HA-SNet: A Best Choice to Solve Ultrasonic Images Based on Hybrid Attention Network

With the advancement of artificial intelligence, a large number of scientific research achievements that previously remained only on paper have been successfully landed. In modern society with uneven distribution of medical resources, it is of great significance to vigorously promote the integration of big data with the medical industry. Although there has been a long-term development of deep learning-based artificial intelligence in ultrasound medicine, the complexity of computing increases significantly while improving accuracy, which poses a considerable challenge in migrating from the laboratory to clinical applications. To overcome this challenge, we propose the HA-SNet (Hybrid Attention Segmentation network) model with a hybrid attention mechanism to achieve coarse-to-fine segmentation of ultrasound medical images. In particular, the hybrid attention module can improve the accuracy and robustness of the network without raising the model parameters and operational complexity. We extensively validate the HA-SNet network on the ultrasound dataset, and finally, we find that it has good prediction on ultrasound datasets.