Deep learning-enhanced zero echo time silent brain magnetic resonance imaging in infants without sedation

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Pediatric Neurology

Background:

• Acoustic noise in infant brain MRI necessitates sedation, impacting patient care.
• Deep learning (DL) offers potential for improving image quality in silent MRI sequences like zero echo time (ZTE).

Purpose of the Study:

• To evaluate DL-based reconstruction's impact on ZTE infant brain MRI image quality.
• To compare DL-reconstructed ZTE MRI with conventional sedated MRI.

Main Methods:

• Retrospective analysis of 78 infants (≤16 months) undergoing brain MRI.
• Comparison of unsedated ZTE MRI with DL reconstruction (ZTE-DL) against sedated MPRAGE and unsedated ZTE.
• Image quality assessment by three radiologists using Likert scales and signal uniformity analysis.

Main Results:

• ZTE-DL demonstrated superior image quality, reduced noise, and better gray-white matter differentiation than unsedated ZTE.
• ZTE-DL image quality was comparable to sedated MPRAGE, with significantly less noise.
• Signal uniformity was highest in ZTE-DL for gray and white matter.

Conclusions:

• DL-based reconstruction significantly improves ZTE MRI quality in infants.
• This technique offers comparable image quality to sedated MRI, potentially eliminating the need for sedation.
• DL-enhanced ZTE MRI represents a promising advancement for pediatric neuroimaging.