Predicting Birth Outcome Using Cardiotocography and Machine Learning
(2022) In Master’s Theses in Mathematical Sciences FMSM01 20212Mathematical Statistics
- Abstract
- Cardiotography, CTG, is a monitoring method that is commonly used during childbirth. The method measures the fetal heart rate, FHR, alongside the uterine contractions, TOCO. Clinicians use this tool to evaluate the health of the infant, and to observe changes which might imply hypoxia, lack of oxygen supply, for the fetus. However, abnormalities in the CTG are often non-specific, making the interpretation difficult.
This master's thesis aims to extract features from the CTG and to use these features in different types of machine learning classifiers to predict the child's health after birth. The target is to make the interpretation of the CTG easier for clinicians and to find patterns that could imply hypoxia.
By using the... (More) - Cardiotography, CTG, is a monitoring method that is commonly used during childbirth. The method measures the fetal heart rate, FHR, alongside the uterine contractions, TOCO. Clinicians use this tool to evaluate the health of the infant, and to observe changes which might imply hypoxia, lack of oxygen supply, for the fetus. However, abnormalities in the CTG are often non-specific, making the interpretation difficult.
This master's thesis aims to extract features from the CTG and to use these features in different types of machine learning classifiers to predict the child's health after birth. The target is to make the interpretation of the CTG easier for clinicians and to find patterns that could imply hypoxia.
By using the gestational age as the first feature, it was decided to split the data into two cases, preterm pregnancies and full term including postterm pregnancies. For each case, the gestational age was tested as a feature in different classifiers, which provided a benchmark for comparison when testing new features. During the investigation of the FHR signal, a total of eight features were tested. The extracted features were tested with the gestational age separately, followed by a test using all derived features, and lastly a test using all features but the gestational age was made. Some features increased the classifiers' sensitivity, or specificity, but none made the predictions significantly more accurate.
The conclusions of this thesis were that the gestational age should only have been used for stratifying the data into two separate cases, since it for the preterm pregnancies carried much information, therefore derived features did not contribute with further information. For the other case, the gestational age did not carry any information which showed in the results of the last two tests when using all features together. The extracted features should have been tested in different combinations of each other to see if they contributed with correlating information or not. Additional research is required to create an accurate prediction model, e.g., by investigating how the FHR signal correlates with the TOCO signal. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9076750
- author
- Öder, Josefine LU
- supervisor
-
- Andreas Jakobsson LU
- Karl Åström LU
- Ida Arvidsson LU
- Karel Marsal LU
- Karin Källén LU
- organization
- course
- FMSM01 20212
- year
- 2022
- type
- H2 - Master's Degree (Two Years)
- subject
- publication/series
- Master’s Theses in Mathematical Sciences
- report number
- LUTFMS-3440-2022
- ISSN
- 1404-6342
- other publication id
- 2022:E13
- language
- English
- id
- 9076750
- date added to LUP
- 2022-03-23 13:23:40
- date last changed
- 2022-03-23 13:23:40
@misc{9076750,
abstract = {{Cardiotography, CTG, is a monitoring method that is commonly used during childbirth. The method measures the fetal heart rate, FHR, alongside the uterine contractions, TOCO. Clinicians use this tool to evaluate the health of the infant, and to observe changes which might imply hypoxia, lack of oxygen supply, for the fetus. However, abnormalities in the CTG are often non-specific, making the interpretation difficult.
This master's thesis aims to extract features from the CTG and to use these features in different types of machine learning classifiers to predict the child's health after birth. The target is to make the interpretation of the CTG easier for clinicians and to find patterns that could imply hypoxia.
By using the gestational age as the first feature, it was decided to split the data into two cases, preterm pregnancies and full term including postterm pregnancies. For each case, the gestational age was tested as a feature in different classifiers, which provided a benchmark for comparison when testing new features. During the investigation of the FHR signal, a total of eight features were tested. The extracted features were tested with the gestational age separately, followed by a test using all derived features, and lastly a test using all features but the gestational age was made. Some features increased the classifiers' sensitivity, or specificity, but none made the predictions significantly more accurate.
The conclusions of this thesis were that the gestational age should only have been used for stratifying the data into two separate cases, since it for the preterm pregnancies carried much information, therefore derived features did not contribute with further information. For the other case, the gestational age did not carry any information which showed in the results of the last two tests when using all features together. The extracted features should have been tested in different combinations of each other to see if they contributed with correlating information or not. Additional research is required to create an accurate prediction model, e.g., by investigating how the FHR signal correlates with the TOCO signal.}},
author = {{Öder, Josefine}},
issn = {{1404-6342}},
language = {{eng}},
note = {{Student Paper}},
series = {{Master’s Theses in Mathematical Sciences}},
title = {{Predicting Birth Outcome Using Cardiotocography and Machine Learning}},
year = {{2022}},
}