Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Impact of Box-Cox Transformation on Machine-Learning Algorithms for Detecting Heartbeat Abnormalities
We aim to increase the accuracy of how we classify heartbeat data gathered by ECG.
Keywords: ECG signal analysis, biomedical signal analysis, data science, medical technologies, and digital health
The purpose of this study is to apply box-cox monotonic power transformation to physiologically relevant (e.g., RR intervals) and signal-based (e.g., noise and power) variables extracted from large electrocardiogram datasets to increase beat-type classification accuracy. Given the advances in machine learning, accurate detection of arrhythmia using electrocardiogram signals is challenging. There is, therefore, a need to develop new methodologies to improve the accuracy of arrhythmia detection.
The purpose of this study is to apply box-cox monotonic power transformation to physiologically relevant (e.g., RR intervals) and signal-based (e.g., noise and power) variables extracted from large electrocardiogram datasets to increase beat-type classification accuracy. Given the advances in machine learning, accurate detection of arrhythmia using electrocardiogram signals is challenging. There is, therefore, a need to develop new methodologies to improve the accuracy of arrhythmia detection.
- Extract features from ECG signals.
- Estimate box-cox parameters
- Test normality of transformed data
- Compare results before and after applying box-cox to machine learning algorithms
Tasks
- Literature review (10%)
- Data analysis (loading data, extracting features, and data transformation using box-cox) (40%)
- Design and implement a basic machine learning algorithm (20%)
- Test, compare and evaluate results (10%)
- Report and presentation (20%)
- Extract features from ECG signals. - Estimate box-cox parameters - Test normality of transformed data - Compare results before and after applying box-cox to machine learning algorithms
Tasks
- Literature review (10%) - Data analysis (loading data, extracting features, and data transformation using box-cox) (40%) - Design and implement a basic machine learning algorithm (20%) - Test, compare and evaluate results (10%) - Report and presentation (20%)
Dr. Moe Elgendi (moe.elgendi@hest.ethz.ch) will supervise the student at the Biomedical and Mobile Health Technology Research Group in ETH Zurich’s D-HEST Department of Health Sciences and Technology.
Dr. Moe Elgendi (moe.elgendi@hest.ethz.ch) will supervise the student at the Biomedical and Mobile Health Technology Research Group in ETH Zurich’s D-HEST Department of Health Sciences and Technology.
