Cao Q, Sun H, Wang H, Liu X, Lu Y, Huo L. Comparative study of neonatal brain injury fetuses using machine learning methods for perinatal data.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE 2023;
240:107701. [PMID:
37480645 DOI:
10.1016/j.cmpb.2023.107701]
[Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Subscribe] [Scholar Register] [Received: 05/04/2021] [Revised: 06/20/2023] [Accepted: 06/28/2023] [Indexed: 07/24/2023]
Abstract
OBJECTIVE
CTG is used to record the fetus's fetal heart rate and uterine contraction signal during pregnancy. The prenatal fetal intrauterine monitoring level can be used to evaluate the fetal intrauterine safety status and reduce the morbidity and mortality of the perinatal fetus. Perinatal asphyxia is the leading cause of neonatal hypoxic-ischemic encephalopathy and one of the leading causes of neonatal death and disability. Severe asphyxia can cause brain and permanent nervous system damage and leave different degrees of nervous system sequelae.
METHODS
This paper evaluates the classification performance of several machine learning methods on CTG and provides the auxiliary ability of clinical judgment of doctors. This paper uses the data set on the public database UCI, with 2126 samples.
RESULTS
The accuracy of each model exceeds 80%, of which XGBoost has the highest accuracy of 91%. Other models are Random tree (90%), light (90%), Decision tree (83%), and KNN (81%). The performance of the model in other indicators is XGBoost (precision: 90%, recall: 93%, F1 score: 90%), Random tree (precision: 88%, recall: 91%, F1 score: 89%), lightGBM (precision: 87%, recall: 93%, F1 score: 90%), Decision tree (precision: 83%, recall: 86%, F1 score: 84%), KNN (precision: 77%, recall: 85%, F1 score: 81%).
CONCLUSION
The performance of XGBoost is the best of all models. This result also shows that using the machine learning method to evaluate the fetus's health status in CTG data is feasible. This will also provide and assist doctors with an objective assessment to assist in clinical diagnosis.
Collapse