Back close

Analysis on the Effectiveness of Transfer Learned Features for X-ray Image Retrieval

Publication Type : Book

Publisher : Springer

Source : Innovative Data Communication Technologies and Application pp 251–265

Url :

Keywords : CBIR, Deep Learning, Transfer Learning, X-ray Image Retrievalm Adaptive Histogram Equalization, Image Enhancement

Campus : Coimbatore

School : School of Engineering

Department : Computer Science

Year : 2022

Abstract : A computer-assisted system for retrieving medical images of identical contents can be used as a data processing method for managing and mining large amount of medical data, as well as in clinical decision support systems. This paper studies the effectiveness of deep features extracted from state-of-the-art deep learning models for the retrieval of X-ray images. The first part of the study explores the effectiveness of transfer learned features generated from DenseNet model, Inception model and Inception-ResNet model for image retrieval. The performance of transfer learned features for image retrieval was analyzed based on the retrieval accuracy. The second part of the paper analyzes the effect of preproccesing using adaptive histogram equalization on image retrieval. The experiment is carried out on a publicly available musculoskeletal radiographs (MURA) dataset which consists of nearly 40,561 bone X-ray images of different body parts in varied angles with 7 classes.

Cite this Research Publication : Krishnan, G., Sikha, O.K. (2022). Analysis on the Effectiveness of Transfer Learned Features for X-ray Image Retrieval. In: Raj, J.S., Kamel, K., Lafata, P. (eds) Innovative Data Communication Technologies and Application. Lecture Notes on Data Engineering and Communications Technologies, vol 96. Springer, Singapore.

Admissions Apply Now