Method Ayres et al. (2000) (Sisporto 2.0)
<bibtex> Ayres-de-Campos, Diogo, Bernades, João, Garrido, Antonio, Marques-de-Sa, Joaquim, Pereira-Leite, Luis - SisPorto 2.0 : A program for automated analysis of cardiotocograms The journal of Maternal-Fetal Medicine pp. 311--318,2000
This method has been developed at the Porto University (Portugal). Following to their works on Fetal Heart Rate, the society Omniview has been created and they distribute the software SisPorto which is one of the most known software for the automated FHR analysis. Their team is highly active on research and they have published numerous papers concerning their analysis. This version 2.0 seems very different from the current version.
The entire method to get the baseline is described on the following flow chart.
The first idea described on the algorithm is to take the mode of the signal other a window of 10 min. The mode is not always the best value for the baseline and it seems that statistically, they can prefer taking another frequent value (a FHR value with more than 5% of the histogram). Globally they try to take a frequent value which is closer of 110bpm but over 110bpm. Depending of the percent abnormal Short Term Variability Point (aSTV=% of successive point with a difference of less than 1bpm), the method will FAVORISE more or less closer value of 110bpm.
The method seems well detailed and we assume that our code correspond to what is written on the paper. However there is numerous unjustified criterions on the method and there may be ambiguity or error on method description. As example aSTV could mean percent of abnormal SVT on some time and means average STV in other times which would be a little bit more justified but following the description we need to assume that aSTV is always the percent of abnormal SVT.
We did not test but we assume there is several difference with the current version of Sisporto.
<syntaxhighlight lang="matlab"> </syntaxhighlight>
Reported results and discussion
The idea of taking the mode is probably a simple and efficient idea. However the mode only works if there is no baseline change (see ????). There is simple improvement which could be done by taking a sliding window with a reduced step. The idea of taking another frequent value depending of values and their frequencies and aSTV can probably improve a bit the result but the criterions might be better chosen.
The criterions for taking those other frequent value are not justified on the paper and they are very strange. We do not know if it is error on the description of the method or if it is voluntary for an obscure reason.
As example, if the most frequent value f1 (with frequency h1) is between 110bpm and 152 bpm. The method will take lower value over 110bpm as long as this lower value have a frequency over 0.05 and over a factor K(aSTV) of h1. The figure shows the curve of K in function of aSTV. We do not understand the discontinuities of this curve and why this curve is lower for some values.