Promising methods of prenatal diagnostics based on passive sensors and machine learning

Abstract

Introduction. Prenatal diagnostics of fetal vital activity includes regular assessment of parameters such as heart rate (HR), motor activity and physiological state. Current methods of fetal monitoring based on using active ultrasound waves have a number of limitations: insufficiently high diagnostic sensitivity and specificity, lack of prognostic significance and impossible long-term use. These limitations necessitate a development of innovative technologies for assessing fetal functional state.Aim: to summarize advanced global developments as an alternative to ultrasound systems for long-term fetal monitoring, allowing continuous real-time recording of fetal vital signs by using passive sensors and trend analysis with potentially high diagnostic and prognostic accuracy.Materials and Methods. The review methodology included an analysis of publications released over the past 10 years, included based on relevance criteria. Publications were selected in accordance with the PRISMA (preferred reporting items for systematic reviews and meta-analyses) guidelines. The following keywords in Russian and English were used for selection: "modern methods of prenatal diagnostics", "fetal monitoring", "assessment of fetus functional state", "passive sensors", "artificial intelligence", "machine learning". The search yielded 69 articles in the PubMed/MEDLINE database, 17,500 – in Google Scholar, 21 – in eLibrary, and 3,563 – in ResearchGate. Such articles were analyzed for relevance, relevance to the review topic, and availability of experimental data. Non-peer-reviewed publications and duplicates were also excluded from the reviewed materials. The most relevant 8 articles were included in the review, which describe promising methods of prenatal diagnostics based on the use of passive sensor experimental devices.Results. The conducted literature analysis allowed to generalize the experimental achievements of current methods of prenatal diagnostics and demonstrated great promise for automated systems to assess fetal vital signs, including monitoring fetal HR, motor activity and general functional state. However, it was found that none of the described systems achieves 100 % accuracy of the results corresponding to fetal cardiotocography and ultrasound examination data. Most experimental systems remain wired, which limits their use for fetal monitoring. Promising passive monitoring systems are based on using accelerometers, microphones and other sensors to assess fetal functional state. A key component of such technologies is the use of artificial intelligence for signal processing and interpreting, which increases the accuracy and monitoring information content. The main problem is generation of effective data processing algorithms for their accurate and unambiguous interpretation. All the technologies under consideration are still experimental, and further work is required to improve the algorithms and integrate various types of sensors to ensure comprehensive analysis.Conclusion. It is noteworthy that technologies employing passive sensors for continuous and long-term monitoring of fetal vital signs, in conjunction with machine learning algorithms for data analysis and interpretation are of particular interest. The use of wearable devices, based on passive sensors such as accelerometers and digital microphones, has the potential to enhance prenatal diagnostics, ensuring both enhanced safety and the early detection of pregnancy complications and fetal conditions.

Keywords

• prenatal diagnostics
• fetal monitoring
• prenatal screening
• fetal distress
• artificial intelligence
• machine learning