Creating your own system of annotations as EDF-XML notes

In the last chapter we translated a set of well-defined annotations into EDF-XML annotations. Now we are going to face a somewhat more difficult problem. We will encode the result of a motor conduction study. Besides the information contained in the header, there are a lot of relevant variables that need to be considered. Some of them describe the stimulus (intensity, location...), some of them describe measurements (latencies, amplitudes...) and some of them describe the procedures (antidromic, peak-to-peak...). We are going to store this information as a XML string.

Basically, a motor conduction study is carried out by attaching some electrodes to the skin surface overlying the muscle that is going to be tested. Then the nerve is stimulated at supramaximal intensity. The response of the muscle in measured to obtain latency, amplitude and other parameters for each stimulation point. Different curves together with the distances between the stimulation points allow to calculate the maximum conduction velocity of the nerve. Some external variables such as height or local temperature are also relevant.

Having in mind that the next items do not constitute any kind of normative rule, we could consider the following suggestions:

• A common element for all the annotations (e.g. EDF_XML) could be used. In some cases, it could contain as attributes the name of the file, the data record, the time stamp or a label defining the whole annotation. These attributes could be utilized to locate the position of the annotation once extracted from the file.
• All the values are coded as content while their units and procedures are coded as attributes. Defining units as attributes allows the limitation of the range of possible units.
• The measurements are designed in such a way that the content of each data record is as independent as possible (e.g., distances are measured from the stimulus to the recording point, even if segmental velocity is calculated).
• When the information is relevant to all data records (e.g., local temperature), it is only coded in one of them.
• Whenever several methods are possible (e.g., peak to peak vs. baseline to peak), this fact is properly indicated.
• When measurements can be made on different signals in the same data record (e.g., motor conduction simultaneously recorded in more than one muscle, a Visual Evoked Potential recorded in several positions), the curve to which the measurements are referred to is clearly indicated. Sometimes it is useful to make measurements not attached to any curve (e.g., position of QRS in a multi-derivation record or Visual Evoked Potential P100 cosidered as a global measurement not attached to any curve).
• Since individual sweeps and averages can be included in the same file, this aspect should be well defined. When traces derive from different sweeps, the criteria to reject individual sweeps ( e.g., abs value greater than 200 uV) as well as the number of accepted/rejected sweeps, and the procedures in use (e.g., median, mean, standard deviation..) should be clearly indicated. Remember that in EDF+, individual sweeps and averages are curves handled in the same way.

We could apply some of these rules in the next annotations, which could be included as an annotation signal (data records 0 and 1) to describe a motor nerve conduction:

 
<EDF_XMLnote>
  <!--data record 0-->
  <temperature unit="degree C" position="left hand">
    33.2
  </temperature>
  <distance mode="stimulus to recording" unit="cm">
    6.5
  </distance>
  <stimulus label="stimulus" mode="constant intensity"
                 duration="0.2 ms" intensity="9.2 mA">
    wrist
  </stimulus>
  <measurements>
    <latency unit="ms">
      3.3
    </latency>
    <amplitude mode="baseline to peak" unit="mV">
      17.4
    </amplitude>
  </measurements>
</EDF_XMLnote>


<EDF_XMLnote>
  <!--data record 1-->
  <distance mode="stimulus to recording" unit="cm">
    30.5
  </distance>
  <stimulus label="stimulus" mode="constant intensity" 
                 duration="0.2 ms" intensity="27.6 mA">
    elbow
  </stimulus>
  <measurements>
    <latency unit="ms">
      7.4
    </latency>
    <amplitude mode="baseline to peak" unit="mV">
      16.8
    </amplitude>
    <velocity mode="segmental" unit="m/s">
      58.5
    </velocity>
  </measurements>
</EDF_XMLnote>

Some lines have been folded and carriage returns and tabulations have been added for clarity. There are two different annotations corresponding to data records 0 and 1. From the annotations, we can see that the first data record was obtined by stimulating the nerve at wrist (6.5 cm from the muscle) while the second data record was obtained by stimulating the nerve at elbow (30.5 cm from the muscle). The characteristics of each stimulus, as well as the measurements, are included together with the methods utilized to obtain them.

Notice that this kind of expression can be the source for the creation of a report or the input of a database with standard tools widely available.

This is only an example. You can define your annotation outline adapted to your needs. Different groups use very different procedures and require very different annotations. EDF+ gives you some very basic tools that allow the annotation of signals, acting like a frame where you can introduce your own specification.