python批量处理edf文件

根据txt文件的时间段在edf中获取对应的心电信号，归纳并写入csv中

首先txt文件的部分数据是这样的：

而edf文件是这样的：

我们要根据txt里面的time和duration确定开始时间和结束时间，在edf中获取这个时间段内的信号（edf的time为时间节点，根据时间节点判断是否在各个时间段内），并将这些信号标签标为对应的type；然后在每个时间段都获取不多于10s的负样本，并将这段时间内的信号标签标为normal。最终这些都归纳写入csv中。

import csv
import os
import mne

txt_path = "/alpha/.../aa.txt"
edf_path = "/alpha/.../edf/aa.edf"
save_path = "/alpha/.../save.csv"

mul = 0.006833

def start_adjust(input):
    rem = input % mul
    if rem == 0:
        return input
    else:
        return input - rem


def end_adjust(input):
    rem = input % mul
    if rem == 0:
        return input
    else:
        return input + (mul - rem)

raw = mne.io.read_raw_edf(edf_path, preload=False)
raw2 = raw.copy()
raw3 = raw.copy()
df = raw.to_data_frame()
print(df)
# # 获取通道数值
raw.pick_channels(["chan 1"])
eeg = raw.to_data_frame()
eeg1 = list(eeg.values[:, 1])
raw2.pick_channels(["chan 2"])
eeg = raw2.to_data_frame()
eeg2 = list(eeg.values[:, 1])
raw3.pick_channels(["chan 3"])
eeg = raw3.to_data_frame()
eeg3 = list(eeg.values[:, 1])
# print(eeg1[-1], eeg2[-1], eeg3[-1])
# # 获取时间值
times = list(df["time"])
# print(list(df["time"])[-5:])

# 写表头
with open(save_path, "w", encoding="utf-8", newline='') as f2:
    writer = csv.writer(f2)
    writer.writerow(["duration", "chan1", "chan2", "chan3", "type"])
    
with open(txt_path, "r", encoding="utf-8") as f1:
    tmp = []
    for idx, line in enumerate(f1.readlines()):
        if idx in [0, 1, 2]:
            continue
        line = line.strip("\n")
        a = line.split()
        if len(a) == 1:
            continue
        # print(a)
        start = int(a[0].split(":")[0])*60*60 + int(a[0].split(":")[1])*60 + int(a[0].split(":")[-1])
        start = start_adjust(start)
        print(start)
        end = start + int(a[2])
        end = end_adjust(end)
        # print(end)

        tmp.append(end)
        neg = []
        if len(tmp) == 2:
            dura = start - tmp[0]
            if dura > 10:
                dura = 10
            neg = [tmp[0], tmp[0] + 10]
            tmp.pop(0)

        pos_eeg = []
        neg_eeg = []
        for idx, time in enumerate(times):
            if start <= time <= end:
                eggs = [eeg1[idx], eeg2[idx], eeg3[idx]]
                pos_eeg.append(eggs)
            if not len(neg) == 0:
                if neg[0] <= time <= neg[1]:
                    eggs = [eeg1[idx], eeg2[idx], eeg3[idx]]
                    neg_eeg.append(eggs)
        with open(save_path, "a", encoding="utf-8", newline='') as f3:
            writer = csv.writer(f3)
            # writer.writerow(["duration", "chan1", "chan2", "chan3", "type"])
            if not len(neg_eeg) == 0:
                for idx, e in enumerate(neg_eeg):
                    data = []
                    if idx == 0:
                        data = [str(neg[0]) + ":" + str(neg[1]), e[0], e[1], e[2], "normal"]
                    else:
                        data = ["", e[0], e[1], e[2], "normal"]
                    writer.writerow(data)
            for idx, e in enumerate(pos_eeg):
                data = []
                if idx == 0:
                    data = [str(start) + ":" + str(end), e[0], e[1], e[2], a[1]]
                else:
                    data = ["", e[0], e[1], e[2], a[1]]
                writer.writerow(data)

最终写入的csv为这样：
比如这是第一段时间内的

第二段

这里表格的duration都是以秒s为单位的，不再是txt里面十分秒的格式。

可能会出现的错误：
使用mne.io.read_raw_edf读取edf文件时，可能会报一个ValueError 的错误，这时候需要下载一个EDFbrowser软件，点Tools下的Edit EDF/BDF header选项（有的新版本也叫EDF header editor）打开每个edf文件再点保存即可，每个文件都操作一遍每个文件就都能正常被读取。