Image Quality and Diagnostic Performance of Accelerated 2D Hip MRI with Deep Learning Reconstruction Based on a Deep...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Standard hip MRI sequences (TSE_S) have long acquisition times, potentially impacting patient comfort and throughput.
• Accelerated MRI techniques often compromise image quality, posing a challenge for routine clinical use.

Purpose of the Study:

• To compare standard 2D turbo spin echo (TSE_S) sequences with accelerated 2D TSE sequences using deep learning (DL) reconstruction (TSE_DL) for hip MRI.
• To evaluate the feasibility, image quality, and diagnostic performance of TSE_DL at 1.5 T and 3 T.

Main Methods:

• Prospective, monocentric study involving 14 patients undergoing hip MRI.
• Comparison of standard TSE_S and accelerated TSE_DL sequences in the same patients.
• Radiologist assessment of image quality, artifacts, noise, sharpness, and diagnostic confidence using a Likert scale.

Main Results:

• TSE_DL demonstrated significantly superior image quality, reduced noise, and improved edge sharpness compared to TSE_S (p ≤ 0.020).
• No significant differences were observed in artifacts, diagnostic confidence, or anatomical structure delineation (p > 0.05).
• Acquisition time reductions of 52% at 3 T and 70% at 1.5 T were achieved with TSE_DL.

Conclusions:

• Deep learning reconstruction (TSE_DL) is clinically feasible for hip MRI.
• TSE_DL provides excellent image quality and significantly reduces scan time.
• The diagnostic performance of TSE_DL is equivalent to standard TSE_S sequences.