Related Experiment Video
Updated: Dec 6, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.2K
Virtual digital subtraction angiography using multizone patch-based U-Net.
Ryusei Kimura1, Atsushi Teramoto2, Tomoyuki Ohno3
1Graduate School of Health Sciences, Fujita Health University, 1-98 Dengakugakubo, Kutsukake-cho, Toyoake-city, Aichi, 470-1192, Japan.
Physical and Engineering Sciences in Medicine
|October 7, 2020
Summary
A novel Virtual Digital Subtraction Angiography (DSA) method uses U-Net deep learning to create clear blood vessel images from single X-ray views. This technique effectively eliminates motion artifacts, improving diagnostic accuracy.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Radiology
Background:
- Digital Subtraction Angiography (DSA) is crucial for visualizing blood vessels.
- Patient motion during DSA imaging introduces artifacts, compromising image quality.
- Existing DSA methods require a mask image, adding complexity.
Purpose of the Study:
- To develop a novel Virtual DSA method for artifact-free blood vessel visualization.
- To generate DSA images from a single live X-ray image without a mask.
- To leverage deep learning for enhanced accuracy in blood vessel extraction.
Main Methods:
- Developed a Virtual DSA technique using the U-Net deep learning architecture.
- Implemented a patch-based processing approach for U-Net to enhance localized blood vessel extraction.
- Utilized distinct networks for different image zones to optimize local information processing.
Main Results:
- The Virtual DSA method successfully generated artifact-free blood vessel images from single live X-ray inputs.
- Accurate blood vessel extraction was confirmed in evaluations using head imaging.
- Quantitative metrics included NMSE of 8.58%, PSNR of 33.86 dB, and SSIM of 0.829.
Conclusions:
- The proposed Virtual DSA method effectively visualizes blood vessels from a single live image.
- This approach eliminates motion artifacts common in traditional DSA.
- Virtual DSA shows promise for improved diagnostic imaging in clinical practice.
