Optimized transfer learning based multi-modal medical image retrieval

Disease diagnosis using the medical image is a very technical and tedious process. Small abnormalities in multiple medical images could be noticed by medical specialists but deep analysis of a medical image is still a complicated task due to the restricted ability of the human visual system. The limitations of the human visual system might lead to medical treatment impairment. This issue, however, may be handled by searching for similar cases in the preceding medical database using an effective content-based medical image retrieval (CBMIR) method. The CBMIR’s main problem is efficient classification but also required retrieval from multi-modal medical imagery information. Most prior attempts at medical image retrieving and classification employ handmade features, that perform poorly across a large collection across multimodal datasets. Even though there has been a few earlier research on using deep characteristics for classification, the total count is quite modest. To address this issue, we offer an upgraded Inceptionv3 network, which is a genetic algorithm-based optimum retrieval system incorporating multimodal medical images from multiple forms of imaging systems. The experimental findings from 5 classes are showing accuracy and optimization with F1. score using our technique is 97.22%, and 89.56% with 98.53%, respectively, each of which is higher than either the accuracy but also F1. score from the prior solution of CBMIR.