Title: A dynamic programming approach for accurate content-based retrieval of ordinary and nano-scale medical images

Authors: Jinhong Sun; Liang Qi; Yinglei Song; Junfeng Qu; Mohammad R. Khosravi

Addresses: School of Electronic and Information Sciences, Jiangsu University of Science and Technology, Zhenjiang, 212003, China ' School of Electronic and Information Sciences, Jiangsu University of Science and Technology, Zhenjiang, 212003, China ' School of Electronic and Information Sciences, Jiangsu University of Science and Technology, Zhenjiang, 212003, China ' Department of Computer Science and Information Technology, Clayton State University, Morrow, GA 30260, USA ' Department of Computer Engineering, Persian Gulf University, Bushehr, 999067, Iran

Abstract: Recently, with the explosive growth in the number of available medical images generated by medical imaging systems, content-based retrieval of medical images has become an important method for the diagnosis and study of many diseases. Most existing methods find medical images similar to a given one based on the extraction and comparison of crucial image features. However, similarity values computed with low level visual features of an image generally do not match the similarity obtained from human observation well. The overall performance of these methods is thus often unsatisfactory. This paper proposes a dynamic programming approach for content-based retrieval of medical images. The approach represents an image with three different histograms that contain both crucial intensity and textural features of the image. The similarity between two images is evaluated with a dynamic programming approach that can optimally align the peaks in the corresponding histograms from both images. Experiments show that the proposed approach is able to generate retrieval results with high accuracy. A comparison with state-of-the-art approaches for content-based medical image retrieval shows that the proposed approach can achieve higher retrieval accuracy in both ordinary and nano-scale medical images. As a result, higher retrieval accuracy may lead to more reliable results for the diagnosis and treatment of many diseases. The proposed approach is thus potentially useful for improving the reliability of many applications in health informatics.

Keywords: medical image retrieval; similarity; alignment of histograms; dynamic programming; intensity features; textual features.

DOI: 10.1504/IJNT.2023.131104

International Journal of Nanotechnology, 2023 Vol.20 No.1/2/3/4, pp.75 - 97

Received: 05 Feb 2021
Accepted: 06 May 2021
Published online: 31 May 2023 *