This project began as a three-month collaboration with Dr. Rifai Chai and evolved into a year-long final year project at Swinburne University of Technology.
The primary issue addressed was the significant increase in Emergency Department (ED) waiting times in Australia, often leading to delayed or missed vital signs monitoring, potentially compromising patient care.
Beyond the ED, coronary heart disease (CHD) remains prevalent, and the use of Ambulatory Blood Pressure Monitoring (ABPM) helps diagnose CHD, but it significantly disrupts patients' routines due to the 24-48 hour continuous cuff measurements.
While novel methods using both Photoplethysmogram (PPG) and Electrocardiogram (ECG) show potential in predicting blood pressure through non-invasive, cuff free methodology, they're not recommended for patients with cardiovascular problems due to assumptions about arterial wall thickness.
A blood pressure prediction method using only single-lead ECG waveforms and a regression-based Machine Learning model was developed in MATLAB. By analysing the instantaneous frequency of an ECG waveform, the model achieved mean absolute errors (MAE) of 6.94 mmHg for systolic and 3.98 mmHg for diastolic blood pressure. These results demonstrate the potential of ECG signal complexity for accurate blood pressure prediction.