动态调整时间尺度的医学信号模型，提升脑电分析精度与速度
NAKUL-Med: Spectral-Graph State Space Models with Dynamics Kernels for Medical Signals
- 通过可变核大小分支与元网络，自适应选择时间尺度
- 在脑电运动想象任务中达91.7%准确率，参数少28%，推理快一倍
- 融合频谱上下文与电极拓扑，适合脑电、心电等多通道医疗信号
状态空间模型（SSMs）虽具线性时间复杂度，但在多通道生理信号分析中存在三大局限：固定核无法捕捉多尺度时序动态（毫秒级准备期与数十毫秒执行瞬变），马尔可夫状态更新限制周期振荡的全局上下文建模，且通道独立处理忽略了电极空间拓扑。本文提出NAKUL，通过三方面改进：（1）动态核生成——并行不同核大小（3,5,7,11步）的SSM分支，由元网络根据输入统计加权，实现自适应时序尺度选择；（2）频谱上下文建模——基于FFT的可学习高斯频带滤波器，以$O(N \log N)$复杂度捕获全局周期模式；（3）图引导空间注意力——利用固定电极拓扑为多头注意力提供空间偏置，实现有原则的跨通道交互。在BCI竞赛IV-2a运动想象任务（主要基准）中，NAKUL达91.7±0.6%准确率，接近EEG-Conformer（92.1±0.7%），但参数减少28%（2.5M vs 3.5M），推理速度提升2.0×（4.3ms vs 8.7ms）。模型泛化至脑电情绪识别（83.6%）、多模态脑电-fMRI（91.4%）及医学影像（超声92.8%），展现架构通用性。消融实验显示动态核贡献+2.6%性能，且其选择模式与已知神经动力学相关联。
原文摘要 · Abstract (English)
State space models (SSMs) achieve linear-time complexity but struggle with multi-channel physiological signals due to three limitations: fixed kernels cannot capture multi-scale temporal dynamics (motor preparation over hundreds of milliseconds vs. execution transients in tens of milliseconds), Markovian state updates restrict global context for periodic oscillations, and channel-independent processing ignores spatial electrode topology. We introduce NAKUL, extending SSMs for medical signal analysis through three contributions: (1) Dynamic Kernel Generation-parallel SSM branches with varying kernel sizes (3, 5, 7, 11 timesteps) are weighted by a meta-network that analyzes input statistics, enabling adaptive temporal scale selection; (2) Spectral Context Modeling-FFT-based operations with learnable Gaussian frequency band filters capture global periodic patterns in $O(N \log N)$ complexity; (3) Graph-Guided Spatial Attention-fixed electrode topology provides spatial biases to multi-head attention for principled cross-channel interaction. On BCI Competition IV-2a motor imagery (our primary benchmark), NAKUL achieves 91.7$\pm$0.6\% accuracy, matching EEG-Conformer (92.1$\pm$0.7\%) while using 28\% fewer parameters (2.5M vs 3.5M) and 2.0$\times$ faster inference (4.3ms vs 8.7ms). The model generalizes to EEG emotion recognition (83.6\%), multimodal EEG-fMRI (91.4\%), and medical imaging (92.8\% on ultrasound), demonstrating architectural versatility. Ablations show dynamic kernels contribute +2.6\% and exhibit interpretable scale selection patterns correlated with known neural dynamics.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。