This project is a part of a research dealing with analysis of fetal Doppler echocardiogram in order to diagnose fetal stress.
Abstract
This project is a part of a research dealing with analysis of fetal Doppler echocardiogram in order to diagnose fetal stress. This is done by analyzing a long fetal heart rate echocardiogram (about 20 min) signal in frequency domain. Our goal is to get the heart rate signal from the Doppler echocardiogram.
In this work we present an algorithm based on an auto-correlation procedure and test its performance on both test signals and real fetal Doppler echocardiogram. The results are tested visually and ,when possible, by statistical measuers.
Note that the Doppler signal we worked with was first rectified and its envelop detected.
Basic Approach
Basically, the Doppler signal is composed of a heart beat pattern which is repeated every heart beat – our goal is to measure the interval between each heart beat. The Doppler signal is very sensitive to the fetus position, the ultrasound probe distance from the fetus heart, etc. And therefore, the pattern is different for each patient and even changes during monitoring. Moreover, the signal contains short sections (about 1-3 sec) in which the signal cannot be used due to external interrupts (fetus movements, probe displacements).
A block scheme of the algorithm is present in the following sketch:
Figure 1. Algorithm block scheme
The parameters used in the algorithm are:
- Number of Beats (NB)
- Window Size (W)
Each NB beats one reference pattern is chosen and used to correlate the next NB beats. The cross-correlation between the reference pattern and the signal has local maxima in places corresponding with the following beats. The distance between two consecutive maxima is the distance between heart beats, the heart beat interval.
Tools
We used the analog output of HP 8040A Cardiotocograph and A/D PC card in order to sample the Doppler echocardiogram.
The sampled signals were of mature fetuses near delivery.
Note that the signals were used to develop an approach algorithm. The fetal stress itself was not analyzed.
The algorithm was developed in MATLAB.
Results
With optimal parameters values (obtained from running the Algorithm on test signals resembling the real Doppler signals), the algorithm was shown to produce a good heart-beat series from Doppler echocardiograms.
In the following figure is is shown that the Algorithm finds the heart beats in the correct places, even when the signal is saturated and its amplitude varies.
Figure 2. Fetal heart beat signal and the algorithm results
Conclusions
As a fetal heart beat detector an auto-correlation algorithm has been proposed. The algorithm was found to be efficient and the results were satisfactory. The algorithm detects the heart beats without dependence on the heart beat pattern. Moreover, it overcomes the noise that usually characterize the fetal heart beat signal, even when the signal gets garbled due to external causes.
Acknowledgments
We would like to thank Dr. Israel Teler for his assistance in sampling the signals, and for explaining the medical aspects of the project. We would also like to thank Dan Adam and Johanan Erez.
This work was supported by the Ollendorff Center Research Fund.