Research on the Timing Selection of NIPT and the Judgment of Fetal Abnormalities Based on Historical Data

Xiangxiang Meng; Xingmin Zhou; Weiqu Cui

doi:10.54097/zwkw6q87

Authors

Xiangxiang Meng
Xingmin Zhou
Weiqu Cui

DOI:

https://doi.org/10.54097/zwkw6q87

Keywords:

NIPT detection; Random Forest; SMOTE algorithm; Y chromosome concentration.

Abstract

With the wide application of non-invasive prenatal testing (NIPT) technology, scientifically determining testing time points and effectively identifying fetal abnormalities have become hot topics. The integration of machine learning and mathematical modeling offers crucial support for related research. Based on actual NIPT data of pregnant women in a region, this paper focuses on two core tasks: Y chromosome concentration prediction and abnormal female fetal identification, constructing and optimizing multiple machine learning models. First, raw data was cleaned, missing values imputed and normality tested. Spearman correlation analysis explored relationships between Y chromosome concentration and relevant factors. A prediction model integrating random forest and gradient boosting regression was then built. After hyperparameter and ensemble optimization, its R2 rose from 0.5993 to 0.7268 with good significance, supporting quantitative NIPT timing selection. To address the scarcity of abnormal female fetal samples, the SMOTE algorithm balanced the database, and a random forest-based identification model was established. Results show the models significantly enhance detection reliability and sensitivity, providing a scientific basis for individualized clinical non-invasive prenatal screening decisions.

References

[1]Blanc G, Lange J, Strassle C, et al. A Distributional-Lifting Theorem for PAC Learning[J]. arXiv preprint arXiv,2025,2506.16651.

[2]Sharma A, Lysenko A, Jia S, et al. Advances in AI and machine learning for predictive medicine[J]. Journal of Human Genetics, 2024, 69(10): 487-497.

[3]Wang G, Zhang Y, Wu T, et al. Serum Zn and Ca correlate with false positive non-invasive prenatal test results[J]. American Journal of Translational Research, 2025, 17(3): 2135.

[4]Wang C, Zhang H, Liu Y, et al. Results and indications changes by prenatal diagnosis of 3,458 pregnant women: a 10-year retrospective study[J]. Annals of Medicine, 2025, 57(1): 2563753.

[5]Guillet S, Loustau V, Boutin E, et al. Immune thrombocytopenia and pregnancy: an exposed/nonexposed cohort study[J]. Blood, 2023, 141(1): 11-21.

[6]Ashoor G., Syngelaki A., Poon L. C., et al. Fetal fraction in maternal plasma cell-free DNA at 11–13 weeks’gestation: relation to maternal and fetal characteristics. [J].Journal of ultrasound Obstet Gynecol, 2013, 41(1): 26–32.

[7]Pergament E., Cuckle H., Zimmermann B., et al. Single-nucleotide polymorphism-based noninvasive prenatal screening in a high-risk and low-risk cohort. [J].Journal of Obstetrics and Gynecology, 2014, 124: 210–218.

[8]Nguyen T H, Nguyen H V, Pham M D, et al. An efficient computational method to create positive NIPT samples with autosomal trisomy[C]2024 12th International Conference on Bioinformatics and Computational Biology (ICBCB). IEEE, 2024: 126-131.

[9]Zhang H., Gao Y., Jiang F., et al. Noninvasive prenatal testing for trisomy 21, 18 and 13—Clinical experience from 146,958 pregnancies. [J].Journal of ultrasound Obstet Gynecol, 2015, 45(5): 530–538.

[10]Gao F., Lin J., Liu X., et al. Application of deep learning in prenatal screening and genetic diagnosis. [J].Journal of Frontiers in Genetics, 2021, 12: 667.

[11]He W., Wang X., Li J., et al. SMOTE-based data augmentation for imbalanced NIPT datasets. [J].Journal of IEEE Access, 2022, 10: 45062–45074.

[12]Zhao Y., Chen H., Zhang X., et al. Feature importance and interpretability in machine learning models for prenatal testing. [J]. Journal of Computers in Biology and Medicine, 2023, 156: 106760.