Herrmann
A fast 5-minute MRI protocol of the shoulder with deep learning (DL) reconstruction is feasible in clinical routine and allows a reduction in scan time of more than 50% compared to the standard protocol, according to research presented Sunday.
MRI of the shoulder has become the most commonly used imaging modality in patients with shoulder pain. It provides a comprehensive noninvasive diagnosis of pathologies such as rotator cuff or biceps tendon tears and glenoid labral abnormalities.
"MRI is often preferred to arthroscopy and offers much more information than a simple shoulder X-ray," said presenter and study lead author Judith Herrmann, MD, from the Department of Diagnostic and Interventional Radiology at the Eberhard Karls University of Tübingen in Tübingen, Germany.
The exam can be time-consuming and may cause patient discomfort. A speedier, more efficient exam of similar quality would be of great benefit to clinicians and patients. Dr. Herrmann and colleagues studied a 5-minute shoulder MRI protocol that exploits the power of DL for faster image reconstruction. The 5-minute protocol consists of accelerated 2D turbo spin echo (TSE) sequences with a DL reconstruction at 1.5T and 3T.
While prior studies have shown good results for DL in this setting, they have been primarily preclinical or retrospective.
For the new study, researchers prospectively compared the image quality of the 5-minute protocol to that of a standard 2D TSE protocol in 30 patients who underwent shoulder MRI at 1.5T and 3T. Each patient had two MRI examinations: one with a standard TSE (TSES) protocol and the other with a fast TSE with DL reconstruction (TSEDL) and an undersampled k space. Two musculoskeletal radiologists independently assessed image quality and visualization of anatomical structures using a Likert scale ranging from 1 to 5, with 5 representing the best quality. The radiologists also assessed diagnostic performance concerning pathologic shoulder lesions.
Superior Results From Fast Protocol
The radiologists judged the overall quality of the images from the TSEDL protocol to be superior to those from the standard protocol. They found noise and edge sharpness to be significantly superior in TSEDL versus TSES. No difference was found in qualitative diagnostic confidence and clarity of anatomical structures and quantitative diagnostic performance for pathologic shoulder lesions when comparing the two sequences.
"The use of the deep learning reconstruction improves the overall image quality by reducing noise and improving edge sharpness, providing interchangeable results concerning the diagnostic performance, and allowing a reduction in scan time of more than 50% compared to the standard shoulder imaging," Dr. Herrmann said.
If larger studies involving multiple centers and different equipment vendors support the results, ultrafast protocols in shoulder MRI could soon become an important option in daily clinical routine, according to Dr. Herrmann.
"This technique increases patient comfort and throughput," she said.
Access the presentation, "Implementation and Evaluation of a Fast 5-Minute Shoulder MRI Protocol With Accelerated TSE-Sequences and Deep Learning Image Reconstruction for the Assessment of Shoulder Pain at 1.5 and 3 T," (S4-SSMK02-1) on demand at Meeting.RSNA.org.
