Author(s)

Haohui Wang, Xinrui Wang, Zhen Xing

Corresponding Author

Haohui Wang

Abstract

In today's era of rapid advancements in medical technology, non-invasive prenatal testing (NIPT) enables the detection of foetal abnormalities. The timing of such testing is crucial in mitigating risks associated with the narrowing treatment window. This study first investigates the relationship between Y chromosome concentration and both gestational age and maternal body mass index (BMI). Employing both multiple linear regression and polynomial regression models, it further establishes the functional relationship between Y chromosome concentration and these maternal parameters, calculating mean squared errors of 0.0978 and 0.0987 respectively for the two models. Building upon this, an optimised model was established incorporating five additional indicators: height, weight, age, detection error, and the proportion of Y chromosome concentrations meeting the standard. Subsequently, a Monte Carlo method was employed to introduce random perturbations of 0.5%–2% to the Y chromosome concentration. Results demonstrated that the three groups defined in this study achieved accuracy rates exceeding 90% under various perturbation levels. Finally, a logistic regression model calculated the regression coefficients and p-values for each feature, yielding a ranked importance order. This enabled the extraction of coefficients for the five features to determine the formula for the classification method.

Keywords

Non-Invasive Prenatal Testing; Multi-Objective Optimization; Multiple Linear Regression; Body Mass Index, Logistic Regression