Deep learning reconstruction improves appendix visualization on pediatric magnetic resonance imaging (MRI): a...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Medicine
• Pediatric Healthcare

Background:

• Thin-slice MRI offers potential for better appendix visualization in pediatric appendicitis.
• However, reduced slice thickness typically lowers signal-to-noise ratio (SNR) and degrades image quality.
• Deep learning reconstruction (DLR) can denoise thin-slice images, potentially overcoming these limitations.

Purpose of the Study:

• To assess if 2-mm T2-weighted MRI with DLR enhances appendix visualization and diagnostic performance compared to standard 4-mm MRI in pediatric appendicitis.
• To evaluate the impact of DLR on image quality and diagnostic confidence.

Main Methods:

• Prospective study of 82 pediatric patients with suspected appendicitis.
• Comparison of standard 4-mm MRI with thin-slice 2-mm MRI enhanced by DLR (2-mm+DLR).
• Radiologist assessment of appendix visualization, diagnostic confidence, image quality, and ancillary findings; objective SNR measurements.

Main Results:

• 2-mm+DLR images showed comparable SNR to standard 4-mm images.
• Improved appendix visualization with 2-mm+DLR (88.2%) versus 4-mm MRI (83.3%).
• Higher overall diagnostic image quality with 2-mm+DLR (median 5 vs 4), with similar diagnostic confidence.

Conclusions:

• DLR-enhanced 2-mm MRI improves appendix visualization and image quality in pediatric appendicitis.
• The technique does not reduce reader confidence or specificity.
• Thin-slice DLR MRI aids in depicting the appendix and other abdominal pathologies in children.