Abstract:Alzheimer’s disease (AD), the most common type of dementia, not only imposes a huge financial burden on the health care system, but also a psychological and emotional burden on patients and their families. There is thus an urgent need to infer trajectories of cognitive performance over time and identify biomarkers predictive of the progression. In this article, we propose the multi-task learning with fused Laplacian sparse group lasso model, which can identify biomarkers closely related to cognitive measures due to its sparsity-inducing property, and model the disease progression with a general weighted (undirected) dependency graphs among the tasks. An efficient alternative directions method of multipliers based optimization algorithm is derived to solve the proposed non-smooth objective formulation. The effectiveness of the proposed model is demonstrated by its superior prediction performance over multiple state-of-the-art methods and accurate identification of compact sets of cognition-relevant imaging biomarkers that are consistent with prior medical studies.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the prediction of cognitive performance in Alzheimer's disease (AD) and the modeling of its disease progression. Specifically, the authors propose a multi - task learning model - Fused Laplacian Sparse Group Lasso (FL - SGL), aiming to improve the prediction performance by exploiting the correlations between tasks and identify biomarkers closely related to cognitive performance. ### Background and Motivation Alzheimer's disease is the most common type of dementia. It not only brings a huge economic burden to the medical system but also causes great psychological and emotional pressure on patients and their families. Therefore, it is particularly important to accurately predict the cognitive performance trajectory of patients and identify biomarkers related to disease progression. ### Research Objectives 1. **Cognitive Performance Prediction**: Through the multi - task learning model, accurately predict the changes in cognitive performance of Alzheimer's disease patients over time. 2. **Biomarker Identification**: Identify imaging biomarkers closely related to cognitive performance, which can be used for early diagnosis and treatment evaluation of the disease. 3. **Disease Progression Modeling**: By constructing a dependency graph between tasks, better capture the correlations between tasks at different time points, and thus more accurately model disease progression. ### Method Innovations - **Multi - task Learning**: The FL - SGL model processes cognitive performance prediction tasks at multiple time points simultaneously through a multi - task learning framework. - **Fused Laplacian Sparse Group Lasso**: Introduce the Fused Laplacian Sparse Group Lasso regularization term, which encourages related tasks to have similar parameters while allowing each task to select a specific feature set. - **Time - Dependent Modeling**: Use the Gaussian kernel function to define the weights between tasks and construct a weighted undirected dependency graph to capture the complex dependency relationships between tasks. ### Model Formula The optimization problem of the FL - SGL model can be expressed as: \[ \min_{\Theta, \Gamma} \sum_{t = 1}^T \left\| y_t - X_t \theta_t \right\|^2+R_{\lambda_1, \lambda_2}(\Theta)+\lambda_3 \left\| \Gamma \right\|_1 \] where: - \( \Theta \in \mathbb{R}^{p \times T} \) is the regression parameter matrix of all tasks. - \( \Gamma \in \mathbb{R}^{p \times T} \) is the residual matrix. - \( R_{\lambda_1, \lambda_2}(\Theta)=\lambda_1 \left\| \Theta \right\|_1+\lambda_2 \left\| \Theta \right\|_{2, 1} \) is the sparse group lasso regularization term. - \( \gamma_t=\theta_t - \hat{\theta}_t=\theta_t - \sum_{\ell = 1}^T w_{\ell, t} \theta_\ell \) is the residual term, and \( w_{\ell, t} \) is the Gaussian kernel weight. ### Experimental Verification The paper conducted extensive experiments using the ADNI dataset to verify the superiority of the FL - SGL model in prediction performance and biomarker identification. The experimental results show that the FL - SGL model is significantly superior to the baseline methods in aggregated performance and can accurately identify stable MRI features related to cognitive performance. ### Conclusion The FL - SGL model successfully improves the accuracy of Alzheimer's disease cognitive performance prediction and identifies biomarkers related to disease progression through multi - task learning and Fused Laplacian Sparse Group Lasso regularization. This research provides new tools and methods for the early diagnosis and treatment of Alzheimer's disease.

Modeling Alzheimer’s Disease Progression with Fused Laplacian Sparse Group Lasso

Modeling Disease Progression via Fused Sparse Group Lasso

Group Guided Fused Laplacian Sparse Group Lasso for Modeling Alzheimer's Disease Progression.

Fused Group Lasso Regularized Multi-Task Feature Learning and Its Application to the Cognitive Performance Prediction of Alzheimer’s Disease

Modeling disease progression via multi-task learning

Group Guided Sparse Group Lasso Multi-task Learning for Cognitive Performance Prediction of Alzheimer’s Disease

High-Order Multi-Task Feature Learning to Identify Longitudinal Phenotypic Markers for Alzheimer's Disease Progression Prediction.

Rethinking modeling Alzheimer's disease progression from a multi-task learning perspective with deep recurrent neural network

Spatio-Temporal Similarity Measure based Multi-Task Learning for Predicting Alzheimer's Disease Progression using MRI Data

Modeling Alzheimer's disease cognitive scores using multi-task sparse group lasso

Evaluating the Predictive Power of Multivariate Tensor-based Morphometry in Alzheimers Disease Progression via Convex Fused Sparse Group Lasso.

Modeling Alzheimers' Disease Progression from Multi-task and Self-supervised Learning Perspective with Brain Networks

Dual Feature Correlation Guided Multi-Task Learning for Alzheimer's Disease Prediction.

Predicting Cognitive Declines Using Longitudinally Enriched Representations For Imaging Biomarkers

Spatio-temporal Tensor Multi-Task Learning for Predicting Alzheimer's Disease in a Longitudinal study

Correlation-aware Sparse and Low-rank Constrained Multi-task Learning for Longitudinal Analysis of Alzheimer's Disease.

Sparse Shared Structure Based Multi-Task Learning for MRI Based Cognitive Performance Prediction of Alzheimer's Disease.

Dynamic Survival Prediction Using Sparse Longitudinal Images via Multi-Dimensional Functional Principal Component Analysis

Feature-aware Multi-task Feature Learning for Predicting Cognitive Outcomes in Alzheimer's Disease

Predicting Alzheimer's Disease Cognitive Assessment Via Robust Low-Rank Structured Sparse Model.

Sparse Learning and Hybrid Probabilistic Oversampling for Alzheimer’s Disease Diagnosis