Can a magnetic resonance imaging simulator generate realistic flow images?
(2025) BMEL01 20252Division for Biomedical Engineering
- Abstract
- Background: Magnetic resonance flow imaging is vital tool for modern cardiovascular medicine. However, it is resource intensive. Simulations can help develop magnetic resonance sequencing methods and can also facilitate use of AI models in the future by providing realistic test models to train on. Simulations of magnetic resonance imaging are often computationally demanding and the accuracy of the images needs to be reliable. The goal of this study was to evaluate whether one such simulation framework, CMRsim, could generate realistic flow images.
Method: One model was implemented and evaluated against literature regarding magnetic resonance flow imaging characteristics such as sequencing, image quality and especially error in flow. 19... (More) - Background: Magnetic resonance flow imaging is vital tool for modern cardiovascular medicine. However, it is resource intensive. Simulations can help develop magnetic resonance sequencing methods and can also facilitate use of AI models in the future by providing realistic test models to train on. Simulations of magnetic resonance imaging are often computationally demanding and the accuracy of the images needs to be reliable. The goal of this study was to evaluate whether one such simulation framework, CMRsim, could generate realistic flow images.
Method: One model was implemented and evaluated against literature regarding magnetic resonance flow imaging characteristics such as sequencing, image quality and especially error in flow. 19 simulations were performed to evaluate partial volume error in correlation to increasing spatial resolution. Mesh independence was evaluated by comparing different mesh resolution at a set spatial resolution and observing its effect on flow rate.
Results: The qualitative results show simulated flow images with realistic characteristics and artifacts. The quantitative results show error convergence in agreement with literature, with a minimum of 6 pixels per diameter spatial resolution. The effect of mesh resolution on flow rate could not be determined due to the non-deterministic reseeding of particles in the software. The compute times were observed to increase proportionally to the spatial resolution.
Conclusions: The images generated were realistic concerning error convergence as well as common qualitatively observable artifacts Further study is necessary to explore the full possibilities of the framework. No conclusions could be drawn regarding mesh independence. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9216108
- author
- Mårtensson, Hjalmar LU
- supervisor
- organization
- alternative title
- Kan en magnetresonans-simulator skapa realistiska flödesbilder?
- course
- BMEL01 20252
- year
- 2025
- type
- M2 - Bachelor Degree
- subject
- language
- English
- additional info
- 2025-21
- id
- 9216108
- date added to LUP
- 2026-01-07 13:58:08
- date last changed
- 2026-01-07 13:58:08
@misc{9216108,
abstract = {{Background: Magnetic resonance flow imaging is vital tool for modern cardiovascular medicine. However, it is resource intensive. Simulations can help develop magnetic resonance sequencing methods and can also facilitate use of AI models in the future by providing realistic test models to train on. Simulations of magnetic resonance imaging are often computationally demanding and the accuracy of the images needs to be reliable. The goal of this study was to evaluate whether one such simulation framework, CMRsim, could generate realistic flow images.
Method: One model was implemented and evaluated against literature regarding magnetic resonance flow imaging characteristics such as sequencing, image quality and especially error in flow. 19 simulations were performed to evaluate partial volume error in correlation to increasing spatial resolution. Mesh independence was evaluated by comparing different mesh resolution at a set spatial resolution and observing its effect on flow rate.
Results: The qualitative results show simulated flow images with realistic characteristics and artifacts. The quantitative results show error convergence in agreement with literature, with a minimum of 6 pixels per diameter spatial resolution. The effect of mesh resolution on flow rate could not be determined due to the non-deterministic reseeding of particles in the software. The compute times were observed to increase proportionally to the spatial resolution.
Conclusions: The images generated were realistic concerning error convergence as well as common qualitatively observable artifacts Further study is necessary to explore the full possibilities of the framework. No conclusions could be drawn regarding mesh independence.}},
author = {{Mårtensson, Hjalmar}},
language = {{eng}},
note = {{Student Paper}},
title = {{Can a magnetic resonance imaging simulator generate realistic flow images?}},
year = {{2025}},
}