A Low-cost Approach for Effective Shape-based Retrieval and Classification of Medical Images

Using relevance feedback to reduce the semantic gap in content-based image retrieval of mammographic masses

“…The database used in the present study is the same as that used by Felipe et al [5], consisting of 250 regions of interest (ROIs) comprising breast tumoral masses, taken from mammograms available in the Digital Database for Screening Mammography (DDSM) [7]. Images were selected randomly so as to have equal number of benign mass and malignant tumors.…”

“…The methodology used for feature extraction to index the images has been described by Felipe et al [5], and comprises techniques of image processing, data mining, and fractal theory. The method uses Zernike moments [9], as Using relevance feedback to reduce the semantic gap in contentbased image retrieval of mammographic masses shape features of ROIs for pattern recognition.…”

“…Zernike polynomials provide a precise mathematical model that captures global shape while preserving information related to local details. This technique was chosen because previous works [10,11] have demonstrated that Zernike moments present advantages over other techniques for image comparison based on shape features, and also because the moments can be extracted without the need for image segmentation [5]. Initially, for the moment order of 30, a set of 256 moments is generated for each image.…”

“…Initially, for the moment order of 30, a set of 256 moments is generated for each image. The StARMiner, a statistical association rule miner is then applied to identify representative features [5]. In order to minimize the computational effort, another algorithm, based on the Fractal Theory, was also applied to reduce the dimensionality of the representative feature space without loss of precision, through the identification and removal of correlated features inside the representative feature set [5].…”

“…Generic features of images may not be suitable for the analysis of similarity of medical images of a specific modality of interest. This work focuses on mammographic masses, using specific features that have been proven to be suitable for CBIR [5]. Although CBIR systems have been proposed for several applications, "the semantic gap" between the quantitative features used to represent the images and the interpretation of the images by users who are experts in the domain of application can impair the results obtained [1].…”