科研 Skill
神经数据集陷阱审查
neuro-dataset-gotchas
Checklist for auditing neural datasets before training: rate and channel consistency, duplicate subjects, label mapping. Use before modeling on a public or lab dataset.
SKILL.md
Dataset audit before you train
Most surprising results on public datasets come from the data, not the model: mixed sampling rates, duplicated people, channel sets that change between files, evaluation labels that differ from training labels. Spend ten minutes on this audit before any modelling.
1. Run the audit
For each file or run, tabulate:
- sampling rate, number and names of channels, units, reference;
- duration, number of events per class, event code mapping;
- subject id, session id, site, date.
Then check:
- Rates and channels identical across files? If not, resample and select a common channel set explicitly, and say so.
- One person, one id? Merged or multi-site sets can list the same person twice. Compare demographics and feature similarity.
- Classes balanced per subject? Pooled balance can hide subjects with one class only, which breaks per-subject CV.
- Labels match the documentation? Recount events against the paper or README; check for off-by-one event codes and swapped class names.
- Train/evaluation split defined by the dataset? Use it for comparison with published numbers; competitions often released evaluation labels separately.
- Data quality: flat or saturated channels, long artifact periods, discontinuities, and recordings with long wake or idle periods.
Record exclusions with reasons in a table you keep with the results.
2. Known issues to check first
These are well-known pitfalls. Confirm each against the dataset's own documentation before acting, since versions change.
| Dataset | Check |
|---|---|
| BCI Competition IV 2a | 22 EEG + 3 EOG channels; separate training (T) and evaluation (E) sessions; evaluation labels shipped separately |
| PhysioNet EEG Motor Movement/Imagery | several subjects are commonly excluded for sampling or annotation irregularities (often S088, S092, S100, S104); runs mix executed and imagined tasks, so select runs by task |
| Sleep-EDF | long wake periods before and after sleep inflate accuracy; trim to sleep period plus margin; older stage 3 and 4 labels map to N3 |
| CHB-MIT | channel montage changes between files of the same patient; pediatric patients; seizure annotations per file |
| TUH EEG corpora | mixed montages and sampling rates across sessions; use the corpus's train/dev/eval split and patient ids |
| MOABB datasets | use MOABB's paradigms and evaluation classes so results compare with the literature |
3. When merging datasets
Harmonise rate, channels, reference, filtering and units first. Keep a dataset id column and test leave-one-dataset-out, because site effects are large and a model can learn the site.
Output
An audit table (file, rate, channels, duration, events per class, issues, action) plus the final list of subjects and runs used.
示例提示词
我下载了CHB-MIT癫痫数据集,帮我检查里面常见的标注和采样陷阱。
所属插件
安装该插件会一并获得同领域的 45 个 Skill / MCP。
同插件的其他科研 Skill
- Allen脑区本体查询Looks up Allen Institute brain region names, acronyms and atlas hierarchy, and picks a BrainGlobe reference atlas. Use when matching region labels across datasets or papers.
- 动物行为姿态追踪Tracks animal pose and behavior from video with DeepLabCut or SLEAP, and extracts kinematic and behavioral event features. Use for markerless tracking or scoring behavior.
- BIDS神经影像数据管理Converts raw EEG or fMRI data to BIDS, validates with bids-validator, and queries or manages datasets with pybids and datalad. Use when organizing or sharing neuroimaging data.
- Brian2脉冲神经网络仿真Simulates LIF, AdEx and Hodgkin-Huxley spiking neuron networks in Brian2, including STDP plasticity and recurrent dynamics. Use when building a spiking network model.
- 钙成像数据分析Processes two-photon or widefield calcium imaging with Suite2p or CaImAn, from motion correction through ROI extraction to dF/F and spike deconvolution. Use for calcium imaging movies.
- 功能连接组图论分析Computes functional connectivity matrices and graph-theory metrics such as modularity and path length from ROI time series. Use when analyzing brain network topology.
- MNE脑电脑磁分析Runs an MNE-Python EEG or MEG pipeline covering preprocessing, ICA cleanup, ERP and time-frequency analysis, and source localization. Use to analyze raw EEG/MEG recordings end to end.
- 神经解码无泄漏评估Guides leakage-free cross-validation of neural decoders, including subject grouping, nested CV, chance levels and confound checks. Use when training or reporting a BCI classifier.
- 电极通道映射校验Checks that EEG, ECoG or probe channel labels and positions match physiology using correlation-distance consistency. Use when montage labels may be swapped or bridged.
- 聚类置换统计检验Explains how to set up and correctly interpret cluster-based permutation tests and TFCE for ERP or time-frequency data. Use to avoid onset or localization overclaims.
- 神经连接性分析核查Guides choice of leakage-robust connectivity measures such as wPLI, imaginary coherency and AEC, with confound controls. Use when reviewing or planning a connectivity analysis.
- 神经数据档案库导航Lists neural data archives such as OpenNeuro, DANDI, PhysioNet and MOABB by paradigm, with access tiers and download guidance. Use when finding a dataset for a research question.
- DBS植入物传感数据分析Loads, time-aligns and analyzes sensing data from DBS and bidirectional implants such as Percept and Summit RC+S. Use with implant LFP exports, packet loss or aDBS biomarkers.
- 神经解码器嵌入式移植Ports a neural signal-processing or BCI decoder from Python or MATLAB to fixed-point embedded C on a microcontroller. Use when moving a decoder onto implant or headstage firmware.
- 神经设备法规合规映射Maps a neurotech device to the regulatory standards and testing it needs and builds a requirements-to-risk traceability matrix. Use when planning verification or a design review.
- EEG预处理流程审查Designs or audits an EEG or MEG preprocessing pipeline in MNE-Python, EEGLAB, FieldTrip or Brainstorm. Use when choosing or reviewing filter, ICA or epoching settings.
- 神经电极电化学表征Characterizes neural electrodes electrochemically via impedance spectroscopy, cyclic voltammetry and charge-injection limits. Use when comparing electrode coatings or charge capacity.
- 神经记录文件格式转换Opens, inspects and converts neural recording files from EEG, MEG, intracranial and fNIRS acquisition systems. Use when a reader errors or units, rates or events look wrong.
- 神经信号基础模型评估Fairly evaluates pretrained EEG or iEEG foundation models against classical baselines with matched data and splits. Use when fine-tuning or linear-probing a pretrained model.
- 神经记录前端电路设计Designs or reviews analog front ends for neural recording headstages, covering noise budget, CMRR and ADC resolution. Use when choosing an amplifier or debugging a noisy front end.
- 脑机解码器长期稳定性Evaluates intracortical BCI decoders across days using zero-shot and few-shot regimes to quantify drift and recalibration cost. Use to assess decoder stability over sessions.
- 植入设备功耗热预算Budgets power, data rate and tissue heating for implanted or wireless neural interfaces against thermal screening limits. Use when sizing a high-channel-count implant design.
- NeuroKit2生理信号处理Processes physiological biosignals such as ECG, EEG, EDA, RSP, EMG and EOG with NeuroKit2, including HRV and complexity metrics. Use for psychophysiology or multi-modal signal analysis.
- NeuroMorpho神经元形态检索Finds and compares digital neuron reconstructions on NeuroMorpho.Org by region, cell type or species and reads their morphometry. Use when real neuron morphologies are needed.
常见问题
神经数据集陷阱审查能做什么?
Checklist for auditing neural datasets before training: rate and channel consistency, duplicate subjects, label mapping. Use before modeling on a public or lab dataset。
如何安装神经数据集陷阱审查?
在科搭 Codar 左侧点击「插件」,在「组件」页签搜索「神经数据集陷阱审查」并安装;在 Claude Code 或 Codex 中可以发送本页的安装提示词,或下载 Skill 包后解压到 ~/.claude/skills 或 ~/.codex/skills。
神经数据集陷阱审查需要哪些权限?
该科研 Skill没有声明额外权限;执行具体操作时仍由 Agent 按自身规则请求确认。
神经数据集陷阱审查经过官方审核吗?
是。SpeedSkills 收录的每个版本都会记录内容摘要与权限声明,当前版本已通过安全审核。