Enhanced Spinal Cord Lesion Detection in MS Using White-Matter-Nulled 3D MPRAGE with Deep Learning Reconstruction.

Area of Science:

• Neurology
• Radiology
• Medical Imaging

Background:

• Spinal cord (SC) lesions are crucial for diagnosing and monitoring multiple sclerosis (MS).
• Conventional MRI sequences (2D T2-weighted FSE, STIR) have limitations in detecting SC lesions, especially in the cervical and thoracic regions.
• Accurate SC lesion detection is vital for MS diagnosis and progression assessment.

Purpose of the Study:

• To evaluate the diagnostic performance of a novel 3D white-matter-nulled (WMn) MPRAGE sequence for SC lesion detection.
• To compare the sensitivity and accuracy of the 3D WMn sequence against conventional MRI techniques.
• To assess the impact of deep learning denoising on the 3D WMn sequence's performance.

Main Methods:

• Prospective evaluation of 38 patients with MS or clinically isolated syndrome using 3T SC MRI.
• Acquisition of 2D T2-weighted FSE, 2D STIR, 3D MPRAGE, and 3D WMn sequences.
• Application of a deep learning denoising method to the 3D WMn sequence.
• Independent assessment of lesion count, detection confidence, and image quality by four blinded neuroradiologists.
• Computation of contrast-to-noise ratio (CNR) for detected lesions.

Main Results:

• The 3D WMn sequence detected significantly more lesions in the cervicothoracic spine (+62% vs. 2D T2 FSE, +47% vs. STIR, +50% vs. 3D MPRAGE).
• In the thoracolumbar spine, 3D WMn also detected more lesions (+53% vs. 2D T2 FSE).
• The 3D WMn sequence showed higher CNR, improved lesion conspicuity, fewer artifacts, and increased rater confidence, particularly among experienced readers.

Conclusions:

• The 3D WMn sequence, enhanced with deep learning denoising, significantly improves SC lesion detection in MS.
• This novel sequence outperforms conventional MRI techniques in sensitivity and diagnostic confidence.
• 3D WMn MPRAGE represents a promising advancement for SC imaging in multiple sclerosis management.