DATA ACQUISITION STUDY FOR DEVELOPING AI (Artificial Intelligence) ALGORITHMS TO PREDICT PRETERM LABOR AND ADVERSE PERINATAL OUTCOME

This study aims to acquire data from around 1500 unselected pregnant women attending antenetal OPD in a tertiary care referral center in South India (single center). Data acquired includes risk factors from history, examination, blood/urine reports, Ultrasound reports - all relevant to prediction of preterm labour. In addition, regularly performed Non-stress tests and uterine activity monitoring are digitally captured during their antenatal visits at or beyond 26 weeks of pregnancy. Once data is collected, study aims to develop and test Artificial Intelligence Algorithms to predict preterm labour in obstetric population.

Preterm labour/delivery (PTL/PTD) is a major health burden and contributes significantly to perinatal mortality/morbidity as well as long term health problems in the children. With a worldwide prevalence of 5-18%, it is an urgent need to predict and prevent preterm delivery. Multiple individual risk factors and risk scoring algorithms have been developed to predict PTL/PTD but none have reached a level of accuracy to be able to be employ in general obstetric practice. In addition, over half of PTD occur in women with no apparent risk factors. Complex interplay of risk factors is also a well-known phenomenon – which is difficult to be interpreted during routine obstetric practice. Hence this study aims to see if incorporating all these multiple risk factors in addition to periodic observation of fetal heart rate/uterine activity patterns – into an AI tool – to see if we can develop a prediction model for preterm delivery that has a better prediction value than the multiple existing risk factors/risk scoring systems. A large data set on fetal heart rate pattern along with uterine activity has not so far been studied by an AI algorithm in an attempt to predict preterm delivery. Also, incorporating multiple clinical, laboratory, imaging parameters to fetal heart+uterine activity patterns may help us develop a more robust tool to predict PTL/PTD - this information is not available in clinical literature. An accurate prediction may help employ preventive interventions in a suitable subset of vulnerable pregnant women in order to prevent PTL/PTD.