edfrd

edfrd is a Python 3 software library to read and write EDF files.

It is designed as a low-level library, that does not interpret the EDF data whenever possible. Therefore, edfrd can read files even if non-standard values are contained in the header.

Data records are loaded as int16 arrays using numpy.

Installation

pip3 install --user edfrd

Reading EDF Header and Data Records

from edfrd import read_header, read_data_records

file_path = 'PATH/TO/FILE.edf'

header = read_header(file_path)
print(header)

for data_record in read_data_records(file_path, header):  # generator
    # iterate through data_records
    break

for signal in data_record:
    # iterate through signal arrays of a single data_record
    print(signal.size)

for signal_header, signal in zip(header.signals, data_record):
    # iterate through signal headers and signal arrays
    print(signal_header.label, signal.size)

If the header of your EDF file does not correctly specifiy the number of data records, you can use the following option to calculate it from the file size.

header = read_header(file_path, calculate_number_of_data_records=True)

You can try parsing the startdate_of_recording and starttime_of_recording as integer tuples. If parsing fails the original string will be returned.

header = read_header(file_path, parse_date_time=True)

day, month, year = header.startdate_of_recording
hours, minutes, seconds = header.starttime_of_recording

The number of data records being read can be limited by specifying an optional start or end index.

for data_record in read_data_records(file_path, header, start=0, end=header.number_of_data_records):
    break

To work with larger chunks of a signal than provided by a data record, consider creating a new numpy array as a buffer.

import numpy as np
from edfrd import read_header, read_data_records

file_path = 'PATH/TO/FILE.edf'

header = read_header(file_path)
start, end = 2, 4
signal_index = 0
signal_header = header.signals[signal_index]
buffer_length = (end - start) * signal_header.nr_of_samples_in_each_data_record
buffer = np.empty(buffer_length, dtype=np.int16)
pointer = 0

for data_record in read_data_records(file_path, header, start=start, end=end):
    buffer[pointer:pointer+signal_header.nr_of_samples_in_each_data_record] = data_record[signal_index]
    pointer += signal_header.nr_of_samples_in_each_data_record

print(buffer)

You can also pass a file descriptor (fr) instead of a string (file_path). Note that read_data_records will continue from the current byte position, where read_header stopped, without performing an additional seek operation.

with open(file_path, 'rb') as fr:
    header = read_header(fr)

    for data_record in read_data_records(fr, header):
        break

Writing EDF Header and Data Records

from edfrd import read_header, read_data_records, write_header, write_data_records

file_path = 'PATH/TO/FILE.edf'
new_file_path = 'PATH/TO/NEW_FILE.edf'

header = read_header(file_path)
data_records = read_data_records(file_path, header)
write_header(file_path, header)
write_data_records(file_path, data_records)

Again, using file descriptors (fr and fw) is possible.

with open(file_path, 'rb') as fr:
    header = read_header(fr)
    data_records = read_data_records(fr, header)
    
    with open(new_file_path, 'wb') as fw:
        write_header(fw, header)
        write_data_records(fw, data_records)