A Study on Optimal NIPT Timing and Fetal Anomaly Detection Based on Multi-Model Approach

Abstract

The accuracy of Non-Invasive Prenatal Testing (NIPT) critically depends on selecting the optimal testing time point, which is influenced by maternal factors such as gestational week and Body Mass Index (BMI). This paper develops a comprehensive modeling framework to address two key challenges: optimizing NIPT timing for male fetuses and accurately detecting chromosomal abnormalities in female fetuses. A quadratic polynomial regression model is established to quantify the relationship between Y-chromosome concentration, gestational week, and BMI. To determine the optimal NIPT time point, K-Means clustering is integrated with a logistic regression-based risk prediction model, which minimizes a composite risk function. For the classification of fetal anomalies in females, a fusion model combining Random Forest and Logistic Regression is proposed. The results demonstrate that Y-chromosome concentration exhibits a weak but statistically significant positive correlation with gestational week and a negative correlation with BMI. The risk optimization model identifies distinct optimal NIPT timings for different maternal BMI and age groups. The fusion classification model achieves an accuracy of 0.9505, a precision of 0.8235, a recall of 0.7000, and an F1-score of 0.7570, with an optimal fusion weight of 0.67 and a decision threshold of 0.4459. This study provides a robust theoretical framework and actionable recommendations for personalized NIPT scheduling and diagnosis in clinical practice.