Background: Computational signal pre-processing is a prerequisite for developing data-driven predictive models for clinical decision support. Thus, identifying the best practices that adhere to clinical principles is critical to ensure transparency and reproducibility which will drive clinical adoption.Objective: This review focuses on the Neonatal intensive care unit (NICU) setting and summarises the state-of-the-art computational methods used for pre-processing neonatal clinical physiological signals for the development of machine learning models in predicting the risk of adverse outcomes.Methods: Five databases (Pubmed, Web of Science, Scopus, IEEE, ACM Digital Library) were searched using a combination of keywords and MeSH terms. 3,585 papers from the year 2013 to January 2023 were identified based on the defined search terms and inclusion criteria. After removing duplicates, 2,994 papers were screened by title and abstract, and 81 were selected for fulltext review. Of these, 52 were eligible for inclusion in the detailed analysis.
Results:The papers included in the review were heterogeneous in design and the selection of adverse outcomes modelled. We found a partial or complete lack of transparency in reporting the setting and the methods used for signal pre-processing. This includes reporting methods to handle missing data, segment size for considered analysis, and details regarding the modification of the state-of-the-art methods for physiological signal processing to align with the clinical principles for neonates.
Conclusions:The review found heterogeneity in techniques and inconsistent reporting of parameters and procedures used for preprocessing neonatal physiological signals, which is necessary to confirm their adherence to clinical practices, usefulness and choice of the best practices. Improving this aspect will ensure transparent reporting and hence facilitate the interpretation and reproducibility of the studies as well as accelerate their clinical adoption. Clinical