PROJECT LEADER
Non-invasive and intravascular identification and drug-eluting balloon treatment of vulnerable lipid-rich plaques
Developing non-invasive AI-based method that detects and characterizes lipid-rich plaque in computed tomography coronary angiography.
ONGOING
2020
|
2024
RESEARCH AREAS
About project
More than half of the acute coronary syndromes result from the rupture of a lipid-rich plaque (LRP), thus the detection hereof is of great importance for prognosis. To date, these LRPs can only be visualized using the invasive imaging techniques of near-infrared spectroscopy (NIRS), combined with intravascular ultrasound (IVUS), which is currently the gold standard. In this project we aim to develop non-invasive AI-based method that allows LRP detection and characterization with computed tomography coronary angiography (CTCA). The project is a collaboration with the Department of Cardiology and funded by Health-Holland.
Members
Our publications & software
Publications
Journal articlesR. L. M. Van Herten, N. Hampe, R. A. P. Takx, K. J. Franssen, Y. Wang, D. Suchá, J. P. Henriques, T. Leiner, R. N. Planken, I. Išgum, "Automatic Coronary Artery Plaque Quantification and CAD-RADS Prediction using Mesh Priors", IEEE Transactions on Medical Imaging, 2023.
InproceedingsR.L.M. van Herten, L. van Harten, N. Planken, I. Isgum, "Generative Adversarial Networks for Coronary CT Angiography Acquisition Protocol Correction with Explicit Attenuation Constraints", Medical Imaging with Deep Learning, 2023; 227: 1288-1303.
Software
Automatic Coronary Artery Plaque Quantification and CAD-RADS Prediction using Mesh Priors
Code for the paper Automatic Coronary Artery Plaque Quantification and CAD-RADS Prediction using Mesh Priors, van Herten et al. 2023, IEEE Transactions on Medical Imaging (IEEE-TMI). This code uses coronary artery centerline priors to create mesh segmentations of the coronary artery lumen, as well as calcified and non-calcified plaque. The mesh segmentations are used to compute plaque volumes and to predict the CAD-RADS score.
Generative Adversarial Networks for Coronary CT Angiography Acquisition Protocol Correction with Explicit Attenuation Constraints
A method for the automatic correction of CTA images acquired with a suboptimal level of contrast enhancement. The method learns to identify and correct the attenuation in coronary arteries by enforcing boundary conditions on coronary artery centerlines, while the GAN architecture constrains the generative model to produce realistic looking images similar to the input image.