Usage

Once you have installed edfAnn you start the program in the usual way (calling the program from the commnand line or clicking the icon). We are going to use the file wav.edf. This file is also avaliable at the download page. You can use the program to face the following processes:

• Reading the annotations of an EDF+ file

  To do it, from inside edfAnn (some tooltips appear if the mouse is maintained over the fields)...

  • Choose the first tab (EDFann > File)
  • Select the EDF+ file (In this case wav.edf)
  • At this time the first annotation signal appears in the combo box and a default output file (name.txt ) appears
  • Modify the default options if needed
  • Push Export Annotations

  This is the output as seen with KWrite ( 3.1). Notice that the units of the EDF+ file are expressed in seconds related to the start time contained in the header of the file.

  Figure 3.1: This is the output of the annotations such as it is seen in KWrite

  

  Notice that the Unicode character appears properly shown (of course to see properly Unicode characters this feature has to be supported by the program and by the Operating System). Also notice that the file contains six point annotations (they have no duration). You can see the code used to create the annotations in the tutorial Coding a sin wave in EDF+ by using Java

  In the next point we are going to use an editor to add/erase/modify the annotations of a file.

• Writing annotations to an EDF+ file

  To do it

  • Prepare the text file. In this case we will modify wav.edf.txt
  • Choose the second tab (File > EDFann)
  • Select the EDF file (In this case wav.edf
  • Select the file containing the annotations (In this case wav.edf.txt)
  • Select the output EDF+ file (In this case wavNew.edf)
  • Click Import Annotations

  Now let's do it in practice. We modify the text file. This is the file as seen with KWrite ( 3.2)

  Figure 3.2: The whole set of annotations that will be included in the file

  

  Once finished the process we can see the new file with RASCHlab ( 3.3)

  Figure 3.3: The file such as is seen with RASCHlab

  

  Notice that the old annotations and the new ones are mixed. We can also see in the figure the full support of EDF+ annotations by RASCHlab (point annotations, annotations with duration -even annotations beginning out of the record- and annotations containing Unicode characters are shown). Also notice that the text file can be modified in many ways. The structure of the file is easy to remember

  onset<TAB>duration<TAB>content<CR>
  onset<TAB>duration<TAB>content<CR>
  onset<TAB>duration<TAB>content<CR>
  ...
  

  and so, the output of any processing package (e.g. Scilab, R, Octave..) can be easily modified to be included in the EDF file.

  You can add annotations to the annotations included in the EDF+ file (such as we did in the previous example) but you can also create an annotation signal from scratch, e.g. to convert an EDF file in an EDF+ file (a plain EDF file does not contain annotations).

• Reading and writing onset labels to an EDF+ file

  In EDF+ files the onset of each data record is coded as the first empty annotation (see the EDF+ documantation). Since this specific annotations have a very different function they are handled by edfAnn in a different (although similar) way. The procedure is exactly the same than the procedure used to read/write annotations. If we export the onsets of wav.edf to the file wav.onsets.txt, this is the result

   
  je@unit1:~/prog/tmp> cat wav.onsets.txt
  0
  1
  2
  3
  4
  

  Each line is the first time empty stamp, marking the onset of the data record. The format is

   
  onset_rec0<CR>
  onset_rec1<CR>
  onset_rec2<CR>
  ...
  

  The number of lines must be coincident with the number of data records (you can find this value by editing the header of the EDF file with edfEdit). You can use this feature to convert continuous EDF+ files into discontinuous EDF+ files. And the inverse process can also be carried out. Be aware that you can also destroy the recording (the file that you will create) by changing the relations between the different data records.

  Unless you want to modify the time relations of the signals, this is not a very important issue.

As a summary think for a moment in the process used to import/export annotations. The magic comes when you modify the text file. Some examples:

• You have an EDF file and a hypnogram in other file in the form

   
  0<TAB>30<TAB>Sleep stage 0<CR>
  30<TAB>30<TAB>Sleep stage 0<CR>...
  

  You can include it as a signal of an EDF+ file.

• You can create a frame to include the analysis of the sleep in this way

   
  0<TAB>30<TAB>Sleep stage ?<CR>
  30<TAB>30<TAB>Sleep stage ?<CR>...
  

• Then you mark with some external program or algorithm the apneas eventually detected and you could have a file in the form

   
  1450<TAB>32.3<TAB>Obstructive apnea...
  

  You append the file containing the sleep stages with the apnea file and you include the new signal that now contains stages and apneas.

• Then you save the file and some months after you want to create a text file containing only the obstructive apneas. Now you do not have access to the original file (you only have the EDF+ file) but the relations between annotations and signals are deeply built in the structure of the EDF+ file. You could export the annotations and filter the lines containing the word obstructive. EDF+ is a nice way to store all the relevant information of the recording.

• You can have some algorithm e.g. to detect apneas... and you want to check its correctness with some viewer. You do not need to program the EDF file you only need to create the file with the annotations

• Once analyzed the file, segments indicating artifacts or segments requiring further analysis could be marked for further processing

And possibly you will be able to imagine some similar ways to process sleep recordings or many other physiological signals... you will share the same procedures for the analysis of en EMG, the averaging of an evoked potential or the marking of a conventional EEG recording.

One last comment on the integration of annotations coming from different origins. Different formats of annotations can be converted by hand or by the appropriate programs. In this way annotations coming from different sources can be combined and included in the EDF+ file.

Let's face an example on how to include Physionet annotations. Of course you have to install Physionet WFDB . We begin by reading the sampling of the file and the annotations

je@unit1:~/prog/tutorials/edfAnn/Code> sampfreq 100s
360
je@unit1:~/prog/tutorials/edfAnn/Code> rdann -r 100s -a atr | head -n 6
    0:00.050       18     +    0    0    0      (N
    0:00.213       77     N    0    0    0
    0:01.027      370     N    0    0    0
    0:01.838      662     N    0    0    0
    0:02.627      946     N    0    0    0
    0:03.419     1231     N    0    0    0

Now we want to convert this annotation to our annotation format.

onset<TAB>duration<TAB>content<CR>
onset<TAB>duration<TAB>content<CR>
onset<TAB>duration<TAB>content<CR>
...

Notice that the original annotations indicate the position of QRS complex and have no duration. We have to read the second column, convert position in samples to position in seconds (in this case sampling rate is 360) and add a duration of 0 to all the annotations. To do it we can use awk (the first line has been folded)

 
je@unit1:~/prog/tutorials/edfAnn/Code> rdann -r 100s -a atr |
           awk '{print $2/360 "\t 0 \t" $3}'
0.05     0      +
0.21388  0      N
1.02778  0      N
1.83889  0      N
2.62778  0      N
3.41944  0      N
4.20833  0      N
5.025    0      N
5.67778  0      A
6.67222  0      N
...

Now we can save the file and include the annotations in the EDF+ file. A similar process could be carried out with other annotations.