[论文解读] BrainCast: A Spatio-Temporal Forecasting Model for Whole-Brain fMRI Time Series Prediction

Yunlong Gao, Jinbo Yang+6 more|arXiv (Cornell University)|Mar 9, 2026

Functional Brain Connectivity Studies被引用 0

一句话总结

BrainCast 通过联合建模 ROIs 间的空间交互与 ROI 内部的时间动态来预测全脑 fMRI 序列，相较基线提高预测准确性并提升下游认知能力预测。

ABSTRACT

Functional magnetic resonance imaging (fMRI) enables noninvasive investigation of brain function, while short clinical scan durations, arising from human and non-human factors, usually lead to reduced data quality and limited statistical power for neuroimaging research. In this paper, we propose BrainCast, a novel spatio-temporal forecasting framework specifically tailored for whole-brain fMRI time series forecasting, to extend informative fMRI time series without additional data acquisition. It formulates fMRI time series forecasting as a multivariate time series prediction task and jointly models temporal dynamics within regions of interest (ROIs) and spatial interactions across ROIs. Specifically, BrainCast integrates a Spatial Interaction Awareness module to characterize inter-ROI dependencies via embedding every ROI time series as a token, a Temporal Feature Refinement module to capture intrinsic neural dynamics within each ROI by enhancing both low- and high-energy temporal components of fMRI time series at the ROI level, and a Spatio-temporal Pattern Alignment module to combine spatial and temporal representations for producing informative whole-brain features. Experimental results on resting-state and task fMRI datasets from the Human Connectome Project demonstrate the superiority of BrainCast over state-of-the-art time series forecasting baselines. Moreover, fMRI time series extended by BrainCast improve downstream cognitive ability prediction, highlighting the clinical and neuroscientific impact brought by whole-brain fMRI time series forecasting in scenarios with restricted scan durations.

研究动机与目标

• 在数据获取受限时，激励扩展和预测全脑 fMRI 序列。
• 将 fMRI 预测表述为跨 ROI 的多变量时序预测问题。
• 联合建模跨 ROI 的空间依赖与 ROI 内部的时间动态。
• 开发一个统一架构（SIA、TFR、SPA）以生成用于预测的全脑信息特征。

提出的方法

• 通过共享的 MLP 将每个 ROI 时序嵌入为一个令牌，形成 X_E ∈ R^{N×D}。
• 使用含 Fourier Analysis Network 的 SIAformer 层、自注意力和 FFN 来学习跨 ROI 依赖并产生 H^{spat}。
• 将 Amplifier 作为Temporal Feature Refinement 模块以在时域谱上进行能量均衡与分解，捕捉低能量与高能量分量以形成 H^{temp}。
• 通过时空模式对齐来对齐空间与时间表示，获得 H^{global}。
• 用来自 H^{global} 的线性预测头预测未来 fMRI 序列到 X̂_F，并以 MSE 损失优化。

实验结果

研究问题

• RQ1 全脑 fMRI 序列在短扫描时长下是否能够提升数据质量？
• RQ2 如何联合建模跨 ROI 的空间交互与 ROI 内部的时间动态以实现准确预测？
• RQ3 当 fMRI 数据受限时，时空表示是否能提升下游认知能力预测？
• RQ4 各模块（SIA、TFR、SPA）对预测性能的贡献各自为何？

主要发现

数据集	指标	Ours	iTransformer	DLinear	Amplifier	PatchTST	CrossFormer	FourierGNN	LSTM	R	R2
HCP-rs-fMRI	MSE (lower is better)	0.380 (0.004)	0.398 (0.002)	0.529 (0.012)	0.407 (0.006)	0.394 (0.003)	0.417 (0.001)	0.562 (0.005)	0.592 (0.006)	-	-	-
HCP-rs-fMRI	MAE (lower is better)	0.461 (0.004)	0.478 (0.002)	0.562 (0.009)	0.481 (0.005)	0.473 (0.004)	0.488 (0.003)	0.580 (0.004)	0.597 (0.008)	-	-	-
HCP-rs-fMRI	R (higher is better)	0.665 (0.002)	0.631 (0.001)	0.446 (0.013)	0.627 (0.006)	0.635 (0.003)	0.628 (0.004)	0.432 (0.002)	0.355 (0.008)	-	-	-
HCP-rs-fMRI	R² (higher is better)	0.438 (0.003)	0.397 (0.003)	0.199 (0.011)	0.390 (0.003)	0.402 (0.005)	0.396 (0.006)	0.187 (0.003)	0.120 (0.007)	-	-	-
HCP-t-fMRI	MSE (lower is better)	0.435 (0.004)	0.488 (0.004)	0.633 (0.005)	0.460 (0.009)	0.485 (0.006)	0.480 (0.010)	0.705 (0.007)	0.681 (0.005)	-	-	-
HCP-t-fMRI	MAE (lower is better)	0.484 (0.006)	0.501 (0.002)	0.587 (0.007)	0.514 (0.015)	0.501 (0.003)	0.496 (0.008)	0.621 (0.005)	0.609 (0.010)	-	-	-
HCP-t-fMRI	R (higher is better)	0.595 (0.003)	0.583 (0.005)	0.376 (0.004)	0.582 (0.021)	0.586 (0.005)	0.589 (0.017)	0.208 (0.009)	0.292 (0.007)	-	-	-
HCP-t-fMRI	R² (higher is better)	0.351 (0.004)	0.338 (0.002)	0.142 (0.003)	0.333 (0.018)	0.341 (0.008)	0.347 (0.013)	0.043 (0.004)	0.076 (0.002)	-	-	-

• BrainCast 在 HCP 静息态与任务态 fMRI 数据集上，在多项指标上超过最先进的基线。
• 用 BrainCast 延长 fMRI 序列可提升认知能力预测效果。
• 消融实验显示三大模块（SIA、TFR、SPA）均对性能有贡献，且 TFR 的单独影响最大。
• 可视化的皮层注意力图揭示了与任务相关的有意义的空间模式及静息态与任务态之间的半球差异。
• 基于 Transformer 的基线在此任务中通常优于 RNN/LSTM 和部分 GNN 基线。