Related Experiment Video
Updated: Jul 11, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
APNet: Adaptive projection network for medical image denoising
Qiyi Song1, Xiang Li2, Mingbao Zhang2
1Department of Endodontics and Periodontics, College of Stomatology, Dalian Medical University, Dalian, China.
Journal of X-Ray Science and Technology
|November 6, 2023
Summary
Adaptive Projection Network (APNet) effectively reduces noise in low-dose medical images. This advanced deep learning method enhances image quality for better disease diagnosis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Low-dose radiographic images suffer from noise, degrading feature quality and potentially impacting disease diagnosis.
- Image noise is a significant challenge in clinical medicine, affecting diagnostic accuracy.
Purpose of the Study:
- To introduce the Adaptive Projection Network (APNet) for effective noise reduction in low-dose medical images.
- To improve the quality of medical images obtained with reduced radiation exposure.
Main Methods:
- APNet utilizes a U-shaped network architecture for multi-scale data capture and end-to-end denoising.
- Dual attention residual blocks and a non-local attention module are integrated for adaptive feature calibration and noise-texture separation.
Main Results:
- APNet demonstrated superior performance over existing methods on lung CT images in both quantitative metrics and visual quality.
- Experiments on dental CT images confirmed the network's generalization capabilities for medical image denoising.
Conclusions:
- The proposed APNet is an effective deep learning solution for reducing noise in low-dose radiographic images.
- APNet successfully preserves essential image details crucial for accurate medical diagnosis.
