The purpose of this work was to evaluate a newly developed content-based retrieval approach for characterizing a range of different white blood cells from a database of imaged peripheral blood smears. Specimens were imaged using a 20x magnification to provide adequate resolution and sufficiently large field of view. The resulting database included a test ensemble of 96 images (1000x1000 pixels each). In this work, we propose a four-step content-based retrieval method and evaluate its performance. The content-based image retrieval (CBIR) method starts from white blood cell identification, followed by three sequential steps including coarse-searching, refined searching, and finally mean-shift clustering using a hierarchical annular histogram (HAH). The prototype system was shown to reliably retrieve those candidate images exhibiting the highest-ranked (most similar) characteristics to the query. The results presented here show that the algorithm was able to parse out subtle staining differences and spatial patterns and distributions for the entire range of white blood cells under study. Central to the design of the system is that it capitalizes on lessons learned by our team while observing human experts when they are asked to carry out these same tasks.Keywords: Content-based image retrieval, white blood cells
INTRODUCTIONThe exponential growth of images and video in last decade has resulted in an increasing need for efficient contentbased image retrieval (CBIR), which enables investigators to detect and locate similar images in large collections given an example query. For medical diagnostic assistance, several state-of-the-art CBIR systems [1][2][3] have been designed to support the processing of queries across separate images. However, often times users are interested in sub-region searching, usually to identify an image patch exhibiting specific patterns or structures or containing an object which is similar to one in the query patch. Given this sub-region, the system should be able to return other patches within the same set of images which contain localized sub-regions exhibiting similar features. In practice, this approach makes it possible for a pathologist to select an area or object of interest within a digitized bio-specimen as a query to reliably search corresponding regions in either the same bio-specimen or within different specimens from large cohorts of patients having the same disease. The results can then be used to enable physicians to draw comparisons among the samples in order to make informed decision regarding the prognosis and preferred treatment regimens.Recently researchers have proposed many state-of-the-art methods to perform CBIR related to both natural and medical images. Luo and Nascimento [4] introduced relevance feedback by applying a tile re-weighting approach to assign penalties to tiles that compose database images and update the penalties for all retrieved images within each iteration. This procedure is time consuming due to the reliance of the algorithm on feedback learning. T...