Multivendor AI for automated liver MRI prescription achieves high accuracy across scanners

An AI model for automated liver MRI prescription achieved median axial IoU3D of 0.963 (IQR 0.946–0.976) in a multicenter multivendor test set, with margins <4 mm to match manual coverage in 90% of cases. Performance differences across vendors, field strengths, and patient factor…

• Coronal and sagittal prescription IoU3D also exceeded 0.940 (median).
• Performance differences across vendors, field strengths, localizers, and patient factors were minimal (rank-biserial r ≤ 0.16, epsilon-squared ≤ 0.01).

Evidence grade: 2

Automated summary

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