Self-supervised learning strategies evaluated for lung ultrasound imaging
Which Pretext Task Transfers? Self-Supervised Pretraining Objectives for Lung Ultrasound
Computer Vision and Pattern Recognition
Summary
Getting labeled lung ultrasound images is hard, so the authors studied ways to teach computers to learn from videos without labels first, then use a few labeled images to recognize lung features. They tested three popular methods on the same datasets and found that which method works best depends on the specific set of lung ultrasound images being used. This suggests that success on one dataset doesn't guarantee the method will do well on others. The authors plan further analysis to understand these differences better and shared their code openly.
What this means in practice
- •For medical image analysis teams: Choose and evaluate pretraining methods for lung ultrasound AI models to improve performance across diverse patient datasets.
- •For ultrasound device software developers: Incorporate self-supervised pretraining strategies for lung ultrasound video processing to reduce dependence on large labeled datasets and adapt to new data.
Authors
Moein Heidari, Junbo Rao, Jai Choraria, Wenjin Chen, David J. Foran, Ilker Hacihaliloglu
Abstract
Self-supervised learning (SSL) can reduce the need for labelled medical images, but the choice of pretext objective remains unclear for lung ultrasound (LUS). Contrastive learning, masked reconstruction, and joint-embedding predictive architectures (JEPA) differ in the space in which their targets are defined, yet existing ultrasound studies compare them under different corpora, backbones, and evaluation protocols. We compare these three objective families using the same encoder backbone, pretraining corpus, optimisation schedule, and frozen-evaluation protocol. Encoders are pretrained on COVID-BLUeS LUS videos and evaluated with linear, $k$NN, and attentive probes at 5\%, 10\%, 50\%, and 100\% label budgets. Evaluation is performed on POCUS using patient-level five-fold cross-validation and on the independently acquired Mendeley-Uganda dataset, which is excluded from both pretraining and probe fitting. At the full label budget under linear probing, VideoMAE and V-JEPA achieve $66.5 \pm 13.1$ and $65.4 \pm 11.7$ balanced accuracy on POCUS, while MoCo achieves $42.1 \pm 1.2$. On Mendeley-Uganda, the ranking reverses: MoCo performs best at $62.7 \pm 1.0$, followed by VideoMAE at $53.8 \pm 2.8$, while V-JEPA falls near chance at $35.1 \pm 4.9$. These results show that POCUS probe accuracy alone does not identify the objective that transfers best across datasets. We also outline planned representation-level analyses to examine this reversal. Code is publicly available at https://github.com/moeinheidari7829/LUSVideoSSL.