Related Experiment Video
Updated: Jul 16, 2026
10:14
3D Scanning Technology Bridging Microcircuits and Macroscale Brain Images in 3D Novel Embedding Overlapping Protocol
Published on: May 12, 2019
Deep learning-based reconstruction for three-dimensional volumetric brain MRI: a qualitative and quantitative
Yeseul Kang1, Sang-Young Kim2, Jun Hwee Kim1
1Department of Radiology, Yongin Severance Hospital, Yonsei University College of Medicine, 363 Dongbaekjukjeon-daero, Giheung-gu, Yongin-si, Gyeonggi-do, 16995, Republic of Korea.
BMC Medical Imaging
|March 28, 2025
Summary
Deep learning reconstruction (DLR) significantly reduces brain MRI scan times by over 50% without compromising image quality or volumetric accuracy. This efficient technique, using Adaptive-Compressed sensing (CS)-Network, proves reliable for clinical applications.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence in Medicine
Background:
- Evaluating deep learning reconstruction (DLR) performance for brain MRI.
- Validating Adaptive-Compressed sensing (CS)-Network in a clinical setting.
Purpose of the Study:
- Assess the efficacy of DLR with Adaptive-CS for accelerated brain MRI.
- Compare image quality and volumetric measurements between standard and accelerated sequences.
Main Methods:
- Prospective enrollment of healthy volunteers and patients.
- Acquisition of 3D brain MRI using varying CS factors with and without DLR.
- Quantitative volumetric analysis and qualitative radiologist assessment of image quality metrics.
Main Results:
- Scan time reduction of 65.4% (CS factor 2) and 33.5% (CS factor 4) achieved.
- No significant differences in subregion volumes across sequences.
- DLR-CS4 demonstrated comparable image quality and volumetric accuracy to CS2, with substantial time savings.
Conclusions:
- DLR enables significant brain MRI scan time reduction (at least 50%) without sacrificing image quality.
- Volumetric quantification accuracy is maintained, supporting DLR's reliability and efficiency.
- DLR is a promising tool for improving clinical workflow in brain MRI.
