TKE-Net: Deep Learning for Estimation of Super-Resolved Turbulent Kinetic Energy Maps from 4D-Flow MRI Data

Arterial stenosis is one of the most prevalent diseases with significant morbidity and mortality requiring accurate quantification of hemodynamic parameters for diagnosis and prognosis. In particular, variations in velocity derivatives have been correlated with variations in pressure gradient, an im...

Full description

Saved in:
Bibliographic Details
Published in2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI) pp. 1 - 4
Main Authors Kazemi, Amirkhosro, Stoddard, Marcus, Amini, Amir A
Format Conference Proceeding
LanguageEnglish
Published IEEE 18.04.2023
Subjects
4D Flow MRI
Arterial Stenosis
Deep Learning
Energy resolution
Hemodynamics
Magnetic resonance imaging
Phantoms
Superresolution
Three-dimensional displays
Training
Turbulent Kinetic Energy
Online AccessGet full text

Cover

More Information
Summary:Arterial stenosis is one of the most prevalent diseases with significant morbidity and mortality requiring accurate quantification of hemodynamic parameters for diagnosis and prognosis. In particular, variations in velocity derivatives have been correlated with variations in pressure gradient, an important marker of hemodynamic significance of stenosis. 4D flow MRI provides time-resolved 3D velocity mapping, however, image denoising and super-resolution techniques are required for precise velocity fluctuation quantification. To address this issue, we propose TKE-Net, a novel network that uses a ResNet convolutional neural network to estimate Turbulent Kinetic Energy (TKE). We trained and tested the network with high-resolution simulated CFD data in a phantom model of arterial stenosis. Our proposed network was further tested on in-vitro 4D flow MRI data in identical geometry, demonstrating good accuracy in estimating TKE.
ISSN:1945-8452
DOI:10.1109/ISBI53787.2023.10230629