Research on a Personalised NIPT Screening Strategy Based on Multi-factor Collaborative Optimisation and a Multi-level Bayesian Network

DOI: https://doi.org/10.62517/jmhs.202605304

Author(s)

Huan Li1, Zhiguo Hu1, Chenyang Hu1, Xiang Li1, Mingming Gong2

Affiliation(s)

1School of Artificial Intelligence and Big Data, Henan University of Technology, Zhengzhou, Henan, China 2iFLYTEK Co., Ltd., Hefei, Anhui, China

Abstract

This paper proposes an integrated optimisation framework based on hierarchical decision trees and multi-layer Bayesian diagnosis. Firstly, a B-spline regression model is employed to analyse the non-linear interaction between foetal cell-free DNA concentration, gestational age and BMI, and robust BMI stratification is achieved via a hierarchical Bayesian model. Secondly, an extended hierarchical decision tree model (HDT-Extension) is established to dynamically adjust the testing thresholds for different subgroups by incorporating factors such as maternal age and body constitution. Finally, with a focus on abnormal conditions in female foetuses, a dual-model collaborative algorithm combining Multi-Hierarchical Bayesian Diagnosis (MHBD) and Feature Engineering Integrated Decision Tree (FEIDT) is developed for comprehensive evaluation. Experiments reveal that the optimal testing timing differs significantly across various BMI subgroups (P<0.0001), with the optimal timing set at 14.3 gestational weeks for the low-BMI cohort and 17.8 gestational weeks for the obese cohort; following strategic optimisation, the overall qualification rate rises from 86.6% to 93.8%. Furthermore, the complementary effect of the two models strikes a balance between high recall and high specificity. The proposed framework effectively reduces the risk of missed diagnoses in high-BMI populations and provides intelligent decision-making support for personalised prenatal screening.

Keywords

Non-Invasive Prenatal Testing (NIPT); Hierarchical Decision Tree; Bayesian Hierarchical Model; Dynamic Threshold; Ensemble Learning; Medical Decision Support System.

References

[1] Jiang L Y, Lu S K, Du J, et al. Development and Application of Non-invasive Prenatal Testing Technology. Clinical Medicine Research and Practice, 2025, 10(23): 191–194. DOI: 10.19347/j.cnki.2096-1413.202523047. [2] Zhang M, Li Q, Xu L, et al. The clinical utility of NIPT in the screening for foetal chromosomal aneuploidy. International Journal of Laboratory Medicine, 2025, 46(23): 2896–2901. [3] Geng Z X, He F J, Xu J J. Analysis of the Screening Value of CMA Combined with NIPT for Foetal Chromosomal Abnormalities in the Second Trimester. Chinese Journal of Family Planning, 2025, 33(02): 473–478. [4] Ye Y, Liu C L. An Analysis of the Clinical Value of Non-invasive DNA Prenatal Testing in Pregnant Women at High Risk of Trisomy 21. Primary Care Forum, 2026, 30(10): 56–58. DOI: 10.19435/j.1672-1721.2026.10.019. [5] Kamath V, Sivakumar B M, Callado Y G, et al. Non-invasive prenatal testing as a prenatal screening test in pregnancies following assisted reproductive technologies: a systematic review and meta-analysis of diagnostic test accuracy. Fertility and Sterility, 2026, DOI:10.1016/J.FERTNSTERT.2026.04.017 [6] Zhou M Y, Chen X, Lin L, et al. The influence of maternal age, gestational age and body mass index on the proportion of foetal cell-free DNA in maternal peripheral blood. Chinese Journal of Prenatal Diagnosis (Electronic Edition), 2021, 13(03): 21–25. [7] Wu Z Y, Jiang S Q, Huang M Z. Construction of a predictive model for factors influencing the proportion of foetal cell-free DNA in maternal peripheral blood based on the random forest algorithm. Medical Theory and Practice, 2026, 39(11): 1908–1911. DOI: 10.19381/j.issn.1001-7585.2026.11.028. [8] Zeng M, Zhang Y, Lu H L, et al. Investigation into the correlation between foetal cell-free DNA concentration in maternal peripheral blood and adverse perinatal outcomes. Chinese Journal of Perinatal Medicine, 2023, 26(05): 409–414. [9] Liscovitch-Brauer N, Mesika R, Rabinowitz T, et al. Machine learning-enhanced non-invasive prenatal testing for monogenic disorders. Prenatal Diagnosis, 2024, 44(5): 657–665. [10] Chawla N V, Bowyer K W, Hall L O, et al. SMOTE: synthetic minority over-sampling technique. Journal of Artificial Intelligence Research, 2002, 16: 321–357. [11] Zhu K D. A Study on Classification for Multi-Class Imbalanced Data and Its Applications. Inner Mongolia University, 2025. DOI:10.27224/d.cnki.gnmdu.2025.001057. [12] Rahimzadeh F, Khanli M L, Salehpour P, et al. From Heuristics to Deep Generative Models: A Critical Review of Protein Sequence Analysis Architectures for Clinical Decision Support Systems (CDSS). Expert Systems with Applications, 2026, 324132559–132559. DOI: 10.1016/J.ESWA.2026.132559. [13] Liang F, Pan G T, Wang Q Q, et al. A Robust Regression Algorithm Based on B-Splines. Journal of the Armoured Forces Engineering Academy, 2013, 27(1): 97–100. [14] Zhang J N, Zou Y H, Liu H Y. Application of Bayesian Hierarchical Models to Estimating Historical Fertility Rates in China. Mathematical Statistics and Management, 1–13 [10 June 2026]. https://doi.org/10.13860/j.cnki.sltj.20260424-002. [15] Zhang Y. A Method for Predicting lncRNA-Disease Associations Based on Ensemble Decision Trees. Yunnan University, 2020. DOI:10.27456/d.cnki.gyndu.2020.000180. [16] Zheng H, Gao X, Yang X, et al. A study on TFT-LCD manufacturing quality prediction methods using SHAP feature selection and ensemble learning. International Journal of Computer Integrated Manufacturing, 2026, 39(6): 825–841. DOI: 10.1080/0951192X.2025.2498083. [17] Dutta S, Bhattacharya S, Saha M M, et al. Oxidative Stress–Mediated Endothelial Dysfunction and Cell-Free Fetal DNA Release in Pre-eclampsia. Indian Journal of Clinical Biochemistry, 2026, (pre-publication): 1–10. DOI: 10.1007/S12291-026-01393-W. [18] Rong H H, Qiu J Y, Bao Z M, et al. Evaluation of the impact of pre-pregnancy body mass index and weight gain during pregnancy on pregnancy outcomes. Contemporary Medical Collection, 2026, 24(6): 43–45.