Related Experiment Video
Updated: Jun 16, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.7K
Large-scale pretrained frame generative model enables real-time low-dose DSA imaging: An AI system development and
Huangxuan Zhao1, Ziyang Xu2, Lei Chen1
1Department of Radiology, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan 430022, China.
Med (New York, N.Y.)
|August 20, 2024
Summary
This study introduces GenDSA, an AI system that generates digital subtraction angiography (DSA) images using significantly less radiation. GenDSA produces high-quality, real-time DSA videos, reducing radiation exposure for doctors and patients.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Digital subtraction angiography (DSA) requires multiple scans, leading to substantial radiation exposure for medical professionals and patients.
- Current DSA procedures necessitate high radiation doses, posing health risks during interventional procedures.
Purpose of the Study:
- To develop GenDSA, a real-time, low-dose digital subtraction angiography (DSA) imaging system utilizing a large-scale pretrained multi-frame generative model.
- To significantly reduce radiation exposure in DSA procedures while maintaining image quality.
Main Methods:
- GenDSA was trained on approximately 3 million DSA images from 27,117 patients across 10 hospitals.
- The system generates 1- to 3-frame sequences following each real frame, validated using objective metrics (SSIM, PSNR) and subjective assessments by five interventional radiologists.
- Evaluations included Likert scale ratings, visual Turing tests, Kendall's W, and Fleiss' kappa for inter-rater reliability.
Main Results:
- GenDSA achieved high performance with PSNR = 36.83, SSIM = 0.911, and a generation time of 0.07 s/frame, surpassing existing algorithms.
- Generated videos were indistinguishable from real videos by expert radiologists, with comparable overall quality and lesion assessment scores.
- The system successfully reduced radiation dose by two-thirds while maintaining video consistency.
Conclusions:
- GenDSA offers significant clinical and translational value by substantially reducing radiation exposure during DSA-guided interventions.
- The developed AI system enhances patient and physician safety in interventional radiology.
- GenDSA represents a breakthrough in low-dose, real-time DSA imaging.
