An extensive study of visual search models on medical databases

Abstract

Due to the rapid growth of medical images, user-specific ROI and object classification are the significant factors in the region-based segmentation instead of a pixel-based segmentation. Manual image annotation, classification, and filtering are not only a difficult task, but also high memory and time usage. In the visual search system, an unknown query image was given as input, relevant visual images with different diagnoses features are retrieved and then used as clinical decisions. The main goal of the visual search engine is to efficiently retrieve user-specific images that are visually identical to a selected ROI query. In this paper, a survey on traditional visual search methods is analyzed in terms of visual features and accuracy are concerned. Based on the survey performed by different visual search systems, the diagnostic efficiency is increased from 30 to 60% for clinical decision.