EEG癫痫信号的切分

 通过pyEDFlib官方提供的代码，实现将医院的24小时原始信号数据逐小时切分，并保存所需的通道信号。

import os

from pyedflib import highlevel

def split_edf(input_file, output_dir):
    # 使用高级API读取EDF文件的信号和头信息
    signals, signal_headers, header = highlevel.read_edf(input_file)

    # 定义需要保留的通道
    target_channels = [
        "Fp1", "Fp2", "F3", "F4", "C3", "C4", "P3", "P4",
        "O1", "O2", "F7", "F8", "T3", "T4", "T5", "T6"
    ]

    # 根据目标通道名称获取这些通道的索引
    target_channel_indices = [i for i, label in enumerate(signal_headers) if label['label'] in target_channels]

    # 创建每小时的数据
    hour_length = 512 * 3600  # 3600秒
    for hour in range(int(len(signals[0]) / hour_length) + 1):
        start_index = hour * hour_length
        end_index = min(start_index + hour_length, len(signals[0]))

        # 获取当前小时内的信号和头信息
        hour_signals = [signal[start_index:end_index] for signal in signals]
        hour_signal_headers = [header for header in signal_headers if header['label'] in target_channels]

        # 构造新的文件名
        output_file_path = os.path.join(output_dir, f"hour_{hour}.edf")

        # 使用高级API写入EDF文件
        highlevel.write_edf(output_file_path, hour_signals, hour_signal_headers, header)

def main():
    # 指定 .edf 文件路径和输出目录
    input_file = r"F:\2024test\癫痫数据集\EEG.edf"
    output_dir = r"F:\2024test\癫痫数据集\hospitals_pit"

    # 拆分 .edf 文件
    split_edf(input_file, output_dir)

if __name__ == "__main__":
    main()

后续：由于保存时间较长 通道数较多的文件切分占用内存较大。故采用分块方式读取和保存一小时数据 

import os
import numpy as np
import pyedflib
from pyedflib import highlevel


def split_edf(input_file, output_dir):
    # 打开 .edf 文件 with使用无需关闭文件
    with pyedflib.EdfReader(input_file) as f:
        total_samples = f.getNSamples()[0]  # 总样本数 (采样频率*采样时间） 与信号数无关
        signal_labels = f.getSignalLabels()  # 信号标签
        # 获取信号头信息
        signal_headers = f.getSignalHeaders()
        # 获取全局头信息
        header = f.getHeader()
    # 定义需要保留的通道
    target_channels = [
        "Fp1", "Fp2", "F3", "F4", "C3", "C4", "P3", "P4",
        "O1", "O2", "F7", "F8", "T3", "T4", "T5", "T6"
    ]
    # 创建每小时的数据
    chunk_size = hour_length = 512 * 3600  # 3600秒

    # # 根据目标通道名称获取这些通道的索引
    target_channel_indices = [i for i, label in enumerate(signal_headers) if label['label'] in target_channels]
    # 目标信号数量
    num_target_signals = len(target_channel_indices)
    # 目标信号头文件（信号头文件）
    target_signal_headers = [signal_headers[i] for i in target_channel_indices]

    # 逐步处理每个信号的数据
    for i in range(0, total_samples, chunk_size):
        chunk_end = min(i + chunk_size, total_samples)
        chunk_signals = np.zeros((num_target_signals, chunk_end - i))
        for j in range(0, num_target_signals):
            if signal_labels[j] in target_channels:  # 判断信号通道标签是否在目标通道中
                with pyedflib.EdfReader(input_file) as f:
                    chunk_signals[j, :] = f.readSignal(j, start=i, n=chunk_end - i)

        # 构造新的文件名
        output_file_path = os.path.join(output_dir, f"hour_{int(i/chunk_size)}.edf")
        # 将分块的数据（1小时）和目标信号头文件，以及全局头文件（不变）导入到新文件中
        highlevel.write_edf(output_file_path, chunk_signals, target_signal_headers, header)


def main():
    # 指定 .edf 文件路径和输出目录
    input_file = r"F:\2024test\癫痫数据集\EEG\test.edf"
    output_dir = r"F:\2024test\癫痫数据集\切分"

    # 拆分 .edf 文件
    split_edf(input_file, output_dir)


if __name__ == "__main__":
    main()