Development of a syncope classification algorithm from physiological signals acquired in tilt-table test
Syncope also known as transient loss of consciousness which caused problem to human daily life. Since machine learning is much more advanced, classification of syncope can be done with machine learning. Head-up tilt table test (HUTT) having a lengthy procedure and might causing patient to feel disco...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Year Project / Dissertation / Thesis |
| Published: |
2023
|
| Subjects: | |
| Online Access: | http://eprints.utar.edu.my/5809/1/BI_1801320_Final_GAN_MING_HONG.pdf http://eprints.utar.edu.my/5809/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Syncope also known as transient loss of consciousness which caused problem to human daily life. Since machine learning is much more advanced, classification of syncope can be done with machine learning. Head-up tilt table test (HUTT) having a lengthy procedure and might causing patient to feel discomfort during the test. Aim of this study is to design an algorithm which able to classify syncope patient based on their physiological signal. In this study, electrocardiogram (ECG) and blood pressure (BP) signal has been collected from 144 subjects with head-up tilt table test (HUTT) by clinicians. Several features have been extracted from heart rate, systolic and diastolic blood pressure. There are 8 set of feature selection model has built and a total of 24 set of classifiers with 3 different type of classification techniques were developed. Additionally, stratified 5-fold cross-validation was performed to evaluate the performance of proposed model. Features that selected for the classification is mean of systolic and diastolic blood pressure, standard deviation of real variability of diastolic blood pressure, and the mean of systolic blood pressure in low and high frequency ratio. The proposed model yielded the following result: 85.71% sensitivity, 91.43% specificity, 88.18% F1-score and 88.57% accuracy. Future work can be focus on utilise more different type of classifier and carry out external cross validation for achieving a better classification model. |
|---|