Oppositional Jellyfish Search Optimizer With Deep Transfer Learning Enabled Secure Content-Based Biomedical Image Retrieval

Abstract

Recently, a drastic increase in medical imaging such as X-rays, Magnetic Resonance Imaging (MRI), and Computed Tomography (CT) offers essential anatomical and functional details related to different body parts for monitoring, treatment planning, detection, and classification. The use of Content-Based Medical Image Retrieval (CBMIR) technologies helps to handle massive amounts of medical images, and also encryption can be considered an effective solution to attain security in the CBMIR process. In this regard, this research develops an optimal deep transfer learning enabled secure CBMIR technique, called ODTL-SCBMIR model. The proposed ODTL-SCBMIR prototype aim is to provide image encryption techniques with retrieval procedures. To accomplish this, the presented ODTL-SCBMIR model initially employs multikey homomorphic encryption (MKHE) with an oppositional jellyfish search optimizer (OJSO) algorithm for security. Next, the image retrieval process includes a series of processes namely capsule network (CapsNet) based feature extraction, chaos game optimization (CGO) based hyperparameter optimizer, and Manhattan distance based similar measurement. The performance validation of the ODTL-SCBMIR prototype is experimented with by employing a set of medical images. The investigational results implied the enhanced performance of the ODTL-SCBMIR prototype over current approaches.

Keywords

• Content-based medical image retrieval
• deep transfer learning
• hyperparameter optimization
• similarity measurement
• security
• encryption