Deep Learning Reconstruction Improves Neonatal Brain MRI Quality at 3T: A Pilot Study

Higher denoising levels with a deep learning reconstruction algorithm progressively improved qualitative and quantitative image quality in 3D T1-weighted neonatal brain MRI, though the pilot study's small sample size (n=15) limits generalizability.

• Radiologists rated image quality (SNR, artifacts, clarity) and conspicuity of early myelinating structures higher as denoising level increased.
• Objective apparent SNR and contrast-to-noise ratio also increased progressively, with significant differences in nearly all pairwise comparisons.
• Limitations: small sample size (n=15), pilot design, and no evaluation of diagnostic accuracy.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.