Semi-supervised Learning of Fetal Anatomy from Ultrasound

TL;DR

This paper investigates semi-supervised learning for fetal anatomy classification in ultrasound images, highlighting challenges with background classes and the varying benefits for different anatomy classes.

Contribution

It demonstrates the limitations of semi-supervised learning in complex ultrasound data and emphasizes the importance of class distinction for effective learning.

Findings

Semi-supervised learning benefits distinct classes more.

Including challenging background classes can hinder performance.

Performance varies across different fetal anatomy classes.

Abstract

Semi-supervised learning methods have achieved excellent performance on standard benchmark datasets using very few labelled images. Anatomy classification in fetal 2D ultrasound is an ideal problem setting to test whether these results translate to non-ideal data. Our results indicate that inclusion of a challenging background class can be detrimental and that semi-supervised learning mostly benefits classes that are already distinct, sometimes at the expense of more similar classes.