Adaptive Spectral Estimation Improves Preeclampsia Prediction when using Short Pulsed Wave Doppler Acquisition Duration

The logarithmic ratio of maternal to fetal spectral peak (LRSP) has been proposed as a potential biomarker for predicting preeclampsia (PEC) from umbilical vein Doppler signals. However, short clinical Doppler acquisitions limit spectral resolution and cause significant spectral leakage, degrading LRSP estimates from discrete Fourier transform (DFT)-based methods. Adaptive spectral estimation techniques, such as Capon and Amplitude and Phase Estimation (APES), theoretically offer improved resolution and robustness compared to DFT. This study compared adaptive spectral estimators with DFT using clinical Doppler data. Capon demonstrated the highest predictive performance (AUC = 0.75), higher than APES (AUC = 0.68) and DFT (AUC = 0.65), while also showing the smallest variance. These results suggest the potential utility of adaptive spectral estimation methods, particularly Capon, to improve the accuracy and robustness of LRSP-based preeclampsia prediction from short-duration Doppler signals in clinical practice.