Related Experiment Video
Updated: Oct 24, 2025
06:33
Construction of a Preclinical Multimodality Phantom Using Tissue-mimicking Materials for Quality Assurance in Tumor Size Measurement
Published on: July 29, 2013
11.5K
Dose reduction potential of vendor-agnostic deep learning model in comparison with deep learning-based image
Hyunsu Choi1, Won Chang2, Jong Hyo Kim3,4
1Department of Radiology, Seoul National University Bundang Hospital, 82, Gumi-ro-173-beon-gil, Bundang-gu, Seongnam-si, Gyeonggi-do, 13620, Republic of Korea.
European Radiology
|August 14, 2021
Summary
A vendor-agnostic deep learning model (DLM) offers comparable dose reduction potential to vendor-specific algorithms, especially at high strengths. This advanced DLM shows superior performance in computed tomography imaging across various radiation doses.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Radiology
- Image Reconstruction
Background:
- Deep learning models (DLMs) are increasingly used to improve image quality in computed tomography (CT).
- Assessing the dose reduction potential (DRP) of these models is crucial for optimizing radiation exposure in patients.
- Vendor-specific deep learning-based image reconstruction (DLR) algorithms exist alongside vendor-agnostic DLMs.
Purpose of the Study:
- To compare the DRP of a vendor-agnostic DLM (ClariCT.AI) against a vendor-specific DLR (TrueFidelity™).
- To evaluate the performance of these algorithms across different radiation dose levels and target sizes in CT imaging.
Main Methods:
- CT images of a phantom were acquired at six dose levels and reconstructed using filtered back projection (FBP), DLM denoising, and three DLR strengths.
- Detectability index (d') was measured for various target sizes, contrast levels, and phantom diameters.
- DRP was calculated as the dose reduction yielding equivalent d' to FBP at full dose.
Main Results:
- The vendor-agnostic DLM achieved a DRP of 86%, comparable to high-strength DLR (87%).
- DLM outperformed DLR at medium (86% vs. 76%) and low strengths (86% vs. 60%).
- For small targets (1-5 mm), high-strength DLR showed slightly higher DRPs (88-89%) than DLM (84-87%).
Conclusions:
- The vendor-agnostic DLM provides significant dose reduction potential, comparable to high-strength vendor-specific DLR.
- DLM demonstrates superior performance over lower-strength DLR algorithms.
- Both DLM and DLR show promise for reducing radiation dose in CT while maintaining diagnostic image quality.
