Coronary Centerline Extraction from CCTA Using 3D-UNet

Valentin Ogrean,Remus Brad,Alexandru Dorobanțiu

doi:10.3390/fi13040101

Valentin Ogrean, Remus Brad + Show 1 more

Open Access

https://doi.org/10.3390/fi13040101

Copy DOI

Journal: Future Internet	Publication Date: Apr 19, 2021
Citations: 5	License type: CC BY 4.0

Affiliation: Lucian Blaga University of Sibiu

Abstract

The mesh-type coronary model, obtained from three-dimensional reconstruction using the sequence of images produced by computed tomography (CT), can be used to obtain useful diagnostic information, such as extracting the projection of the lumen (planar development along an artery). In this paper, we have focused on automated coronary centerline extraction from cardiac computed tomography angiography (CCTA) proposing a 3D version of U-Net architecture, trained with a novel loss function and with augmented patches. We have obtained promising results for accuracy (between 90–95%) and overlap (between 90–94%) with various network training configurations on the data from the Rotterdam Coronary Artery Centerline Extraction benchmark. We have also demonstrated the ability of the proposed network to learn despite the huge class imbalance and sparse annotation present in the training data.

Highlights

The idea for creating a 3D U-Net network came after observing the results of U-Net convolutional networks for biomedical image segmentation [21], which allowed a fully convolutional neural network to provide a good segmentation even when trained with a small training dataset
The belief was further consolidated when observing the results of the U-Net applied for liver segmentation and vessel exclusion [22], and for retina vessel segmentation, a similar task of centerline segmentation was conducted
To test the implementation of the proposed neural network, we chose the dataset from the Rotterdam Coronary Artery Algorithm Evaluation Framework [30]