Related Experiment Video
Updated: Oct 23, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Deep Learning Algorithm for Simultaneous Noise Reduction and Edge Sharpening in Low-Dose CT Images: A Pilot Study
Hyunjung Yeoh1, Sung Hwan Hong2, Chulkyun Ahn3
1Department of Radiology, Seoul National University College of Medicine, Seoul, Korea.
Korean Journal of Radiology
|August 25, 2021
Summary
A deep learning algorithm effectively reduced noise and sharpened edges in low-dose lumbar spine CT scans. This advanced technique improves image quality while maintaining anatomical detail for better diagnostic accuracy.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence
Background:
- Low-dose CT scans are crucial for reducing radiation exposure in lumbar spine imaging.
- However, low-dose CT often results in increased image noise and reduced sharpness, potentially impacting diagnostic accuracy.
- Deep learning (DL) offers potential solutions for image reconstruction challenges.
Purpose of the Study:
- To evaluate the efficacy of a deep learning algorithm in simultaneously reducing noise and sharpening edges in low-dose lumbar spine CT.
- To compare the image quality of DL-reconstructed low-dose CT with standard-dose CT and non-denoised low-dose CT.
Main Methods:
- A retrospective study included 52 patients undergoing CT-guided lumbar bone biopsy.
- Low-dose (50-mAs) and standard-dose (100-mAs) CT images were reconstructed.
- A vendor-agnostic DL model (ClariCT.AI™) was used for denoising the 50-mAs images.
- Image noise, signal-to-noise ratio (SNR), and edge rise distance (ERD) were measured and compared.
- Radiologists assessed the visualization of anatomical structures.
Main Results:
- Denoised 50-mAs images showed significantly lower noise (36.38 ± 7.03 HU) compared to 50-mAs (93.33 ± 25.36 HU) and 100-mAs (63.33 ± 16.09 HU) images (p < 0.001).
- The denoised 50-mAs images achieved higher SNRs (1.46 ± 0.54) than 100-mAs (0.99 ± 0.34) and 50-mAs (0.58 ± 0.18) images (p < 0.001).
- DL-enhanced images demonstrated improved edge sharpness (ERD) at the vertebral body and psoas, and significantly better visualization of anatomical structures (p < 0.001).
Conclusions:
- Deep learning-based reconstruction can simultaneously reduce noise and enhance image quality, preserving edge sharpness in low-dose lumbar spine CT.
- This technique holds promise for improving diagnostic accuracy in low-dose CT imaging.
- Further research is needed to explore greater radiation dose reduction and clinical applications.
