Author(s)

Yu’ang Zhou, Xiaolin Liu, Jiayi He

Corresponding Author

Yu’ang Zhou

Abstract

With the widespread adoption of NIPT prenatal testing technology for assessing foetal health, unhealthy foetuses can be detected more rapidly, thereby extending the therapeutic window. However, NIPT can only guarantee its validity when performed under specific conditions. This paper therefore commences by applying k-means clustering to relevant data using the Cox proportional hazards model, thereby determining the optimal testing timeframe based on existing data. An error analysis of various influencing factors revealed that height exhibits the lowest sensitivity. Considering the differences between male and female chromosomes, this study constructed features from foetal data and developed a neural network model based on a multi-layer perceptron (MLP) for training and learning. This model achieved a sensitivity of 91%, specificity of 82%, and accuracy of 86%.

Keywords

Cox proportional hazards model; k-means clustering; MLP; Neural networks