Document Type
5 minute oral
Start Date
6-3-2026 1:30 PM
Embargo Date
12-1-2027
Description
Heart failure affects approximately 64 million people worldwide and is associated with impaired cardiac function, frequent hos- pitalisations, and reduced quality of life. Cardiac output is a key indicator of heart function, but current gold-standard measurement techniques are invasive and unsuitable for continuous or long-term monitoring.
In this work, we describe the development of a statistical algorithm to predict cardiac output from pulmonary artery pressure (PAP) waveforms, combined with routinely collected clinical data. Individual cardiac beats are treated as functional observations, and multilevel functional principal components analysis (mFPCA) is used to summarise beat-level and patient-level variation in waveform shape. An automatic signal-processing pipeline enables automatic determination of cardiac cycles, waveform alignment, and dicrotic notch identification, thereby supporting physiologically meaningful feature extraction. Results from an initial cohort demonstrate the promise of this approach while highlighting key challenges. This work demonstrates how functional data analysis can transform high-frequency physiological signals into clinically relevant metrics, with the potential to improve remote heart failure management
Included in
Engineering Commons, Life Sciences Commons, Medicine and Health Sciences Commons, Physical Sciences and Mathematics Commons
Predicting Cardiac Output from Pulmonary Artery Pressure
Heart failure affects approximately 64 million people worldwide and is associated with impaired cardiac function, frequent hos- pitalisations, and reduced quality of life. Cardiac output is a key indicator of heart function, but current gold-standard measurement techniques are invasive and unsuitable for continuous or long-term monitoring.
In this work, we describe the development of a statistical algorithm to predict cardiac output from pulmonary artery pressure (PAP) waveforms, combined with routinely collected clinical data. Individual cardiac beats are treated as functional observations, and multilevel functional principal components analysis (mFPCA) is used to summarise beat-level and patient-level variation in waveform shape. An automatic signal-processing pipeline enables automatic determination of cardiac cycles, waveform alignment, and dicrotic notch identification, thereby supporting physiologically meaningful feature extraction. Results from an initial cohort demonstrate the promise of this approach while highlighting key challenges. This work demonstrates how functional data analysis can transform high-frequency physiological signals into clinically relevant metrics, with the potential to improve remote heart failure management