Study on Detection Time Selection and Fetal Abnormality Determination of NIPT Based on Multi-Model Fusion

DOI: https://doi.org/10.62517/jmhs.202505410

Author(s)

Yiqing Zhang, Siheng Yang, Zihe Yang, Zhenting Chen*

Affiliation(s)

School of Artificial Intelligence, Guangzhou Huashang College, Guangzhou, Guangdong, China *Corresponding Author.

Abstract

The accuracy of Non-invasive prenatal testing (NIPT) is highly dependent on optimal timing, especially for pregnant women with high Body Mass Index (BMI). Using regional NIPT data from high-BMI subjects, this study developed a multi-model fusion framework to optimize testing timing and improve detection of fetal chromosomal abnormalities. The framework integrates: weighted mixed-effects linear regression (for Y-chromosome correlation), quantile optimization (for BMI-grouped timing), Gradient Boosted Regression Tree-Bayesian (GBRT) risk integration (for multi-factor timing), and statistical threshold-logistic regression fusion (for female fetal diagnosis). Results indicate that Gestational Age (GA) positively and BMI negatively affect Y-chromosome concentration. Optimal timing varied by BMI group (13.89–20 weeks). The GBRT model achieved a prediction Mean Absolute Error (MAE) of 0.98 weeks, with BMI identified as the most influential factor. The female fetal model attained an Area Under the Curve (AUC) of 0.867 and demonstrated 94.03% recall for medium- to high-risk cases. This framework offers robust, data-driven guidance for personalized NIPT timing in high-BMI pregnancies.

Keywords

Non-Invasive Prenatal Testing (NIPT); Mixed-Effects Model; Quantile Optimization; GBRT; Risk Minimization; Anomaly Detection

References

[1] Jiang Liya, Lu Shaokan, Du Jiaen, et al. Development and Application of Non-Invasive Prenatal Testing Technology. Clinical Medical Research and Practice, 2025, 10 (23): 191–194. [2] Lo, Yuk-Ming Dennis, Corbetta, Nicola, Chamberlain, Peter F., Rai, Vandana, Sargent, Ian L., Redman, Christopher W., & Wainscoat, James S. (1997). Presence of fetal DNA in maternal plasma and serum. The Lancet, 350 (9076), 485–487. [3] Li Yanfang, Zhang Lanzhen, Tian Geng, et al. Clinical study on fetal free DNA for non-invasive prenatal detection of trisomy 21, 18, and 13 syndromes. China Journal of Obstetrics and Gynecology, 2015, 16 (02): 126–129. [4] Pan Xiaoli, Pan Yun, Li Haibo. Factors Influencing Low Concentration of Fetal Free DNA and Its Relationship with Pregnancy Outcomes. China Journal of Prenatal Diagnosis (Electronic Edition), 2023, 15 (04): 18–21. [5] Pinheiro, Jose, Bates, Douglas, DebRoy, Saikat, Sarkar, Deepayan, & R Core Team. (2023). nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.1-162. [6] Xing Lingling, Liu Hongqian. Factors Influencing Fetal Free DNA Concentration and Discussion on Related Prenatal Screening/Diagnosis Strategies. China Journal of Prenatal Diagnosis (Electronic Edition), 2023, 15 (03): 11–19. [7] Friedman, Jerome Harold. (2001). Greedy function approximation: A gradient boosting machine. Annals of Statistics, 29 (5), 1189–1232. [8] American College of Obstetricians and Gynecologists (ACOG). (2020). Committee Opinion No. 828: Indications for outpatient antenatal fetal surveillance. Obstetrics & Gynecology, 135 (6), e237–e250. [9] Guo Zhiyuan, Hou Dongxia, Wang Jie, et al. Clinical Application and Research Progress of Non-Invasive Prenatal Genetic Testing Technology. Inner Mongolia Medical Journal, 2021, 53 (02): 180–183. [10] Hui, Li, & Bianchi, Diana W. (2021). Recent advances in the prenatal interrogation of the human fetal genome. Trends in Genetics, 37 (2), 103–114.