SC-TauPath 结构连接归因框架用于映射阿尔茨海默病 Tau 传播路径。结合网络扩散模型增强 MLP 与梯度×输入归因,生成多尺度路径图谱(骨干边、高流量路由、枢纽 ROI),验证 Braak 分期解剖学。
Scanned 9/11/2026
Install to Claude Code
npx -y skills add hiyenwong/ai_collection --skill sc-taupath-alzheimer-tau-propagation --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Sc Taupath Alzheimer Tau Propagation?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/hiyenwong-sc-taupath-alzheimer-tau-propagation)More formats (shields.io, HTML) on the badges page.
---
name: sc-taupath-alzheimer-tau-propagation
description: SC-TauPath 结构连接归因框架用于映射阿尔茨海默病 Tau 传播路径。结合网络扩散模型增强 MLP 与梯度×输入归因,生成多尺度路径图谱(骨干边、高流量路由、枢纽 ROI),验证 Braak 分期解剖学。
version: 1.0.0
category: neuroscience
authors:
- Jing Zhang
- Norman Scheel
- Minheng Chen
arxiv_id: 2606.04066
created: 2026-06-09
activation_keywords:
- tau propagation
- Alzheimer's disease
- structural connectivity
- network diffusion
- attribution framework
- Braak staging
related_skills:
- brain-network-controllability
- alzheimer-pet-suvr-network-models
---
# SC-TauPath: Structural Connectivity Attribution Framework for Alzheimer's Tau Propagation
## Overview
SC-TauPath 是一个结构连接(SC)归因框架,从体内神经影像数据映射 Tau 传播路径。结合网络扩散模型增强的 MLP 与梯度×输入归因,生成多尺度路径图谱,验证 Braak 分期解剖学。
## Core Methodology
### 1. Network Diffusion Model-Augmented MLP
结合网络扩散理论的可解释神经网络架构:
- 输入:DTI 结构连接矩阵 + 18F-Flortaucipir PET Tau 分布
- 增强:网络扩散模型(NDM)约束损失函数
- 预测:区域 Tau 分布预测(交叉验证)
### 2. Gradient × Input Attribution
可解释的边贡献评分:
- 物理意义:每条 SC 边对 Tau 预测的贡献度
- 归因映射:将评分转换为多尺度路径图
### 3. Multi-Scale Pathway Maps
三级路径图谱:
- Backbone edges:高归因评分的核心传播通道
- High-traffic routes:多边组合的高流量路径
- Hub ROIs:关键传播枢纽节点
### 4. Braak Staging Validation
与经典 Braak 分期的解剖学一致性验证:
- Braak I-II:内嗅皮层-海马通路
- Braak III-IV:杏仁核-颞叶新皮层扩展
- Braak V-VI:新皮层广泛传播
## Implementation Workflow
### Step 1: Data Preparation
需要的数据:
- DTI structural connectivity matrix (N×N)
- 18F-Flortaucipir PET Tau SUVR maps
- ROI parcellation (e.g., AAL, Desikan-Killiany)
### Step 2: NDM-Augmented MLP Training
架构设计:
```
Input: SC matrix (N×N)
↓
SC Encoder (flatten → hidden_dim)
↓
NDM Constraint Loss (network diffusion theory)
↓
Predictor (hidden_dim → N ROI tau values)
↓
Output: Tau distribution prediction
```
### Step 3: Attribution Computation
算法流程:
1. 计算预测输出对 SC 输入的梯度
2. Gradient × Input 得到边归因评分
3. 重塑为 N×N 归因矩阵
### Step 4: Pathway Extraction
提取策略:
- extract_backbone_edges: percentile threshold (e.g., top 30%)
- identify_hub_rois: top-k ROI importance
- extract_high_traffic_routes: multi-hop path aggregation
## Key Findings
### Cross-Validated Performance
- 234 ADNI 参与者的交叉验证 Tau 预测
- 结构连接编码区域 Tau 分布的空间特异性信息
### Anatomical Consistency
归因路径图与 Braak 分期解剖学的对应:
- 传播起点:内嗅皮层 → 海马(Braak I-II)
- 关键枢纽:杏仁核、颞叶内侧结构
- 扩展路径:额叶、顶叶新皮层(Braak V-VI)
## Applications
1. Tau 传播机制理解:从结构连接角度解释疾病进展
2. 治疗靶点识别:高归因枢纽节点作为干预候选
3. 早期诊断:Tau 传播路径预测用于风险评估
4. 临床试验设计:基于传播路径的患者分层
## Advantages Over Existing Methods
- Biophysical models:Heavy assumptions, Low interpretability
- Pure ML models:No assumptions, Low interpretability
- SC-TauPath:Light NDM constraint, High interpretability, Multi-scale pathway maps
## Limitations
1. DTI 限制:结构连接估计的可靠性依赖 DTI 质量
2. Tau PET 成本:18F-Flortaucipir 成本较高
3. 静态模型:未捕获动态传播过程
4. 样本依赖:ADNI 样本可能不代表所有 AD 亚型
## Future Directions
1. 纵向数据整合:添加时间序列 Tau PET 数据
2. 多模态融合:结合功能连接、代谢成像
3. 个体化路径:开发个性化传播预测模型
4. 药物试验应用:用于抗 Tau 药物试验的患者选择
## References
- arXiv:2606.04066
- Braak H, Braak E. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol (1991)
- Iturria-Medina Y, et al. Network diffusion model of brain atrophy in Alzheimer's disease. NeuroImage (2017)
## Related Work
- brain-network-controllability:网络控制理论用于脑状态转移
- alzheimer-pet-suvr-network-models:Alzheimer PET 网络建模
- time-varying-brain-connectivity:时变脑连接分析方法Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!