Related Experiment Video
Updated: Aug 27, 2025
07:00
Diffusion Tensor Magnetic Resonance Imaging in Chronic Spinal Cord Compression
Published on: May 7, 2019
9.0K
Ultrafast cervcial spine MRI protocol using deep learning-based reconstruction: Diagnostic equivalence to a
Nobuo Kashiwagi1, Mio Sakai1, Akio Tsukabe2
1Department of Diagnostic and Interventional Radiology, Osaka International Cancer Institute, Japan.
European Journal of Radiology
|September 30, 2022
Summary
An ultrafast cervical spine magnetic resonance imaging (MRI) protocol significantly reduces scan time. This deep learning-based reconstruction (DLR) method achieves diagnostic results comparable to conventional MRI.
Area of Science:
- Radiology
- Medical Imaging
- Deep Learning
Background:
- Magnetic Resonance Imaging (MRI) is crucial for cervical spine evaluation.
- Limited imaging speed is a significant drawback of conventional MRI techniques.
- Ultrafast MRI protocols aim to improve efficiency without compromising diagnostic quality.
Purpose of the Study:
- To introduce and evaluate an ultrafast cervical spine MRI protocol using deep learning-based reconstruction (DLR).
- To compare the diagnostic performance of the ultrafast protocol against conventional MRI protocols.
- To assess the feasibility of significantly reducing MRI acquisition time for cervical spine imaging.
Main Methods:
- Fifty patients underwent cervical spine MRI using both conventional and ultrafast protocols (sagittal T1/T2-weighted, STIR; axial T2*-weighted).
- The ultrafast protocol utilized reduced phase matrix, oversampling rate, number of excitations, and compressed sensing.
- Deep learning-based reconstruction (DLR) was applied for noise reduction to compensate for accelerated acquisition.
- Neuroradiologists assessed degenerative changes (stenosis, endplate/disc degeneration, disc hernia) and other pathologies.
- Interchangeability was tested using 95% confidence intervals of the individual equivalence index; inter-protocol agreement was assessed using kappa statistics.
Main Results:
- The ultrafast MRI protocol achieved an acquisition time of 2 minutes 57 seconds, compared to 12 minutes 54 seconds for the conventional protocol.
- Except for endplate degeneration, the 95% confidence intervals for all assessed variables did not exceed 5%, indicating interchangeability.
- Kappa values ranged from 0.600 to 0.977, demonstrating substantial to almost perfect inter-reader agreement between the two protocols.
Conclusions:
- The proposed ultrafast MRI protocol provides diagnostic results that are nearly equivalent to conventional MRI protocols for the cervical spine.
- Deep learning-based reconstruction effectively compensates for signal-to-noise ratio reduction in accelerated imaging.
- This ultrafast approach offers a significant improvement in imaging efficiency for cervical spine MRI.
