Related Experiment Video
Updated: Sep 29, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Maskless 2-Dimensional Digital Subtraction Angiography Generation Model for Abdominal Vasculature using Deep
Hiroki Yonezawa1, Daiju Ueda1, Akira Yamamoto1
1Department of Diagnostic and Interventional Radiology, Graduate School of Medicine, Osaka City University, Osaka, Japan.
Journal of Vascular and Interventional Radiology : JVIR
|March 21, 2022
Summary
A new deep learning model creates artifact-free, synthetic subtraction angiograms from native abdominal images. This AI-driven approach improves vascular visualization and clinical usefulness compared to traditional methods.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Motion artifacts in 2D digital subtraction angiography (2D-DSA) can hinder accurate diagnosis.
- Developing artifact-free imaging techniques is crucial for improving diagnostic quality.
Purpose of the Study:
- To develop a deep learning (DL) model for generating synthetic, artifact-free 2D subtraction angiograms from native abdominal angiograms.
- To evaluate the performance of the DL model in reducing motion artifacts and improving image quality.
Main Methods:
- A retrospective study collected native angiograms and 2D-DSA images.
- An image-to-image translation model was trained using native angiograms to generate DL-based subtraction angiography (DLSA) images.
- Quantitative (PSNR, SSIM) and qualitative (radiologist visual assessment) evaluations were performed.
Main Results:
- DLSA images demonstrated high agreement with original 2D-DSA images (mean PSNR: 43.05 dB, mean SSIM: 0.98).
- Radiologists found DLSA images had fewer motion artifacts than 2D-DSA images.
- DLSA images were rated similarly or higher for vascular visualization and clinical usefulness.
Conclusions:
- The developed DL model successfully generates synthetic, motion-free subtraction images from abdominal angiograms.
- The DL model produces images with characteristics comparable to traditional 2D-DSA.
- This AI approach offers a promising solution for artifact reduction in abdominal angiography.
