Related Experiment Video
Updated: Nov 11, 2025
Improved Registration of 3D CT Angiography with X-ray Fluoroscopy for Image Fusion During Transcatheter Aortic Valve Implantation
Published on: June 3, 2018
Deep Learning-based Angiogram Generation Model for Cerebral Angiography without Misregistration Artifacts
Daiju Ueda1, Yutaka Katayama1, Akira Yamamoto1
1From the Departments of Diagnostic and Interventional Radiology (D.U., A.Y., S.L.W., H. Tatekawa, H. Takita, T.H., A.S., Y.M.), Neurosurgery (T. Ichinose, H.A., Y.W., T.G.), and Medical Statistics (D.K.), Graduate School of Medicine, Osaka City University, 1-4-3 Asahi-machi, Abeno-ku, Osaka 545-8585, Japan; and Department of Radiology, Osaka City University Hospital, 1-5-7 Asahi-machi, Abeno-ku, Osaka, 545-8586, Japan (Y.K., T. Ichida).
A new deep learning (DL) model generates clear digital subtraction angiography (DSA)-like cerebral angiograms from dynamic angiograms, overcoming patient movement artifacts. This AI tool offers clinically useful images for medical procedures.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Radiology
Background:
- Digital subtraction angiography (DSA) is crucial for visualizing blood vessels but can be degraded by patient movement artifacts, leading to unclear images.
- These misregistration artifacts in DSA can interrupt critical medical procedures, necessitating improved imaging techniques.
- Current DSA methods require precise patient immobility, posing challenges in real-world clinical scenarios.
Purpose of the Study:
- To develop and validate a deep learning (DL) model capable of generating DSA-like cerebral angiograms directly from dynamic angiograms.
- To quantitatively and visually assess the clinical usefulness of DL-generated angiograms, particularly in the presence of artifacts.
- To provide a robust AI solution for artifact-free cerebral angiography.
Main Methods:
- A retrospective study utilized dynamic and DSA image pairs from January to April 2019.
- A DL model was trained on a large dataset, validated, and then tested quantitatively (PSNR, SSIM) and visually on separate datasets, including those with misregistration artifacts.
- Radiologists evaluated the visual quality of DL-generated angiograms using a numerical rating scale.
Main Results:
- The DL model achieved high quantitative scores, with mean PSNR of 40.2 dB and SSIM of 0.97, indicating strong similarity to gold-standard DSA images.
- Visual evaluation demonstrated that DL-generated angiograms were rated as similar to or better than original DSA images across all tested sequences.
- The model successfully produced artifact-free cerebral angiograms, even from images with significant misregistration.
Conclusions:
- The developed deep learning model effectively generates clinically useful cerebral angiograms directly from dynamic angiograms, free from significant artifacts.
- This AI-driven approach offers a promising solution for improving the quality and reliability of cerebral angiography in clinical practice.
- The model's ability to handle motion artifacts enhances its potential for broader application in interventional radiology.
