DINOv3-based model reaches Dice 0.945 on adnexal mass segmentation in 7777 ultrasound frames and holds performance when trained on only 25% of the data.
Clinical significance of us artifacts.Radiographics, 37(5):1408–1423, 2017
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Adapting Foundation Models for Annotation-Efficient Adnexal Mass Segmentation in Cine Images
DINOv3-based model reaches Dice 0.945 on adnexal mass segmentation in 7777 ultrasound frames and holds performance when trained on only 25% of the data.