Leakage-Controlled Multimodal Learning for Diagnosis and Progression Prediction in Alzheimer's Disease Research

Key points

• This study presents a multimodal multitask framework combining an adapted SFCN MRI encoder, four causal clinical Transformers, shared fusion and task-specific ODE-GRU dynamics.1
• Fine-tuning and LoRA adapt the final two MRI blocks.1
• Branch-specific input controls, subject-grouped partitions and empirical causality checks support longitudinal evaluation.1
• Across 2,649 subjects and 17,317 visits from ADNI, OASIS-2 and MIRIAD, internal validation yields diagnosis, stage-1 progression and first-stage-1-visit progression AUROCs of 0.935 +/- 0.002, 0.884 +/- 0.003 and 0.870 +/- 0.005, respectively (mean +/- SD across three seeds).1

Sources (1)

• [1]Leakage-Controlled Multimodal Learning for Diagnosis and Progression Prediction in Alzheimer's Disease Research
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 03:38 PM

  Alzheimer's disease prediction involves irregular visits, heterogeneous measurements and incomplete modalities.

  This study presents a multimodal multitask framework combining an adapted SFCN MRI encoder, four causal clinical Transformers, shared fusion and task-specific ODE-GRU dynamics.

Extractive summary: sentences quoted from the sources.