Development and evaluation of automated organ segmentation models for abdominal ultrasound in injured children

Deep learning models segment six abdominal organs on pediatric FAST with mean Dice-Sørensen Coefficient of 0.79 (95% CI: 0.75-0.84), highest for kidney/bladder at 0.91.

• Why it matters: Manual pediatric FAST organ labeling is time-consuming and variable, limiting scalable dataset creation and automated image analysis quality checks.
• What they did: In 200 injured children at a Level I pediatric trauma center, physicians manually segmented six organs; 1,807 image-mask pairs trained U-Net variants with an 80:20 clip-split and evaluated Dice-Sørensen Coefficient.
• The result: EVIT U-Net led for kidney, liver, and spleen and averaged 54 milliseconds per image; in previously unlabeled clips, experts rated localization as acceptable in 73%-95% across organs.

AI summary of the abstract · may contain errors · not medical advice · read the paper.