A Multiple Regression-Based Method for NIPT Timing Selection and Fetal Abnormality Detection

Abstract

Non-invasive prenatal testing (NIPT) is a widely applied clinical technique for detecting fetal chromosomal abnormalities by analyzing cell-free fetal DNA fragments in maternal plasma. The accuracy and reliability of NIPT are affected by multiple factors, including gestational age and maternal body mass index (BMI). This paper investigates the correlation between Y chromosome concentration, gestational weeks, and BMI in male fetuses, as well as the feasibility of predicting female fetal abnormalities in the absence of Y chromosome signals. To address these issues, this article establish a comprehensive framework combining interaction regression analysis, gradient boosting decision trees (GBDT), rule-based grouping, random forests, and logistic regression models. Based on historical data, optimal NIPT time points are estimated for different BMI groups by calculating the gestational weeks at which 95% of pregnant^[1] women achieve the required fetal DNA concentration, with additional safety margins introduced to enhance robustness. Experimental results show that the optimal testing window for male fetuses is approximately 11 ± 3 weeks, while the proposed model achieves a precision of 0.8178 and a recall of 0.8621 for female fetal abnormality detection. The findings provide practical clinical recommendations to minimize potential testing risks, improve detection success rates across maternal BMI groups, and offer effective judgment methods for female fetuses lacking Y chromosome markers.