Related Experiment Video
Updated: Jun 29, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.7K
DP-GAN+B: A lightweight generative adversarial network based on depthwise separable convolutions for generating CT
Xinlong Xing1, Xiaosen Li2, Chaoyi Wei3
1Postgraduate Training Base Alliance of Wenzhou Medical University, Wenzhou, Zhejiang, 325000, China; Wenzhou Institute, University of Chinese Academy of Sciences, Wenzhou, Zhejiang, 325000, China.
Computers in Biology and Medicine
|April 6, 2024
Summary
This study introduces DP-GAN+B, a novel AI method that converts 2D X-rays into 3D CT scans. This approach reduces radiation exposure and costs while improving diagnostic imaging quality.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- X-rays offer low radiation and cost-efficiency but struggle with visualizing overlapping organs.
- Computed Tomography (CT) provides 3D views but involves higher radiation doses and costs.
- Reconstructing 3D CT images from 2D X-rays holds significant clinical and practical value.
Purpose of the Study:
- To introduce DP-GAN+B, a novel method for reconstructing 3D lung CT volumes from 2D frontal and lateral X-ray images.
- To enhance diagnostic imaging by overcoming the limitations of traditional X-ray and CT scans.
- To reduce radiation exposure and healthcare costs associated with medical imaging.
Main Methods:
- Utilized a Generative Adversarial Network (GAN) architecture, DP-GAN+B.
- Employed depthwise separable convolutions instead of traditional convolutions for network efficiency.
- Introduced innovative vector and fusion loss functions to improve reconstruction performance.
Main Results:
- DP-GAN+B significantly reduced generator network parameters by 21.104 M and discriminator network parameters by 10.82 M (44.17% total reduction).
- Experimental results demonstrated the effective generation of clinically relevant, high-quality 3D CT images from 2D X-ray data.
- The method shows promise in enhancing diagnostic capabilities while mitigating cost and radiation concerns.
Conclusions:
- DP-GAN+B presents a pioneering approach for 2D X-ray to 3D CT image reconstruction.
- The developed method offers a cost-effective and lower-radiation alternative for detailed anatomical visualization.
- This technique has the potential to significantly benefit both patients and healthcare providers by improving diagnostic accuracy and accessibility.
