Impact of Deep Learning Image Reconstruction Methods on MRI Throughput.

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Deep learning reconstruction (DLR) algorithms are emerging technologies in MRI.
• Evaluating their real-world clinical efficiency is crucial for adoption.

Purpose of the Study:

• To assess the impact of two distinct DLR algorithms on MRI examination efficiency.
• To compare DICOM-based and k-space-based DLR methods in a multicenter setting.

Main Methods:

• Retrospective analysis of 7346 MRI examinations across 10 scanners.
• Comparison of scan and room times before and after DLR implementation.
• Utilized Wilcoxon test to compare DLR and non-DLR groups.

Main Results:

• DICOM-based DLR reduced scan times by up to 53% and room times by up to 41%.
• k-space-based DLR reduced scan times by up to 27% but did not significantly impact room times.
• Efficiency gains varied by examination type.

Conclusions:

• DLR methods can decrease MRI scan and room times in clinical settings.
• The effectiveness of DLR is heterogeneous and depends on the specific algorithm and examination type.
• Careful evaluation of institutional case mix is recommended before DLR integration.