A novel content-based medical image retrieval method based on query topic dependent image features (QTDIF)

摘要

Medical image retrieval is still mainly a research domain with a large variety of applications and techniques. With theImageCLEF 2004 benchmark, an evaluation framework has been created that includes a database, query topics andground truth data. Eleven systems (with a total of more than 50 runs) compared their performance in variousconfigurations. The results show that there is not any one feature that performs well on all query tasks. Key to successfulretrieval is rather the selection of features and feature weights based on a specific set of input features, thus on the querytask. In this paper we propose a novel method based on query topic dependent image features (QTDIF) for content-basedmedical image retrieval. These feature sets are designed to capture both inter-category and intra-category statisticalvariations to achieve good retrieval performance in terms of recall and precision. We have used Gaussian MixtureModels (GMM) and blob representation to model medical images and construct the proposed novel QTDIF for CBIR.Finally, trained multi-class support vector machines (SVM) are used for image similarity ranking. The proposed methodshave been tested over the Casimage database with around 9000 images, for the given 26 image topics, used forimageCLEF 2004. The retrieval performance has been compared with the medGIFT system, which is based on the GNUImage Finding Tool (GIFT). The experimental results show that the proposed QTDIF-based CBIR can providesignificantly better performance than systems based general features only.