Improvement of peripheral nerve visualization using a deep learning-based MR reconstruction algorithm.

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Medicine
• Neurology

Background:

• Standard-of-care (SOC) MR neurography is crucial for diagnosing peripheral neuropathies.
• Assessing nerve morphology, including outer epineurium and fascicular architecture, is vital for accurate diagnosis.
• Image artifacts can complicate interpretation in MR neurography.

Purpose of the Study:

• To evaluate a novel deep learning-based MR reconstruction method, DLRecon, for clinical peripheral nerve assessment.
• To compare the image quality and diagnostic utility of DLRecon-MRIs against standard-of-care (SOC) MRIs.

Main Methods:

• Sixty peripheral nerves in 29 patients with suspected neuropathy were prospectively evaluated.
• Standard-of-care MRIs and DLRecon-MRIs were acquired for blinded, random evaluation by two radiologists.
• Evaluated features included outer epineurium conspicuity, fascicular architecture, and artifacts (pulsation, ghosting, bulk motion).

Main Results:

• DLRecon-MRIs demonstrated superior conspicuity of the outer epineurium (OR=1.9, p=0.007) and fascicular architecture (OR=1.8, p<0.001) compared to SOC-MRIs.
• DLRecon-MRIs showed increased ghosting (OR=2.8, p=0.004) and pulsation artifacts (OR=1.6, p=0.004).
• Inter-reconstruction method agreement was substantial to almost-perfect (AC=0.73-1.00), with improved interrater agreement for DLRecon-MRI in outer epineurium conspicuity (AC=0.71).

Conclusions:

• DLRecon-MRI significantly enhances the conspicuity of critical nerve morphological features, aiding in the evaluation of nerve injuries.
• Despite increased artifacts, DLRecon-MRI does not negatively impact overall image interpretation.
• DLRecon shows promise as a valuable tool for improving clinical MR neurography.