Medical image retrieval based on deep hashing

Abstract

The deep hash method has been proved to be the most efficient and approximate nearest neighbor search technique for large-scale image retrieval. However, the existing deep hashing methods have few cases in medical image retrieval, so we propose a deep hash network model for medical image retrieval. Firstly, in order to enhance the feature extraction and learning ability in the network model, we add SENet visual attention mechanism into each residual module of the backbone network ResNet50 for feature extraction. Then the extracted high-dimensional semantic features are added to the Cauchy hash module to generate a compact, centralized hash code, and then complete medical image retrieval. So as to enhance experimental comparison, the proposed method was validated experimentally on NUS-WIDE and MURA data sets. A large number of experiments verify that the proposed method is reliable and can be used for medical image retrieval. Compared with other existing hash algorithms, the retrieval performance of the proposed method has certain improved. The framework of the proposed method is shown in Figure 1.