Research on Optimal NIPT Timing Decision and Intelligent Fetal Abnormality Detection Based on Machine Learning and Optimization Algorithms

Abstract

With the popularization of non-invasive prenatal testing (NIPT) technology, determining a personalized optimal testing time for pregnant women with different characteristics. To address this issue, this paper proposes a comprehensive intelligent decision-making framework that integrates prediction, optimization, and classification. First, the framework constructs a quadratic multivariate nonlinear regression model to quantify the complex relationships between key factors such as maternal BMI and gestational age and the concentration of the Y chromosome in male fetuses. On this basis, it designs an optimal timing decision model aimed at minimizing potential risks for pregnant women, which is solved using the particle swarm optimization (PSO) algorithm, thus generating NIPT gestational age recommendations for groups of pregnant women in different BMI ranges. Finally, addressing the issue of detecting chromosomal aneuploidy in female fetuses, this study establishes and trains a support vector machine (SVM) binary classification model. Experimental results show that this framework can effectively formulate personalized testing plans, with the SVM classifier demonstrating superior discriminative ability on independent test sets, achieving an area under the receiver operating characteristic curve (AUC) of 0.924, thus enabling efficient and precise identification of abnormal conditions. The findings of this study provide a scientific, data-driven theoretical basis and methodological support for NIPT clinical decision-making, showcasing the application value of integrating “prediction” with “decision-making.”