摘要：阿尔茨海默病预测面临随访时间不规律、测量数据异质以及模态不完整等问题。本研究提出一个多模态多任务框架，结合经适配的 SFCN MRI 编码器、四个因果临床 Transformer、共享融合机制和任务特异性 ODE-GRU 动态模型。研究通过微调和 LoRA 适配 MRI 模型的最后两个模块；Task-DRO 用于平衡各任务损失，Group-CVaR 则面向队列和共病分层进行优化。分支特异性输入控制、按受试者分组的数据划分以及经验性因果检查共同支持纵向评估。在来自 ADNI、OASIS-2 和 MIRIAD 的 2,649 名受试者、17,317 次访视数据上，内部验证得到诊断、阶段1进展和首次阶段1访视进展的 AUROC 分别为 0.935 +/- 0.002、0.884 +/- 0.003 和 0.870 +/- 0.005（3 个随机种子的均值 +/- 标准差）。对应的混合模型 AUROC 分别为 0.951、0.909 和 0.896。下一次访视 MMSE 的平均绝对误差（MAE）为 1.61 分；最差分层的诊断 AUROC 为 0.827 +/- 0.008。对 ADNI 抽样样本的解释结果识别出任务特异性的输入依赖关系。OASIS-3 外部验证中，网络模型和混合模型的诊断 AUROC 分别为 0.763 和 0.767，混合模型的下一次访视进展 AUROC 为 0.764，诊断校准误差从 0.197 降至 0.052，下一次访视 MMSE 的 MAE 为 0.86。针对六个组件进行的 seed-42 配对消融实验显示，汇总诊断和进展 AUROC 的差异介于 -0.004 至 +0.004 之间；移除临床编码器输入会使诊断 AUROC 降低 0.272。该框架在统一流程中整合了纵向预测、缺失模态处理、辅助共病建模和亚组评估。

Abstract: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. Fine-tuning and LoRA adapt the final two MRI blocks. Task-DRO balances task losses, while Group-CVaR targets cohort and comorbidity strata. Branch-specific input controls, subject-grouped partitions and empirical causality checks support longitudinal evaluation. 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). Corresponding hybrid AUROCs are 0.951, 0.909 and 0.896. Next-visit MMSE mean absolute error (MAE) is 1.61 points; worst-stratum diagnosis AUROC is 0.827 +/- 0.008. Sampled ADNI explanations identify task-specific input dependence. OASIS-3 external validation yields network and hybrid diagnosis AUROCs of 0.763 and 0.767, hybrid next-visit progression AUROC of 0.764, diagnosis calibration error decreasing from 0.197 to 0.052, and next-visit MMSE MAE of 0.86. Seed-42 paired ablations of six components yield pooled diagnosis and progression AUROC differences between -0.004 and +0.004; removing clinical encoder inputs lowers diagnosis AUROC by 0.272. The framework integrates longitudinal prediction, missing-modality handling, auxiliary comorbidity modelling and subgroup evaluation within a common pipeline.