Catálogo de publicaciones - libros

Precise image and text indexing requires domain knowledge and a learning process. In this paper, we present the use of an ontology to filter medical documents and of visual concepts to describe and index associated images. These visual concepts are meaningful medical terms with associated visual appearance from image samples that are manually designed and learned from examples. Text and image indexing processes are performed in parallel and merged to answer mixed-mode queries. We show that fusion of these two methods are of a great benefit and that external knowledge stored in an ontology is mandatory to solve precise queries and provide the overall best results.

In this paper the methods we used in the 2005 ImageCLEF content-based image retrieval evaluation are described. For the medical retrieval task, we combined several low-level image features with textual information retrieval. Combining these two information sources, clear improvements over the use of one of these sources alone are possible.

Additionally we participated in the automatic annotation task, where our content-based image retrieval system, FIRE, was used as well as a second subimage based method for object classification. The results we achieved are very convincing. Our submissions ranked first and the third in the automatic annotation task out of a total of 44 submissions from 12 groups.

Images are difficult to classify and annotate but the availability of digital image databases creates a constant demand for tools that automatically analyze image content and describe it with either a category or set of words. We develop two cluster-based cross-media relevance models that effectively categorize and annotate images by adapting a cross-lingual retrieval technique to choose the terms most likely associated with the visual features of an image.

Image retrieval has great potential for a variety of tasks in medicine but is currently underdeveloped. For the ImageCLEF 2005 medical task, we used a text retrieval system as the foundation of our experiments to assess retrieval of images from the test collection. We conducted experiments using automatic queries, manual queries, and manual queries augmented with results from visual queries. The best performance was obtained from manual modification of queries. The combination of manual and visual retrieval results resulted in lower performance based on mean average precision but higher precision within the top 30 results. Further research is needed not only to sort out the relative benefit of textual and visual methods in image retrieval but also to determine which performance measures are most relevant to the operational setting.

This paper presents the approaches used by the MIRACLE team to image retrieval at ImageCLEF 2005. Text-based and content-based techniques have been tested, along with combination of both types of methods to improve image retrieval. The text-based experiments defined this year try to use semantic information sources, like thesaurus with semantic data or text structure. On the other hand, content-based techniques are not part of the main expertise of the MIRACLE team, but multidisciplinary participation in all aspects of information retrieval has been pursued. We rely on a publicly available image retrieval system (GIFT 4) when needed.

This paper presents the methods and experimental results for the automatic medical image annotation and retrieval task of ImageCLEFmed 2005. A supervised machine learning approach to associate low-level image features with their high level visual and/or semantic categories is investigated. For automatic image annotation, the input images are presented as a combined feature vector of texture, edge and shape features. A multi-class classifier based on pairwise coupling of several binary support vector machine is trained on these inputs to predict the categories of test images. For visual only retrieval, a combined feature vector of color, texture and edge features is utilized in low dimensional PCA sub-space. Based on the online category prediction of query and database images by the classifier, pre-computed category specific first and second order statistical parameters are utilized in a Bhattacharyya distance measure. Experimental results of both image annotation and retrieval are reported in this paper.

A combination of several classifiers using global features for the content description of medical images is proposed. Beside well known texture histogram features, downscaled representations of the original images are used, which preserve spatial information and utilize distance measures which are robust with regard to common variations in radiation dose, translation, and local deformation. These features were evaluated for the annotation task and the retrieval task in ImageCLEF 2005 without using additional textual information or query refinement mechanisms. For the annotation task, a categorization rate of 86.7% was obtained, which ranks second among all submissions. When applied in the retrieval task, the image content descriptors yielded a mean average precision (MAP) of 0.0751, which is rank 14 of 28 submitted runs. As the image deformation model is not fit for interactive retrieval tasks, two mechanisms are evaluated with regard to the trade-off between loss of accuracy and speed increase: hierarchical filtering and prototype selection.

In this paper we describe the technologies and experimental results for the medical retrieval task and automatic annotation task. We combine textual and content-based approaches to retrieve relevant medical images. The content-based approach containing four image features and the text-based approach using word expansion are developed to accomplish these tasks. Experimental results show that combining both the content-based and text-based approaches is better than using only one approach. In the automatic annotation task we use Support Vector Machines (SVM) to learn image feature characteristics for assisting the task of image classification. Based on the SVM model, we analyze which image feature is more promising in medical image retrieval. The results show that the spatial relationship between pixels is an important feature in medical image data because medical image data always has similar anatomic regions. Therefore, image features emphasizing spatial relationship have better results than others.

This article describes the use of and for three of four 2005 tasks. All results rely on two systems: the GNU Image Finding Tool () for visual retrieval, and for text. For ad–hoc retrieval, two visual runs were submitted. No textual retrieval was attempted, resulting in lower scores than those using text retrieval. For medical retrieval, visual retrieval was performed with several configurations of Gabor filters and grey level/color quantisations as well as combinations of text and visual features. Due to a lack of resources no feedback runs were created, an area where performed best in 2004. For classification, a retrieval with the target image was performed and the first = 1; 5; 10 results used to calculate scores for classes by simply adding up the scores for each class. No machine learning was performed, so results were surprisingly good and only topped by systems with optimised learning strategies.

We present a report on our participation in the English-Indonesian image ad-hoc task of the 2005 Cross-Language Evaluation Forum (CLEF). We chose to translate an Indonesian query set into English using a commercial machine translation tool called We used an approach that combines the retrieval results of the query on text and on image. We used query expansion in our effort to improve the retrieval effectiveness. However, worse retrieval effectiveness was resulted.