Automatic computerized diagnosis of fetal sinusoidal heart rate

Area of Science:

• Perinatology
• Biomedical Engineering
• Signal Processing

Background:

• Fetal heart rate (FHR) monitoring is crucial for assessing fetal well-being.
• Distinguishing pathologic fetal sinusoidal heart rate (FSHR) from physiologic FSHR is clinically significant for timely intervention.
• Current methods for FSHR detection may lack objectivity and require expert interpretation.

Purpose of the Study:

• To develop and validate a computerized method for automatic detection of pathologic FSHR.
• To differentiate pathologic FSHR from physiologic FSHR and normal FHR using objective criteria.
• To apply these findings for the objective evaluation of FHR with artificial neural networks.

Main Methods:

• FHR tracings were digitized and analyzed using Fast Fourier Transform (FFT) to obtain power spectrums.
• Key spectral parameters, including peak power spectrum density (PPSD) and the ratio of spectral area (La/Ta), were calculated.
• Pathologic FSHR, physiologic FSHR, and normal FHR groups were compared based on these spectral parameters.

Main Results:

• The La/Ta ratio and PPSD were significantly higher in pathologic FSHR compared to physiologic FSHR and normal FHR.
• A diagnostic criterion combining La/Ta ratio (≥39%) and PPSD (≥300) achieved 100% true positive rate with 0% false negative and false positive rates.
• FFT frequency analysis effectively separated pathologic FSHR from other FHR patterns.

Conclusions:

• FFT-derived spectral parameters, specifically La/Ta ratio and PPSD, can reliably distinguish pathologic FSHR.
• The developed method enables automatic and objective diagnosis of pathologic FSHR.
• This automated detection system is suitable for integration into neural network-based FHR evaluation systems.