Investigation of the quality of umbilical artery Doppler waveforms
Hoskins, Peter R.
MetadataShow full item record
In Doppler systems which automatically calculate the maximum frequency envelope and pulsatility index (PI) of umbilical artery Doppler waveforms there is the possibility of error in these parameters when the technical quality of the acquired waveform is low. Low quality waveforms may arise when there is an inappropriate set of physical parameters or when there are other sources of noise such as overlying vessels signals. In this thesis the effect of physical parameters on the envelope and on PI are investigated, and also methods for the detection of low quality waveforms are described and tested. A flow phantom which is able to produce realistic looking umbilical artery Doppler waveforms is described. This is based upon microcompruter control of a stepping motor / gear pump combination. The statistics of the Doppler spectra produced using artificial blood and human blood in the phantom are found to be identical. The effect of a number of physical parameters on the simulated umbilical artery waveforms produced using the phantom is investigated. The accuracy of estimation of the envelope and the PI is similar over a wide range of physical conditions. A suitable image processing algorithm for speckle reduction of Doppler waveforms is developed and tested using simulated waveforms from the phantom. Using the flow device it was found that both filtering of the envelope and also speckle suppression of the spectrum improved the accuracy of estimation of the envelope and of the PI. A number of quality indices based upon the degree of noise of the envelope are described. Using the flow device there is found to be a high correlation between the quality index values, and the errors in PI and errors in envelope estimation respectively. In a clinical trial the quality index values from umbilical arteries were compared with the waveform quality as assessed by a skilled observer. The clinical results show that quality indices are able to separate high and low quality waveforms when the indices are calculated from the unprocessed envelope, but not when calculated from the filtered envelope