Research4d ago
Augmented Ultrasound Segmentation Boosts Accuracy
The augmentation added 3,000 synthetic variations, raising Dice scores by 12% on small structures. Researchers applied size‑scaling and texture remixing to existing scans, preserving anatomy while diversifying training data. The method avoids costly manual labeling and synthetic generation. Practitioners can integrate this technique to improve segmentation accuracy in routine ultrasound workflows. The technique builds on the Segment Anything framework.