Improved Resolution and Image Quality of Musculoskeletal Magnetic Resonance Imaging using Deep Learning-based...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Musculoskeletal (MSK) magnetic resonance imaging (MRI) is crucial for diagnosis.
• Improving resolution and image quality in MSK MRI can enhance diagnostic accuracy.
• Current reconstruction methods may have limitations in achieving optimal image quality.

Purpose of the Study:

• To prospectively evaluate a deep learning-based denoising reconstruction (DLR) technique.
• To assess the impact of DLR on resolution and image quality in MSK MRI.
• To compare DLR with conventional reconstruction filters.

Main Methods:

• Prospective evaluation of 137 contrast-weighted MSK MRI sequences in 40 patients.
• Comparison of routine reconstruction filter (REF) images with DLR and conventional filters (NL2, GA43, GA53).
• Blind review by MSK radiologists, quantitative SNR, CNR, and FWHM measurements, and statistical analysis using GEE.

Main Results:

• DLR improved resolution by 47.61% (0.39 mm² to 0.20 mm²; p < 0.001).
• Average scan time was reduced by 7.93% with DLR (165.58 s to 152.45 s; p < 0.001).
• DLR images received significantly higher scores for all image quality criteria and forced ranking (p < 0.001).

Conclusions:

• DLR significantly enhances resolution and image quality in MSK MRI.
• The DLR technique offers approximately two times finer resolution than standard-of-care images.
• DLR represents a promising advancement for MSK MRI.