Author(s)

Xiaoxi Li, Yiyang Zhong, Zihan Xu

Corresponding Author

Xiaoxi Li

Abstract

Accurate assessment of fetal chromosome concentrations in non-invasive prenatal testing (NIPT) is crucial for prenatal screening. This paper focuses on the analysis of factors affecting sex chromosome concentration, and establishes a mathematical model based on clinical data to study the important factors affecting the accuracy of detection. The first step was to explore the relationship between fetal Y chromosome concentration and gestational age, BMI and other indicators, and the nonlinear relationship was revealed by Spearman correlation analysis and random forest model, and the influence of various factors was quantified by PDP analysis and ternary polynomial fitting. The second step was to minimize the potential risk of pregnant women, K-means clustering was carried out according to BMI indexes, a risk model was constructed, and the best detection time was determined by quantile regression and comprehensive risk function. In the third step, multifactorial was introduced, Gaussian mixed model clustering and Cox proportional hazard model were used to optimize the risk function in combination with a variety of optimization strategies, and the Monte Carlo simulation gave the optimal screening time of BMI in the four groups (22 weeks for the first two groups, 17 and 18 weeks for the last two groups). This paper constructs a complete NIPT detection data analysis and optimization system, and innovatively combines statistical analysis and machine learning methods to solve the key problems of clinical testing.

Keywords

NIPT; Spearman Correlation Analysis; K-Means Clustering; Monte Carlo Algorithm; Machine Learning