Secure Content-Based Medical Image Retrieval Using Hybrid Optimized Deep Learning and Revised Genetic Rice Optimization Algorithm

Presently, medical imaging is critical to medical care, study, and education. Large-scale image banks were established for these objectives through worldwide cooperation. However, up until now, searching through extensive medical image collections was challenging, laborious, and mostly restricted by word search engines. Encryption is a viable method to achieve security in the Content- Based Medical Image Retrieval (CBMIR) procedure, and it assists in managing large volumes of medical images. An ideal deep transfer learning facilitated secured CBMIR method utilizing the multikey homomorphic encryption (MKHE) paradigm is developed in previous study in this area. Nevertheless, there is an issue with MKHE systems that needs to be resolved: their processing speed is not optimal. Therefore, Bald Eagle Search Optimal Pixel Selection with Henon chaotic map (BESOPS-HCE) was employed in the suggested study. Following that, a number of procedures are involved in the image retrieval (IR) process, including Densenet-121 driven feature extraction, Revised Genetic Rice Optimization Algorithm (RGRO) based hyperparameter optimizer, and Bidirectional LSTM (Bi-LSTM) classifier to obtain the relevant images and named as Hybrid Optimized Deep Learning (HODL). An image dataset of diseases is employed to test the suggested HODL-SCBMIR prototype's efficacy. The outcomes of the study suggested that the BESOPS-HCE-HODL-SCBMIR model outperformed existing methods in the area of efficiency