Paper Number
AM17
Session
AI and ML Based Rheological Characterization
Title
Rheology of human blood and the connection to physiology: A data-driven approach
Presentation Date and Time
October 10, 2022 (Monday) 5:25
Track / Room
Track 6 / Mayfair
Authors
- Farrington, Sean M. (University of Delaware, Chemical and Biomolecular Engineering)
- Wagner, Norman J. (University of Delaware, Chemical and Biomolecular Engineering)
- Beris, Antony N. (University of Delaware, Chemical & Biomolecular Engineering)
Author and Affiliation Lines
Sean M. Farrington, Norman J. Wagner and Antony N. Beris
Chemical and Biomolecular Engineering, University of Delaware, Newark, DE 19716
Speaker / Presenter
Farrington, Sean M.
Keywords
experimental methods; bio-fluids; ML based
Text of Abstract
Blood rheology plays a critical role in the transport of oxygen, nutrients, waste, and immune response with potential to assess cardiovascular disease risk in early stages [Beris et al., Soft Matter 2021, 17 (47), 10591-10613.]. Large variations in rheology occur for both healthy and diseased blood. Diseases like hypertension, sickle cell anemia, and diabetes increase the viscosity of blood. The aggregation of red blood cells varies with physiology such as their volume fraction and the concentration of fibrinogen, a protein that promotes interactions in blood.
In this work, we aim to develop a quantitative connection between blood physiology and rheology. Establishing this connection provides two benefits. First, the connection can unlock rheology information from physiology in standard blood tests. With this information blood rheology can become a more powerful diagnostic technique which provides rapid data to complement biochemical analyses. Second, dynamic simulations of blood flow can be informed from this connection. A data-driven approach is used for developing nonlinear relationships between physiological and rheological parameters. The first objective is to use the model for Casson yield stress and viscosity developed by Apostolidis and Beris [Apostolidis et al. Journal of Rheology 2014, 58 (3), 607-633.] as a source for generating pseudo data. The purpose of using pseudo data is to compare machine learning algorithms on a model blood rheology system with any size dataset. The second objective builds a connection between rheological constitutive relationships and physiology using steady state and transient blood rheology data by Horner et al. [Horner et al. Journal of Rheology 2019, 63 (5), 799-813.]. Thus, a model to predict rheology properties from blood physiology is established. If this work is successful, a method for building this connection will be validated. Future work can amplify the connection by collecting larger sets of well-posed blood rheology data.