Peer-Reviewed Research
Deep Learning-Based High-Resolution Magnetic Resonance Angiography (MRA) Generation Model for 4D Time-Resolved Angiography with Interleaved Stochastic Trajectories (TWIST) MRA in Fast Stroke Imaging
Our summary
This study investigated the improvement of image quality and diagnostic performance utilizing SwiftMR for time-resolved, contrast-enhanced magnetic resonance angiography (CE-MRA) in fast stroke imaging. Specifically, relatively low signal-to-noise ratio and spatial resolution of time-resolved angiography with interleaved stochastic trajectories (4D-TWIST-MRA) was explored. CE-MRA images from 520 patients were processed with SwiftMR, and were evaluated by two board-certified radiologists for overall image quality, aneurysm size measurement, detection accuracy, and diagnostic confidence in patients suspected of intracranial large vessel occlusion (LVO) were assessed. Results indicated that SwiftMR-processed images exhibited superior signal-to-noise ratio (SNR), overall image quality, sharpness, and vascular conspicuity compared to conventional images. In terms of aneurysm assessment, two additional aneurysms were detected, and the average aneurysm size was closer to that of the reference images compared to those obtained from conventional images. The level of diagnostic confidence among radiologists increased, accompanied by reduced decision-making time with SwiftMR-processed images. This study demonstrates that SwiftMR can enhance both image quality and diagnostic performances even when applied retrospectively on conventional clinical images.
Why this matters
Time-resolved CE-MRA with its short acquisition time is crucial for patients with suspected hyperacute ischemic stroke. However, this imaging method is limited by lower SNR due to high temporal resolution compared to the time-of-flight (TOF) MRA. This research highlights the significant role of SwiftMR in a real-world clinical setting, demonstrating its ability to improve both image quality and clinical performance.
For a thorough examination of the methodology, results, and potential impacts, you can read the full research article. It offers an in-depth analysis of how deep learning-based models can enhance 4D time-resolved MRA, improving the speed and accuracy of stroke imaging.
Kim BK, You S-H, Kim B, Shin JH. Deep Learning-Based High-Resolution Magnetic Resonance Angiography (MRA) Generation Model for 4D Time-Resolved Angiography with Interleaved Stochastic Trajectories (TWIST) MRA in Fast Stroke Imaging. Diagnostics. 2024; 14(11):1199. https://doi.org/10.3390/diagnostics14111199
