Cardiovascular Digital Twins from Time-Resolved CT

It is known that atrial fibrillation is related to other cardiovascular diseases, such as stroke. However, the direction of the relation is uncertain. Therefore, there is a need for novel methods to perform further in-depth investigations on the cardiovascular system and its mechanisms. Such a novel method is the use of computational mathematical models. By leveraging the excellent spatial resolution and rapid acquisition times of time-resolved CT, lumped parameter models can be made subject-specific with data about haemodynamic variables obtained with time-resolved CT. This provides a novel and effective method for in-depth investigations of the cardiovascular system. In this study, haemodynamic variables of a cohort of nine subjects, who all had been diagnosed with early stage of atrial fibrillation, were used together with a lumped parameter computational mathematical model. Together with subject-specific haemodynamic variables and a lumped parameter model, a cardiovascular digital twin was created. The aim in this study was to investigate the possibility of using haemodynamic variables obtained from time-resolved CT of subjects diagnosed with early stage atrial fibrillation with time-resolved CT as estimation data for a lumped parameter model, and whether the lumped parameter model can fit to the data from time-resolved CT. Furthermore, it was investigated whether there was a difference in how well the digital twin based on time-resolved CT data, compared to a digital twin based on 4D flow MRI data, could provide trustworthy simulations and predictions of hemodynamic variables of the left side of the heart. The study showed that a lumped parameter model can fit to data about haemodynamic variables obtained with time-resolved CT, and that it is achievable to apply subject-specific data from subjects diagnosed with early stage atrial fibrillation in sinus rhythm into a lumped parameter computational model. Additionally, the study has shown the difference in estimation data between the digital twins, and that the digital twin based on time-resolved CT data outperforms the digital twin based on 4D flow MRI data in terms of how well the model could fit the data it was given. However, it is important to mention that to get a conclusive comparison of the digital twins, greater computational power is needed. Together with greater computational power, an implementation of further validation methods is required. In conclusion, this master’s thesis has shown promising results in the use of zero-dimensional computational models together with subject-specific data obtained from time-resolved CT. The results of the study have shown that it is possible to use a lumped parameter model, made subject-specific with subject-specific haemodynamic variables acquired with time-resolved CT from subjects diagnosed with early stage AF. Additionally, the results of the study have shown the difference in estimation data between CT and MRI, together with that the CT-based digital twin performs more trustworthy simulations and predictions of haemodynamic variables of the left side of the heart than the MRI-based digital twin. The results yield a promising future in cardiovascular digital twins of the left side of the heart, to assist in diagnosis, treatment planning, disease predictions, and in studying the mechanisms of the left side of the heart.