Title: Fast retrieval of similar images of pulmonary nodules based on deep multi-index hashing

Authors: Rui Hao; Yaxue Qin; Yan Qiang

Addresses: School of Information, Shanxi University of Finance & Economics, Taiyuan, Shanxi, China ' School of Information, Shanxi University of Finance & Economics, Taiyuan, Shanxi, China ' College of Information and Computer, Taiyuan University of Technology, Taiyuan, Shanxi, China

Abstract: CT image retrieval of pulmonary nodules mainly uses neural network embedded in hash layer to extract hash codes, in which deep h are directly as the index address for linear search. With the increasing number of clinical lung CT images and the complexity of image expression, the traditional retrieval methods are inefficient. We propose a multi-index hash retrieval algorithm based on deep hash features, which reduces the retrieval cost from linear search to sub-linear search. First, a hash layer is added to the Convolutional Neural Network (CNN) which can simultaneously learn the high-level semantic features of images and the corresponding hash function expression. Secondly, the hash codes extracted are effectively divided and multi-index tables are constructed. The query algorithm is designed based on the drawer principle. Finally, the complexity analysis of the whole index algorithm and experimental results show that the proposed algorithm can effectively reduce the retrieval cost while maintaining the accuracy.

Keywords: pulmonary nodule; deep learning; hash feature; image retrieval; multi-index hashing.

DOI: 10.1504/IJWMC.2023.135380

International Journal of Wireless and Mobile Computing, 2023 Vol.25 No.4, pp.303 - 308

Received: 07 Feb 2022
Received in revised form: 29 Mar 2022
Accepted: 13 Apr 2022
Published online: 08 Dec 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article