University of Leeds
Efficient ensemble simulation methods for in-silico trials of endovascular medical devices
Abstract
dc:description.abstractIn-silico trials (ISTs) use computational modelling and simulation in virtual patients to evaluate medical device performance. Despite early promise, various challenges prevent the use of ISTs from becoming common practice in medical device development. Three significant challenges are: (i) Prohibitive costs due to complex computational models that require excessive resources and time to execute. (ii) Lack of exemplar ISTs demonstrating their effectiveness in generating evidence for medical device performance. (iii) Lack of scalability and reproducibility due to computational modelling pipelines requiring significant expert manual input. In this thesis, challenge (i) was addressed through a comprehensive literature review into reduced order modelling and machine learning techniques that can accelerate the computational models that are essential in ISTs. Challenge (ii) was addressed by performing the FD-PComA IST into flow diversion (FD) of posterior communicating artery (PComA) aneurysms, which are a common sub-group currently not approved for treatment with the most widely used flow diverter. PComA aneurysm treatment is complicated by the presence of fetal posterior circulation (FPC), which has an estimated prevalence of 4-29% and is more common in black than white ethnicities. Given these factors, FD-PComA demonstrates the effectiveness of ISTs in generating evidence for less-studied scenarios and demographics. Challenge (iii) was addressed in FD-PComA through automation of the modelling steps. The results of FD-PComA demonstrate that flow diversion is less effective in FPC patients and that PComA and aneurysm morphology do not influence treatment performance. Challenge (i) was addressed further through the development of a machine learning reduced order model (ML-ROM) for evaluating aneurysm blood flow subject to physiological variation, which is a relevant problem for IST applications. The ML-ROM was approximately 98% accurate in evaluating the velocity solution and provided an acceleration of five orders of magnitude relative to a computational fluid dynamics model for the same problem.
Degree
thesis:*- Name dc:type.qualificationname
- Ph.D
- Level dc:type.qualificationlevel
- doctoral
- Grantor dc:publisher.institution
- University of Leeds
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- MacRaild, Michael
- Advisors dc:contributor.advisor
-
- Frangi, Alejandro
- Lassila, Toni
- Sarrami-Foroushani, Ali
- ORCID person:identifier.orcid
Rights
- Language dc:language.dcterms_iso639_2
- eng
Identifiers
dc:identifier.*- Dc Identifier Grantnumber
- P/L01615X/1