learning Predicts Conversion to Alzheimer'S Disease from Neuroimaging Genomics Data.

A Transformer-based Multi-features Fusion Model for Prediction of Conversion in Mild Cognitive Impairment