Related Experiment Video
Updated: Jul 17, 2025
Improved Registration of 3D CT Angiography with X-ray Fluoroscopy for Image Fusion During Transcatheter Aortic Valve Implantation
Published on: June 3, 2018
Automated cross-sectional view selection in CT angiography of aortic dissections with uncertainty awareness and
Antonio Pepe1, Jan Egger2, Marina Codari3
1Graz University of Technology, Institute of Computer Graphics and Vision, Inffeldgasse 16/II, 8010 Graz, Austria; Stanford University, School of Medicine, 3D and Quantitative Imaging Lab, 300 Pasteur Drive Stanford, CA 94305, USA; Computer Algorithms for Médicine (Café) Laboratory, Graz, Austria.
Insights
This study introduces novel deep learning methods to automatically determine the orientation of aortic cross-sectional planes for surveillance imaging. These AI-driven approaches offer faster and more reproducible measurements for chronic aortic diseases than manual methods.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Cardiovascular Imaging
Background:
- Chronic aortic diseases require regular surveillance imaging with serial diameter measurements.
- Manual determination of cross-sectional plane orientation is time-consuming and operator-dependent.
- Current centerline-based methods are unreliable for chronic aortic dissections due to complex anatomy and flow dynamics.
Purpose of the Study:
- To develop and evaluate novel automated methods for predicting the orientation of cross-sectional planes in the aorta.
- To improve the efficiency and reproducibility of aortic surveillance imaging, particularly for chronic aortic dissections.
- To leverage convolutional neural networks and uncertainty quantification for this task.
Main Methods:
- Development of three alternative approaches (INS, MCDS, MCDbS) using convolutional neural networks.
- Application of uncertainty quantification methods to predict plane orientation (ϕ,θ).
- Training and validation on a dataset of 162 CTA volumes with 3273 manual annotations.
Main Results:
- The proposed methods provide faster and more reproducible results compared to expert users and centerline methods.
- Despite interoperator variability in training data, the models achieve performance comparable to the state of the art.
- The remaining disagreement aligns with the variability observed among expert annotators.
Conclusions:
- Automated prediction of aortic cross-sectional plane orientation using deep learning is feasible and efficient.
- These methods can significantly enhance the speed and reproducibility of aortic disease monitoring.
- The approach effectively handles the complexities of chronic aortic dissections, offering a valuable tool for clinical practice.
Abstract:
Surveillance imaging of patients with chronic aortic diseases, such as aneurysms and dissections, relies on obtaining and comparing cross-sectional diameter measurements along the aorta at predefined aortic landmarks, over time. The orientation of the cross-sectional measuring planes at each landmark is currently defined manually by highly trained operators. Centerline-based approaches are unreliable in patients with chronic aortic dissection, because of the asymmetric flow channels, differences in contrast opacification, and presence of mural thrombus, making centerline computations or measurements difficult to generate and reproduce. In this work, we present three alternative approaches - INS, MCDS, MCDbS - based on convolutional neural networks and uncertainty quantification methods to predict the orientation (ϕ,θ) of such cross-sectional planes. For the monitoring of chronic aortic dissections, we show how a dataset of 162 CTA volumes with overall 3273 imperfect manual annotations routinely collected in a clinic can be efficiently used to accomplish this task, despite the presence of non-negligible interoperator variabilities in terms of mean absolute error (MAE) and 95% limits of agreement (LOA). We show how, despite the large limits of agreement in the training data, the trained model provides faster and more reproducible results than either an expert user or a centerline method. The remaining disagreement lies within the variability produced by three independent expert annotators and matches the current state of the art, providing a similar error, but in a fraction of the time.
Related Concept Videos
01:28Imaging Studies for Cardiovascular System V: CT
01:21Aneurysm II: Clinical Manifestations and Diagnostic Studies