Deep Learning Model Suppresses Venous Signal on MR Lymphangiography, Cutting Read Time from 17 to 4 Minutes

A deep learning tool (LympClear) suppressed obscuring venous signal on MR lymphangiography in 1022 patients with lower extremity lymphedema, boosting lymphatic reflux visibility (8.3 vs 6.0 on a 10-point scale) and slashing reading time from 16.7 to 3.6 minutes per exam. Readers…

• Retrospective and prospective multicenter study across four sites; 1022 patients (6162 dynamic scans), mean age 49.3 years, 560 female.
• LympClear achieved a Dice score of 0.94 for venous segmentation; venous-suppressed images significantly improved image quality, venous signal suppression score, and lymphatic reflux visibility (all P < .001).
• Key limitation: The reference standard relied on consensus radiologist annotations without an independent histopathologic or invasive lymphographic gold standard, and prospective validation was limited to an internal test set.

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