Wavelet-based algorithms for noninvasive fetal ECG post-processing: a methodological review

Objective: Downstream of noninvasive fetal electrocardiography (fECG) extraction, a post-processing signal enhancement step is generally required. To date, wavelet denoising (WD) was successfully adopted but typically parameterizations are largely empiric. This comprehensive review aims to systematize the knowledge by presenting and assessing WD algorithms for this application. Methods: 17 WD algorithms for fECG enhancement were identified, presented and quantitatively compared on the same public datasets. Overall performance, effectiveness in noise reduction and signal morphology preservation were separately assessed by using a novel methodology based on principal component analysis, to synthetize different performance metrics by a single unbiased index. Results: The comparison reveals different best solutions according to the different analyzed performance, which can be used for a reasoned selection of the right algorithm for a given enhancement problem. The proposed unbiased performance index was effective in supporting the analysis. Conclusions: WD represents a powerful tool for fECG enhancement, but the parameterizations severely affect the algorithm performance. Our findings can be used for a reasoned selection of an algorithm or for the proposal of new WD approaches. Significance: This review fills a knowledge gap by providing a comprehensive literature review for WD post-processing of noninvasive fECG. Beyond presenting the state-of-the-art, it is also a ready benchmark for the development of new solutions and a guide for researchers in the field, as well as in other biomedical signal processing applications.