Source Reconstructed MEG Data for Adaptive Circuit Dynamics Across Human Cortex During Evidence Accumulation in Changing Environments

This dataset contains source reconstructed MEG data for: Murphy PR, Wilming N, Hernandez Bocanegra DC, Prat Ortega G &amp; Donner TH (2021). Adaptive circuit dynamics across human cortex during evidence accumulation in changing environments. <em>Nature Neuroscience</em>. Online ahead of print.<br><br> Each "*source_reconstructions*" .zip contains files for trial onset-aligned epochs (full-length trials composed of 12 evidence samples only), separately for low (1-35 Hz in steps of 1 Hz; "LF") and high (36-160 Hz in steps of 4 Hz; "HF") frequency TFR decompositions. Furthermore, each session is spread over a number of files that contain 100 trials each. Files from one epoch type can be safely concatenated in pandas.<br>Individual files can be read by using `pandas.read_hdf`. This will return a table that contains individual ROIs as columns and a multi-index that labels each data point. Specifically, the index contains a trial identifier ('trial'), a time identifier ('time', seconds relative to trial onset), an identifier for the TFR settings ('est_key') and a frequency identifier ('est_val'). See https://github.com/DonnerLab/2021_Murphy_Adaptive-Circuit-Dynamics-Across-Human-Cortex/tree/main/source_reconstruct/pymeg for code that makes and further processes datasets of this form. They are made with lcmv_peter.py and an example of further processing is sr_agg_parallel.py (in this case, aggregation of reconstructed over vertices within specified ROIs).<br>Each “*sr_behav.zip” contains behavioural (‘choices’), task (sample locations: ‘stimIn’; change-point positions: ‘pswitch’; generative distributions at end of each trial: ‘fdist’; and generative distributions per sample position: ‘distseq’) and minimal eye-tracking data (‘pupil’, ‘Xgaze’, ‘Ygaze’, all from only 0.57 s following sample onset) from the same trials in the source reconstructed datasets. Use the ‘trialID’ variable in combination with the ‘trial’ identifier in the source reconstructed datasets to align trials.<br>

本数据集包含经源重构的脑磁图（Magnetoencephalography, MEG）数据，对应研究：Murphy PR、Wilming N、Hernandez Bocanegra DC、Prat Ortega G及Donner TH于2021年发表于《自然-神经科学》（Nature Neuroscience）的论文《变化环境下证据累积过程中人脑皮层的自适应环路动力学》，目前处于在线优先出版状态。

每个名为`source_reconstructions`的压缩包内含与试次起始对齐的时段数据（仅由12个证据样本构成的完整试次），分别对应低频（1~35 Hz，步长1 Hz，标记为"LF"）与高频（36~160 Hz，步长4 Hz，标记为"HF"）的时频域分解（Time-Frequency Representation, TFR）结果。此外，每个实验会话的数据被拆分至多个文件中，每个文件包含100个试次。同一类时段数据可通过pandas库安全拼接。

单个文件可通过`pandas.read_hdf`读取，返回的表格将以各感兴趣脑区（Region of Interest, ROI）作为列，并附带用于标记每个数据点的多级索引。具体而言，索引包含试次标识符（`trial`）、时间标识符（`time`，即相对于试次起始的秒数）、时频域设置标识符（`est_key`）以及频率标识符（`est_val`）。可参考以下链接中的代码生成并进一步处理此类格式的数据集：https://github.com/DonnerLab/2021_Murphy_Adaptive-Circuit-Dynamics-Across-Human-Cortex/tree/main/source_reconstruct/pymeg。该数据集由`lcmv_peter.py`生成，进一步处理的示例可参考`sr_agg_parallel.py`（本例用于对指定感兴趣脑区内的顶点重构结果进行聚合）。

每个名为`sr_behav.zip`的压缩包包含与源重构数据集对应试次完全一致的行为学数据（`choices`）、任务相关数据（样本位置：`stimIn`；变点位置：`pswitch`；每个试次结束时的生成分布：`fdist`；每个样本位置对应的生成分布：`distseq`）以及精简眼动数据（`pupil`、`Xgaze`、`Ygaze`，仅采集自样本起始后0.57秒的时段）。可通过`trialID`变量与源重构数据集中的`trial`标识符对齐对应试次。