Multivariate functional mixed model with MRI data: An application to Alzheimer's disease.

Area of Science:

• Neuroscience
• Biostatistics
• Medical Imaging

Background:

• Alzheimer's Disease (AD) is a leading cause of dementia, necessitating early intervention strategies for mild cognitive impairment (MCI) patients.
• Multimodal data, including neuroimaging and longitudinal assessments, are crucial for understanding AD progression.
• Accurate personalized prediction models are needed to guide interventions for MCI subjects.

Purpose of the Study:

• To propose a novel multivariate functional mixed model with MRI data (MFMM-MRI) for predicting dementia onset in MCI subjects.
• To investigate two functional forms (random-effects and instantaneous models) for linking longitudinal and survival processes.
• To develop a dynamic prediction framework for personalized longitudinal trajectories and survival probabilities.

Main Methods:

• Developed the MFMM-MRI model integrating longitudinal neurological assessments, baseline MRI data, and dementia onset (survival outcome).
• Employed Markov Chain Monte Carlo (MCMC) with the No-U-Turn Sampling (NUTS) algorithm for posterior sample estimation.
• Applied the model to data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) study.

Main Results:

• Identified significant associations between longitudinal outcomes, MRI data, and dementia onset risk in MCI subjects.
• The instantaneous model utilizing whole-brain voxels demonstrated superior prediction performance.
• Simulation studies validated the model's estimation and dynamic prediction capabilities.

Conclusions:

• The MFMM-MRI model provides accurate personalized predictions for MCI patients, aiding in early intervention planning.
• Integrating multimodal data, particularly MRI, significantly enhances AD progression prediction.
• The developed dynamic prediction framework offers valuable insights into individual patient trajectories and dementia risk.