说明
转到末尾下载完整示例代码。
将 Epochs 导出为 Pandas DataFrame#
本教程展示如何将 Epochs 对象中的数据导出为 Pandas DataFrame，并应用典型的 Pandas split-apply-combine（拆分-应用-合并）工作流，考察各个 epoch 与各种条件下反应最大值的潜伏期。
我们将使用 Sample 数据集，但加载的是原始文件的一个版本，该版本已经过滤波和降采样，并且其 EEG 通道已施加平均参考。和往常一样，我们先导入所需的模块并加载数据：
# Authors: The MNE-Python contributors.
# License: BSD-3-Clause
# Copyright the MNE-Python contributors.
import matplotlib.pyplot as plt
import seaborn as sns
import mne
sample_data_folder = mne.datasets.sample.data_path()
sample_data_raw_file = (
sample_data_folder / "MEG" / "sample" / "sample_audvis_filt-0-40_raw.fif"
)
raw = mne.io.read_raw_fif(sample_data_raw_file, verbose=False)
接下来，我们从文件中加载事件列表，用事件字典将其映射为条件名称，设置一些信号拒绝阈值（参见 Rejecting Epochs based on peak-to-peak channel amplitude），然后将连续数据分割为 epoch：
sample_data_events_file = (
sample_data_folder / "MEG" / "sample" / "sample_audvis_filt-0-40_raw-eve.fif"
)
events = mne.read_events(sample_data_events_file)
event_dict = {
"auditory/left": 1,
"auditory/right": 2,
"visual/left": 3,
"visual/right": 4,
}
reject_criteria = dict(
mag=3000e-15, # 3000 fT
grad=3000e-13, # 3000 fT/cm
eeg=100e-6, # 100 µV
eog=200e-6,
) # 200 µV
tmin, tmax = (-0.2, 0.5) # epoch from 200 ms before event to 500 ms after it
baseline = (None, 0) # baseline period from start of epoch to time=0
epochs = mne.Epochs(
raw,
events,
event_dict,
tmin,
tmax,
proj=True,
baseline=baseline,
reject=reject_criteria,
preload=True,
)
del raw
Not setting metadata
288 matching events found
Setting baseline interval to [-0.19979521315838786, 0.0] s
Applying baseline correction (mode: mean)
Created an SSP operator (subspace dimension = 4)
4 projection items activated
Loading data for 288 events and 106 original time points ...
Rejecting epoch based on EEG : ['EEG 003']
Rejecting epoch based on EEG : ['EEG 001', 'EEG 002', 'EEG 003', 'EEG 004', 'EEG 006', 'EEG 007']
Rejecting epoch based on EEG : ['EEG 001', 'EEG 002', 'EEG 003', 'EEG 007']
Rejecting epoch based on EEG : ['EEG 001']
Rejecting epoch based on EEG : ['EEG 003', 'EEG 007']
Rejecting epoch based on EEG : ['EEG 007']
Rejecting epoch based on EEG : ['EEG 003', 'EEG 007']
Rejecting epoch based on EEG : ['EEG 007']
Rejecting epoch based on MAG : ['MEG 1411', 'MEG 1421', 'MEG 1441']
Rejecting epoch based on MAG : ['MEG 1711']
Rejecting epoch based on EEG : ['EEG 001', 'EEG 002', 'EEG 003', 'EEG 007']
Rejecting epoch based on EEG : ['EEG 007']
Rejecting epoch based on MAG : ['MEG 1411', 'MEG 1421']
Rejecting epoch based on MAG : ['MEG 1711']
Rejecting epoch based on EOG : ['EOG 061']
Rejecting epoch based on EEG : ['EEG 008']
Rejecting epoch based on EEG : ['EEG 008']
Rejecting epoch based on EOG : ['EOG 061']
Rejecting epoch based on EEG : ['EEG 007', 'EEG 008']
Rejecting epoch based on EEG : ['EEG 001', 'EEG 002', 'EEG 003', 'EEG 007']
Rejecting epoch based on EEG : ['EEG 001', 'EEG 003', 'EEG 007']
Rejecting epoch based on MAG : ['MEG 1411', 'MEG 1421']
Rejecting epoch based on MAG : ['MEG 1411']
Rejecting epoch based on EEG : ['EEG 001', 'EEG 002', 'EEG 003', 'EEG 007']
Rejecting epoch based on EEG : ['EEG 001', 'EEG 002', 'EEG 003', 'EEG 007']
Rejecting epoch based on EEG : ['EEG 007']
Rejecting epoch based on EEG : ['EEG 003', 'EEG 007']
Rejecting epoch based on EEG : ['EEG 003', 'EEG 007']
Rejecting epoch based on EEG : ['EEG 007']
Rejecting epoch based on MAG : ['MEG 1331', 'MEG 1421']
Rejecting epoch based on EEG : ['EEG 007']
Rejecting epoch based on EOG : ['EOG 061']
Rejecting epoch based on EEG : ['EEG 001', 'EEG 002', 'EEG 003', 'EEG 007']
Rejecting epoch based on EEG : ['EEG 001', 'EEG 002', 'EEG 003', 'EEG 007']
Rejecting epoch based on EEG : ['EEG 001', 'EEG 002', 'EEG 003', 'EEG 007']
35 bad epochs dropped
将 Epochs 对象转换为 DataFrame#
有了 Epochs 对象之后，将其转换为 DataFrame 就很简单：只需调用 epochs.to_data_frame()。每个通道的数据都会成为新 DataFrame 的一列，此外还有三列分别记录事件名称、epoch 编号和采样时间。这里我们只展示前几行和前几列：
df = epochs.to_data_frame()
df.iloc[:5, :10]
缩放时间与通道数值#
默认情况下，时间值会从秒转换为毫秒，然后四舍五入到最接近的整数；如果你不希望这样，可以传入 time_format=None，以秒为单位将时间保留为 float 浮点值，也可以通过 time_format='timedelta' 将其转换为 Timedelta 值。
另请注意，默认情况下通道测量值会被缩放，使 EEG 数据转换为 µV、磁力计数据转换为 fT、梯度计数据转换为 fT/cm。这些缩放可以通过 scalings 参数自定义，或者通过传入 scalings=dict(eeg=1, mag=1, grad=1) 来取消。
df = epochs.to_data_frame(time_format=None, scalings=dict(eeg=1, mag=1, grad=1))
df.iloc[:5, :10]
注意，时间值不再是整数，而且与前一个 DataFrame 相比，通道数值变化了好几个数量级。
设置索引#
还可以通过向 index 参数传入一个字符串或字符串列表，将一个或多个指示列（事件名称、epoch 编号和采样时间）移入索引。这里我们还会演示 time_format='timedelta' 的效果，在“time”列中生成 Timedelta 值。
df = epochs.to_data_frame(index=["condition", "epoch"], time_format="timedelta")
df.iloc[:5, :10]
宽格式与长格式 DataFrame#
另一个参数 long_format 决定每个通道的数据是放在 DataFrame 的单独一列中（long_format=False），还是将测量值透视为单个 'value' 列，并额外增加一列用于指示通道名称（long_format=True）。传入 long_format=True 还会额外创建一列 ch_type，用于指示通道类型。
long_df = epochs.to_data_frame(time_format=None, index="condition", long_format=True)
long_df.head()
Converting "epoch" to "category"...
Converting "channel" to "category"...
Converting "ch_type" to "category"...
在使用其他 Python 模块进行后续分析或绘图时，以长格式生成 DataFrame 会很有帮助。例如，这里我们取出“auditory/left”条件的数据，挑选几个 MEG 通道，并使用 seaborn.lineplot() 自动绘制每个通道的均值和置信带，其中的置信度是在所选条件的各个 epoch 上计算得到的：
plt.figure()
channels = ["MEG 1332", "MEG 1342"]
data = long_df.loc["auditory/left"].query("channel in @channels")
# convert channel column (CategoryDtype → string; for a nicer-looking legend)
data["channel"] = data["channel"].astype(str)
data.reset_index(drop=True, inplace=True) # speeds things up
sns.lineplot(x="time", y="value", hue="channel", data=data)
现在，我们还可以充分利用 Pandas 的强大功能来对数据分组和变换。这里，我们找出 2 个梯度计（一个靠近听觉皮层，一个靠近视觉皮层）峰值激活的潜伏期，并用 violinplot() 绘制每个通道峰值出现时间的分布：
plt.figure()
df = epochs.to_data_frame(time_format=None)
peak_latency = (
df.filter(regex=r"condition|epoch|MEG 1332|MEG 2123")
.groupby(["condition", "epoch"])
.aggregate(lambda x: df["time"].iloc[x.idxmax()])
.reset_index()
.melt(
id_vars=["condition", "epoch"], var_name="channel", value_name="latency of peak"
)
)
ax = sns.violinplot(
x="channel",
y="latency of peak",
hue="condition",
data=peak_latency,
palette="deep",
saturation=1,
)
脚本总运行时间：（0 分钟 4.167 秒）