The changes introduced in EDF+: handling discontinuous recordings

Imagine that we need to store the sweeps used to get a Brain Auditory Evoked Potential by using two channels. We sample the EEG at 20 kHz and want a window of 10 ms. We give the stimulus every 100 ms (equivalent to an stimulus rate of 10 Hz). If we see the internal structure, we note that we are going to use only about 10 percent of the file. The solution of EDF+ is very simple: eliminating the data records that are not going to be used. It introduces a division between continuous and discontinuous recordings 4.1

Figure 4.1: In EDF, all the data records are continuous. In EDF+ data records can be discontinuous. The discontinuous record in the lower part of the figure includes only the data records labeled as 1,2,3

EDF+ is by no means limited to the periodic extraction of data records. We can extract any slice of the continuous record to form a discontinuous record. The only limitation is that all the data records in the file must have the same duration (and of course the same characteristics of recording). This is not a very important limitation because inside a discontinuous file there can be prolonged segments of continuous data records with no limitation of duration.

Obviously, by using this method we save a lot of storing space. We could see a continuous EMG recording as analogous to free running and discontinuous recording as single sweep or triggered sweep depending on the use of an external stimulus or internal triggers.

Notice that by eliminating the continuity we need some way of describing the time at which the data record begins. This information would be lost if we do not use some convention. We will describe the method proposed to face this problem in the next chapter.

It is important to stress that the stimulus is not necessarily coincident with the start of the window; sometimes we use a delay line (e.g., to see the start of a motor unit potential), sometimes we need a baseline previous to the stimulus and even sometimes we can be interested in beginning the acquisition some time after the stimulus. All these variants are treated with the same mechanism: the annotations described in the next chapter.

EDF+ is not limited to discontinuous recordings. A continuous recording can also be coded as EDF+ and the mechanisms created to annotate records can be used.

This has been a very easy chapter. A data record was made equivalent to the window of our instruments. This is the core of EDF+ and has no equivalence in EDF. It has some consequences:

• When an EDF viewer handles a EDF+ file, it understands the file as if it were continuous.
• The algorithm has also to take into account this feature. Imagine a smoothing algorithm. It must stop at the end of each data record and begin a new smoothing in the next data record. Unless this fact is considered, some erroneous results could be produced.

To minimize these and other similar problems, EDF+ includes in the reserved field of the global header the string EDF+C for continuous recordings and EDF+D for discontinuous ones.