Related Experiment Video
Updated: Jan 18, 2026
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
43.4K
Reducing motion artifacts in craniocervical background subtraction angiography with deformable registration and
Chaochao Zhou1, Ramez N Abdalla1,2, Dayeong An1
1Department of Radiology, Northwestern University and Northwestern Medicine, Chicago, IL, 60611, United States.
Radiology Advances
|September 10, 2025
Summary
This study introduces a deep learning method to reduce motion artifacts in digital subtraction angiography (DSA), significantly improving image quality and reducing artifacts for better craniocervical angiography.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Healthcare
- Neuroradiology
Background:
- Digital subtraction angiography (DSA) is prone to misregistration artifacts caused by patient motion.
- Existing registration methods for DSA often lack real-time application due to iterative optimization.
Purpose of the Study:
- To develop a fast, unsupervised deep learning deformable registration model for craniocervical angiography.
- To reduce DSA misregistration without compromising spatial resolution or introducing new artifacts.
Main Methods:
- Extended the HyperMorph deep learning framework for deformable registration.
- Introduced novel image similarity loss functions with vessel layer estimation for robust background registration.
- Utilized a dataset of 5,240 angiographic series for training and testing.
Main Results:
- Learning-based background subtraction angiography (BSA) significantly improved vascular fidelity, reduced artifacts, and enhanced overall image quality compared to traditional DSA.
- BSA outperformed affine registration-based methods (P < .0001).
- Achieved an average inference time of 30 milliseconds per frame.
Conclusions:
- Deep learning deformable registration effectively reduces motion artifacts in DSA.
- The developed method enhances image quality in craniocervical angiography.
