Deep learning reconstruction for optimized bone assessment in zero echo time MR imaging of the knee

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Zero echo-time (ZTE) MRI offers potential for detailed bone assessment.
• Standard reconstruction methods may limit image quality and conspicuity of bone pathology.
• Deep learning-based reconstruction (DLRecon) is emerging as a tool to improve MRI quality.

Purpose of the Study:

• To evaluate the impact of DLRecon on bone assessment in 1.5 Tesla knee ZTE MRI.
• To compare bone depiction, artifact levels, and pathology conspicuity between DLRecon and standard reconstruction.
• To assess quantitative improvements in signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) with DLRecon.

Main Methods:

• Retrospective analysis of 48 knee MRI exams (1.5 Tesla) using isotropic ZTE sequences.
• Reconstruction of ZTE images using standard (non-DL) and prototype DLRecon methods.
• Qualitative grading by blinded readers for bone depiction, artifacts, and pathology conspicuity; quantitative SNR and CNR measurements.

Main Results:

• DLRecon significantly improved bone depiction quality compared to non-DL reconstruction (p < 0.05).
• Conspicuity of osseous pathologies and diagnostic confidence were significantly higher with DLRecon (p ≤ 0.03).
• DLRecon demonstrated significantly enhanced bone SNR and CNR (p < 0.001) with no significant increase in artifacts.

Conclusions:

• DLRecon significantly enhances bone depiction in knee ZTE MRI compared to non-DL methods.
• DLRecon improves the conspicuity of osseous findings and increases diagnostic certainty.
• DLRecon represents a valuable advancement for bone assessment in knee MRI.