edfRead

A simple file reader for European Data Formatted (EDF-) files.

Description

Read European Data Format file into MATLAB
 
[hdr, record] = edfread(fname)
Reads data from ALL RECORDS of file fname ('*.edf'). Header information is returned in structure hdr, and the signals (waveforms) are returned in structure record, with waveforms associated with the records returned as fields titled 'data' of structure record.
 
[...] = edfread(fname, 'assignToVariables', assignToVariables)
Triggers writing of individual output variables, as defined by field 'labels', into the caller workspace.

[...] = edfread(fname, 'targetSignals', targetSignals,...)
Allows the user to specify by name or index the subset of signals to import.
 
FORMAT SPEC: Source: http://www.edfplus.info/specs/edf.html

• MATLAB release: MATLAB 7.12 (R2011a)

[Pelle Krøgholt]

 

[Cindy]

Works great, thanks Brett!

[Cindy]

 

[Brett Shoelson]

Hi all,
Please note that the new version (5/23/13) supports reading of a user-specified subset of signals.
Regards,
Brett

[Cindy]

Hi Brett,

I think I'm getting somewhere by only reading the necessary info (instead of all channel data) into tmpdata. I'm not sure if I'm getting the syntax of fread right though, specifically the SKIP argument. e.g. if I only want one channel (i.e. channel #k):

tmpdata(recnum).data{1} = fread(fid,hdr.samples(k),'int16',hdr.ns*2) * scalefac(k) + dc(k);

which doesn't quite work. I'm probably missing something...

[Brett Shoelson]

Hi Cindy,

Clearly I'm missing something...

If you can share a file with me and explicitly explain what you'd like to read (and what you'd like to skip), I will dig in and see if I can improve this reader.

Thanks!
Brett
char(cumsum([98 16 -13 15 0 -70 69 -11 7 -10 7 7 -4 -1 -46 45 -12 19 -12 15 -8 3 -7 8 -69 53 12 -2]))

[Cindy]

Thanks for this, very handy indeed.

As a previous poster has noted though, if you're only after specific channels, the code uses up a great deal of memory.

Replacing
for ii = 1:hdr.ns
with
for ii = 1:[1,3,5:7]

doesn't quite work at the point when tmpdata is being generated (where fread is used), nor does it reduce the size of the matrix being created. I run into Out of Memory problems at:

record = zeros(hdr.ns, hdr.samples(1)*hdr.records);

because tmpdata is already taking up too much memory space.

Any thoughts on overcoming these problems? I thought about deleting the irrelevant channels from tmpdata, but got a bit stuck with the syntax. Help would be appreciated!

[Brett Shoelson]

Farid, I thought that's what I did. Perhaps if my response is unclear, or if I misunderstood your question, you can email me with a better explanation of what you would like to do with EDFREAD.
Cheers,
Brett
char(cumsum([98 16 -13 15 0 -70 69 -11 7 -10 7 7 -4 -1 -46 45 -12 19 -12 15 -8 3 -7 8 -69 53 12 -2]))

[Farid]

Thanks Brett for your quick and clear response. Could you please suggest a way to read segments of data (e.g specific signals in shorter segments).
Thanks

Farid

[Brett Shoelson]

Farid,
Number of channels (signals?) is specified in line 124:
hdr.ns = str2double(fread(fid,4,'*char')');

In subsequent lines (125,128,132,...) the files reads 1:hdr.ns. If you wanted to, you could modify those calls to read (for example) signals [1,3,5,6,7] by changing instances of

for ii = 1:hdr

to

for ii = [1,3,5:7]

HTH,
Brett

[Farid]

Thanks for useful code. I wonder to know how I can use the edfread to read specific channels? I have huge file of several channels and am interested on some of them. The comments on MATLAB code is not clear to me.
Thanks

[Brett Shoelson]

@Oleksandr (and Ibraheem):
Thanks for the reviews/comments. My goal in writing this was to facilitate using MATLAB to analyze data stored in EDF format.

I encourage you to share your EDFWrite function, if you write one; creating that capability is a bit outside of my job description. ;)

I'm curious, though: what is your rationale for exporting to EDF after you've brought data in to MATLAB for analysis? Is there something you need to do with them that you can't do with MathWorks' tools? (Or did you just want to share with non-MATLAB users?)

Brett (the MATLAB guy)

[Oleksandr Makeyev]

Hi Brett,
This is a great function, thank you very much for making and sharing it!
Sorry to repeat a question that already has been asked before but is there an edfWrite (or edfSave) by any chance? I'd love to save myself some time making one if at all possible.
Once again, thank you for your work!
Regards,

[Hisham]

Thanks Brett. I discovered the problem. For some reason when I was downloading the file into my Matlab directory The file got corrupted in some way that its much different than the original page. I just copied the text Matlab function into a blank m file in my Matlab directory and so I was able to run the function without errors. Thanks

[Brett Shoelson]

Hisham, Would you either try to download the function again and re-try to read the file, or send me the file you downloaded. I'd like to understand what's going on.
Thanks,
Brett
char(cumsum([98 16 -13 15 0 -70 69 -11 7 -10 7 7 -4 -1 -46 45 -12 19 -12 15 -8 3 -7 8 -69 53 12 -2]))

[Hisham]

Thanks Brett for your quick Reply. I am using Matlab R2012a and I am not sure where is the problem. I tried the r01.edf you recommended but the same thing happended. when I run the file I got the message 'Matlab unexpected operator' on every line when I comment out everything before it. I am not doing an coding else than invoking the file by edfread('r01.edf') is this right? Also I don't think line 13 is a comment (although it seems not doing anything in the code). I have downloaded the file edfread directly. So is there anything I am missing? Thanks alot for your help.

[Brett Shoelson]

Hisham, line 13 is already a comment, isn't it? Can you tell me what platform you're using, and what version of MATLAB? And can you download a signal from Physionet and try to read it? (For instance:

http://www.physionet.org/physiobank/database/adfecgdb/r01.edf

).

If you can read that signal, it would suggest there's something in your data file that might be problematic.

Thanks,
Brett

[Hisham]

I am trying to use this function to read an edf file but I am getting the following error always
Error: File: edfread.m Line: 13 Column: 1
Unexpected MATLAB operator.
Even if I comment out Line: 13 the same error occurs on following lines. Is there anything I am missing to run the edf code.

[Ibraheem Al-Dhamari]

Thanks for sharing. It would be great if you make it two directions by adding export feature (write matlab variables to edf).
Best regards!

[Behnaz]

 

[Behnaz]

 

[Joseph Isler]

Excellent, and timing is perfect I just needed it now!

[Andrey Shapkin]

Great piece of work, thanks.

function [data,header] = edfread(filename)
% Read European Data Format file into MATLAB

fid = fopen(filename,'r','ieee-le');
% PART1
hdr = char(fread(fid,256,'uchar')');

header.ver=str2num(hdr(1:8)); % 8 ascii : version of this data format (0)
header.patientID = char(hdr(9:88)); % 80 ascii : local patient identification
header.recordID = char(hdr(89:168)); % 80 ascii : local recording identification
header.startdate=char(hdr(169:176)); % 8 ascii : startdate of recording (dd.mm.yy)
header.starttime = char(hdr(177:184)); % 8 ascii : starttime of recording (hh.mm.ss)
header.length = str2num (hdr(185:192)); % 8 ascii : number of bytes in header record
reserved = hdr(193:236); % [EDF+C ] % 44 ascii : reserved
header.records = str2num (hdr(237:244)); % 8 ascii : number of data records (-1 if unknown)
header.duration = str2num (hdr(245:252)); % 8 ascii : duration of a data record, in seconds
header.channels = str2num (hdr(253:256));% 4 ascii : number of signals (ns) in data record

%%%% PART2

header.label=cellstr(char(fread(fid,[16,header.channels],'char')')); % ns * 16 ascii : ns * label (e.g. EEG FpzCz or Body temp)
header.transducer =cellstr(char(fread(fid,[80,header.channels],'char')')); % ns * 80 ascii : ns * transducer type (e.g. AgAgCl electrode)
header.units = cellstr(char(fread(fid,[8,header.channels],'char')')); % ns * 8 ascii : ns * physical dimension (e.g. uV or degreeC)
header.physmin = str2num(char(fread(fid,[8,header.channels],'char')')); % ns * 8 ascii : ns * physical minimum (e.g. -500 or 34)
header.physmax = str2num(char(fread(fid,[8,header.channels],'char')')); % ns * 8 ascii : ns * physical maximum (e.g. 500 or 40)
header.digimin = str2num(char(fread(fid,[8,header.channels],'char')')); % ns * 8 ascii : ns * digital minimum (e.g. -2048)
header.digimax = str2num(char(fread(fid,[8,header.channels],'char')')); % ns * 8 ascii : ns * digital maximum (e.g. 2047)
header.prefilt =cellstr(char(fread(fid,[80,header.channels],'char')')); % ns * 80 ascii : ns * prefiltering (e.g. HP:0.1Hz LP:75Hz)
header.samplerate = str2num(char(fread(fid,[8,header.channels],'char')')); % ns * 8 ascii : ns * nr of samples in each data record
reserved = char(fread(fid,[32,header.channels],'char')'); % ns * 32 ascii : ns * reserved

% DATA read

Ch_data = fread(fid,'int16');

if header.records<0, % number of data records (-1 if unknown)
R=sum(header.duration*header.samplerate);
header.records=fix(length(Ch_data)./R);
end

% reshape data
Ch_data=reshape(Ch_data, [], header.records);

% scale set
sf = (header.physmax - header.physmin)./(header.digimax - header.digimin);
dc = header.physmax - sf.* header.digimax;

data=cell(1, header.channels);
Rs=cumsum([1; header.duration.*header.samplerate]); % index of block data -> Rs(k):Rs(k+1)-1

for k=1:header.channels
data{k}=reshape(Ch_data(Rs(k):Rs(k+1)-1, :), [], 1);
% scale data
data{k}=data{k}.*sf(k)+dc(k);
end

% data=cell2mat(data);