Coronary Centreline Extraction from Coronary CT-Angiograms
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Saudi Digital Library
Abstract
Key Massages
• A coronary centreline extraction workflow was created in MeVisLab and validated on
MICCAI`08 Challenge database.
• The workflow showed promise, accuracy=0.4 mm, when applied on a dataset from
InSilc EU project.
Abstract
Objectives
An accurate coronary centreline extraction from computed tomography angiography (CTA) images is a key in developed computational workflows, analysing haemodynamic blood flow in the coronary arteries and myocardium. This study aimed to validate a MeVisLab workflow for coronary centreline extraction from CTA images based on the MICCAI`08 Coronary Centreline Extraction Challenge dataset. Once validated, the workflow used to extract coronary centrelines from the InSilc EU project.
Methods
A MeVisLab workflow of Multiple Hypothesis Tracking (MHT) method was generated and used to extract coronary centrelines from CTA images in MICCAI`08 Challenge. The extracted centrelines were compared against the MICCAI`08 Challenge reference centrelines to ensure they fall into the inter-observer variability range. The InSilc coronary centrelines were extracted using the same workflow and compared against those extracted by Shape Regression method. The distance between centrelines obtained from the methods was calculated as measure of accuracy. A library of 3D InSilc centrelines was generated.
Results
The MeVisLab MHT workflow was validated by reproducing the MICCAI`08 Challenge results with high accuracy (93.5%). Nine InSilc datasets were extracted with the same MeVisLab workflow. Once compared against those extracted by Shape Regression method, an overall distance of 0.49 mm, 0.36 mm, 0.52 mm, and 0.42 mm were observed between LAD, LCX, 1D and Ramus centrelines obtained from the two methods.
Conclusions
A MHT MeVisLab workflow was created and validated based on two datasets of coronary CTA images. A coronary centreline library was generated that can be used to develop computational models of blood flow within the InSilc EU project.