USCL: Pretraining Deep Ultrasound Image Diagnosis Model through Video Contrastive Representation Learning
This work provides a method to improve the accuracy of deep learning models for ultrasound image diagnosis, benefiting medical practitioners by enabling more robust and accurate diagnoses.
The paper addresses the performance bottleneck in deep neural networks for ultrasound image analysis caused by the domain gap between natural and medical images. They constructed a large ultrasound dataset (US-4) and proposed USCL, a semi-supervised contrastive learning method for pretraining, which achieved over 94% accuracy on the POCUS dataset, a 10% improvement over ImageNet pretraining.
Most deep neural networks (DNNs) based ultrasound (US) medical image analysis models use pretrained backbones (e.g., ImageNet) for better model generalization. However, the domain gap between natural and medical images causes an inevitable performance bottleneck. To alleviate this problem, an US dataset named US-4 is constructed for direct pretraining on the same domain. It contains over 23,000 images from four US video sub-datasets. To learn robust features from US-4, we propose an US semi-supervised contrastive learning method, named USCL, for pretraining. In order to avoid high similarities between negative pairs as well as mine abundant visual features from limited US videos, USCL adopts a sample pair generation method to enrich the feature involved in a single step of contrastive optimization. Extensive experiments on several downstream tasks show the superiority of USCL pretraining against ImageNet pretraining and other state-of-the-art (SOTA) pretraining approaches. In particular, USCL pretrained backbone achieves fine-tuning accuracy of over 94% on POCUS dataset, which is 10% higher than 84% of the ImageNet pretrained model. The source codes of this work are available at https://github.com/983632847/USCL.