Dr. Vasiliki Bikia
1 Dec 2023, 17:30 Athens time, 145Π58
Zoom link: https://tuc-gr.zoom.us/j/99893977299?pwd=NzdZM214NE1NczI2bjhMMVBCcVY2dz09
Harnessing Physics-Driven Modelling and Artificial Intelligence for Non-Invasive Cardiovascular Health Monitoring
In a progressively aging population, it is of utmost importance to develop reliable, non-invasive, and cost-effective tools to estimate biomarkers that can be indicative of cardiovascular risk. Clinical parameters directly measured in the heart or the aorta are crucial for the diagnosis and management of disease. However, their clinical use is severely hampered by their invasive nature, cost, or need for special equipment.
In cardiovascular medicine, therapeutic decisions hinge on easily accessible indicators, such as ventricular ejection fraction (EF) and blood pressure (BP). While these metrics have stood the test of time, offering a snapshot of the current hemodynamic state of the heart and vascular system, there is a widespread acknowledgment among experts that they provide only a partial view of the true physiological intricacies of an individual's cardiovascular system. Why do BP and EF continue to rank among the most frequently used biomarkers? The answer lies in their practicality in clinical use.
Therefore, there exists a pressing need for convenient, non-invasive, and cost-efficient predictive methodologies to assist the clinician with cardiovascular assessment. This talk will present original predictive algorithms suitable for estimating major cardiovascular biomarkers from commonly measured clinical data. In particular, we will talk about monitoring tools that advance non-invasive methods for estimating central hemodynamics, including aortic systolic BP and cardiac output, by employing an innovative inverse problem-solving technique and leveraging measurements of cuff pressure and pulse wave velocity. Supervised learning may further enhance the precision of these estimations, offering a promising, time-efficient alternative for monitoring these critical parameters. Additionally, we will addresses the non-invasive prediction of end-systolic elastance (Ees), a key indicator of left ventricular systolic function. By incorporating EF and clinically relevant systolic time intervals, we can demonstrate significant improvements in the accuracy of Ees estimation, paving the way for a practical and easily implementable method to assess left ventricular function. Furthermore, we will present a novel approach utilizing the morphology of brachial BP waveforms and convolution neural networks to predict Ees and will test its accuracy on a synthetic population of virtual subjects. This new technique shows remarkable promise, providing accurate predictions across a wide range of contractility values and loading conditions. The talk will also focus on ways to enhance the assessment of global and local indices of arterial tree’s elasticity with the use of pressure data from multiple arterial locations. These advancements represent a significant leap forward in non-invasive hemodynamic monitoring, with potential applications ranging from efficient clinical assessments to integration into wearable technologies.
About the Speaker
Dr. Vasiliki Bikia is a Postdoctoral Researcher in the Bioengineering Department, at the Swiss Federal Institute of Technology, Switzerland. She specializes in non-invasive cardiovascular health assessment and develops clinical and digital monitoring tools, driving advancements in machine learning for improved patient management. She received her Advanced Diploma degree in Electrical and Computer Engineering with honors from the Aristotle University of Thessaloniki (AUTH), Greece, in 2017, and her Ph.D. degree in Bioengineering from the Swiss Federal Institute of Technology of Lausanne, Switzerland, in 2021, respectively.
Her research interests include machine learning, non-invasive health monitoring, clinical language models, digital biomarkers, and physiological ageing. Dr. Bikia is developing new methods to assess cardiovascular health via the prediction of key hemodynamic and cardiac parameters. She developed a novel modelling approach to predict mean aortic flow from readily available physiological data, facilitating non-invasive, continuous monitoring inside and outside the clinic. In addition, she is working on biomedical applications in the fields of patient-specific physic-based modelling, synthetic data generation, and enhanced prediction of patient outcomes via multimodal integration of clinical data and language representations.
Dr. Bikia is an active member of the wider academic communities, including IEEE, the Artery Society, and the European Network for Research in Vascular Ageing.