Diagnostic value of a second-generation super-resolution deep learning-based reconstruction combined with a metal
Taku Takaishi1, Koichiro Yasaka2, Kazuyoshi Miyamoto1
1Department of Radiology, The University of Tokyo Hospital, 7-3-1 Hongo, Bunkyo-Ku, Tokyo, 113-8655, Japan.
Skeletal Radiology
|November 15, 2025
Summary
Combining deep learning-based reconstruction (DLR2) with metal artifact reduction (MAR) significantly improves CT image quality in patients with hip implants. This enhanced visualization aids in better assessment of pelvic structures and overall diagnostic usability.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Healthcare
- Biomedical Engineering
Background:
- Metal implants in CT scans cause artifacts, degrading image quality and hindering diagnosis.
- Deep learning-based reconstruction (DLR) algorithms offer potential for artifact reduction.
- Evaluating advanced reconstruction techniques is crucial for improving diagnostic accuracy in patients with orthopedic implants.
Purpose of the Study:
- To assess the effectiveness of combining a second-generation deep learning-based reconstruction (DLR2) with metal artifact reduction (MAR) for CT images in patients with hip implants.
- To compare DLR2+MAR with other reconstruction methods, including first-generation DLR (DLR1) and conventional hybrid iterative reconstruction (HIR), both with MAR.
- To evaluate quantitative metrics and qualitative reader assessments of image quality and artifact reduction.
Main Methods:
- Retrospective analysis of CT images from 40 patients with metal hip implants.
- Image reconstruction using DLR2+MAR, DLR1+MAR, HIR+MAR, and DLR2-only.
- Quantitative analysis of signal-to-noise ratio (SNR) in regions of interest and artifact indices (AImuscle, AIfat).
- Qualitative assessment by three radiologists rating pelvic structure visualization, artifact reduction, and diagnostic usability.
Main Results:
- DLR2+MAR demonstrated significantly lower standard deviations (SDs) across all regions of interest compared to other MAR-combined methods and DLR2-only (p < 0.01-0.02).
- DLR2+MAR yielded the lowest artifact indices for muscle and fat, showing comparable results to DLR1+MAR.
- Radiologists rated DLR2+MAR significantly higher for visualizing the femoral artery and rectum compared to other MAR methods (p < 0.01-0.04).
- Substantial inter-rater agreement (0.67–0.77) was observed for qualitative assessments.
Conclusions:
- The combination of DLR2 and MAR significantly enhances CT image quality in patients with metal hip implants.
- This advanced reconstruction technique improves the visualization of critical pelvic structures.
- DLR2+MAR offers superior diagnostic performance compared to conventional and earlier-generation reconstruction methods.
