Related Experiment Video
Updated: Nov 1, 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
Image quality improvement with deep learning-based reconstruction on abdominal ultrahigh-resolution CT: A phantom
Takashi Shirasaka1, Tsukasa Kojima1, Yoshinori Funama2
1Division of Radiology, Department of Medical Technology, Kyushu University Hospital, Fukuoka, Japan.
Journal of Applied Clinical Medical Physics
|June 23, 2021
Summary
Deep learning-based reconstruction (DLR) reduces noise in ultrahigh-resolution CT (U-HRCT) abdominal imaging. DLR maintains consistent image quality across decreasing radiation doses compared to other methods.
Area of Science:
- Medical Imaging
- Radiology
- Computational Imaging
Background:
- Ultrahigh-resolution CT (U-HRCT) offers enhanced detail but often suffers from increased image noise.
- Deep learning-based reconstruction (DLR) algorithms aim to improve image quality by reducing noise without compromising spatial resolution.
- Assessing DLR performance at varying radiation doses is crucial for optimizing abdominal CT protocols.
Purpose of the Study:
- To evaluate the impact of a novel deep learning-based reconstruction (DLR) algorithm on image quality in abdominal ultrahigh-resolution CT (U-HRCT).
- To compare DLR performance against traditional reconstruction methods (MBIR, FBP, HIR) at different radiation dose levels.
- To determine the consistency of DLR in maintaining image quality as radiation dose is reduced.
Main Methods:
- Abdominal phantom models were scanned using U-HRCT at 100%, 50%, and 25% of the standard radiation dose.
- Images were reconstructed using DLR, model-based iterative reconstruction (MBIR), filtered back projection (FBP), and hybrid iterative reconstruction (HIR).
- Quantitative analysis included measurement of image noise (standard deviation of CT number) and spatial resolution (modulation transfer function).
Main Results:
- DLR demonstrated lower image noise compared to FBP and HIR across all radiation doses.
- While DLR had higher noise than MBIR at 100% and 50% doses, it showed lower noise at the 25% dose.
- Spatial resolution (MTF at 10% contrast) remained largely consistent for DLR, even at reduced radiation doses, with values around 1.0 cycles/mm.
Conclusions:
- Deep learning-based reconstruction (DLR) offers a robust solution for noise reduction in abdominal U-HRCT.
- DLR exhibits superior consistency in maintaining image quality at lower radiation doses compared to MBIR, FBP, and HIR.
- The findings support the potential of DLR for dose reduction strategies in abdominal CT imaging.
