Related Experiment Video
Updated: Jun 1, 2025
13:45
Hybrid µCT-FMT imaging and image analysis
Published on: June 4, 2015
13.1K
A comparative analysis of image harmonization techniques in mitigating differences in CT acquisition and
Anil Yadav1,2, Spencer Welland3, John M Hoffman3
1Department of Bioengineering, Samueli School of Engineering, University of California, Los Angeles, CA 90095, United States of America.
Physics in Medicine and Biology
|January 17, 2025
Summary
Image harmonization using convolutional neural networks (CNNs) and generative adversarial networks (GANs) improves CT scan consistency. GANs excel at reproducing quantitative features, crucial for machine learning models in clinical use.
Area of Science:
- Medical Imaging
- Radiology
- Artificial Intelligence in Medicine
Background:
- Variations in computed tomography (CT) parameters like radiation dose and reconstruction kernels can significantly impact image quality and feature reproducibility.
- This variability poses a challenge for the development of generalizable machine learning models and consistent diagnostic interpretation in lung imaging.
Purpose of the Study:
- To systematically characterize the effects of CT parameter variations on lung images, radiomic features, and deep features.
- To evaluate the effectiveness of various image harmonization techniques, including CNNs and GANs, in mitigating these variations.
Main Methods:
- Retrospective analysis of 100 low-dose chest CT scans reconstructed with varying radiation doses (100%, 25%, 10%) and kernels (smooth, medium, sharp).
- Training of image processing, CNN-based, and GAN-based methods to harmonize images to a reference condition (100% dose, medium kernel).
- Evaluation of image similarity using PSNR, SSIM, LPIPS, and feature reproducibility using CCC.
Main Results:
- CNNs demonstrated superior image similarity metrics, significantly improving PSNR, SSIM, and LPIPS for challenging conditions like Sharp/10%.
- Texture-based radiomic features showed higher variability (CCC of 0.500) compared to intensity-based features (CCC of 0.972).
- GANs achieved the highest concordance correlation coefficients (CCC) for both radiomic (0.969) and deep features (0.841), indicating excellent reproducibility.
Conclusions:
- CNNs are suitable for applications requiring visual image interpretation, while GANs are more effective for generating reproducible quantitative features for machine learning.
- Effective harmonization is critical for optimizing diagnostic accuracy and developing robust, generalizable AI models for clinical deployment in medical imaging.
Related Concept Videos
01:10Computed Tomography
4.2K
Tomography refers to imaging by sections. Computed tomography (CT) is a non-invasive imaging technique that uses computers to analyze several cross-sectional X-rays to reveal minute details about structures in the body.
The technique was invented in the 1970s and is based on the principle that as X-rays pass through the body, they are absorbed or reflected at different levels. In the technique, a patient lies on a motorized platform while a computerized axial tomography (CAT) scanner rotates...
The technique was invented in the 1970s and is based on the principle that as X-rays pass through the body, they are absorbed or reflected at different levels. In the technique, a patient lies on a motorized platform while a computerized axial tomography (CAT) scanner rotates...
4.2K
01:14Imaging Studies I: CT and MRI
187
Introduction: MRI and CT scans are crucial advancements in medical imaging techniques, playing a vital role in diagnosing conditions related to the gastrointestinal (GI) system. Each scan serves distinct purposes, targets specific areas, and requires unique nursing duties.
Description of the Procedures
Computed Tomography (CT) scan:
Computed Tomography (CT) scans use X-ray technology to generate detailed images of bones, organs, and tissues. During the scan, the patient lies on a moving table...
Description of the Procedures
Computed Tomography (CT) scan:
Computed Tomography (CT) scans use X-ray technology to generate detailed images of bones, organs, and tissues. During the scan, the patient lies on a moving table...
187