We propose the application of two-dimensional distortion models for comparisons of medical images in a distance-based classifier.

Abstract. In this paper, the automatic annotation task of the 2005 CLEF cross-language image retrieval campaign (ImageCLEF) is described. This paper focuses on the database used, the task setup, and the plans for further medical image annotation tasks in the context of ImageCLEF. Furthermore, a short summary of the results of 2005 is given. The automatic annotation task was added to ImageCLEF in 2005 and provides the first international evaluation of state-of-the-art methods for completely automatic annotation of medical images based on visual properties. The aim of this task is to explore and promote the use of automatic annotation techniques to allow for extracting semantic information from little-annotated medical images. A database of 10.000 images was established and annotated by experienced physicians resulting in 57 classes, each with at least 10 images. Detailed analysis is done regarding the (i) image representation, (ii) classification method, and (iii) learning method. Based on the strong participation of the 2005 campain, future benchmarks are planned. Keywords: content-based image retrieval, medical image annotation, evaluation in computer vision52 Deselaers et al. 1.

Methods for the recognition of multiple objects in images using local representations are introduced. Starting from a straight forward approach, we combine the use of local representations with region segmentation and template matching. The performance of the classifiers is evaluated on four image databases of different difficulties. All databases consist of images containing one, two or three objects and differ in the backgrounds which are used. Also, the presence or absence of occlusions of the objects in the scenes is considered. Classification results are promising regarding the difficulty of the task.

Abstract—This work presents an image representation and matching framework for image categorization in medical image archives. Categorization enables to determine automatically, based on the image content, the examined body region and imaging modality. It is a basic step in content-based image retrieval (CBIR) systems, the goal of which is to augment textbased search with visual information analysis. CBIR systems are currently being integrated with Picture-Archiving and Communication Systems (PACS) for increasing the overall search capabilities and tools available to radiologists. The proposed methodology is comprised of a continuous and probabilistic image representation scheme using Gaussian mixture modeling (GMM) along with information-theoretic image matching via the Kullback Leibler (KL) measure. The GMM-KL framework is used for matching and categorizing x-ray images by body regions. A multi-dimensional feature space is used to represent the image input, including intensity, texture and spatial information. Unsupervised clustering via the GMM is used to extract coherent regions in feature space which are then used in the matching process. A dominant characteristic of the radiological images is their poor contrast and large intensity variations. This presents a challenge to matching between images and is handled via an illumination invariant representation. The GMM-KL framework is evaluated for image categorization and image retrieval on a dataset of 1500 radiological images. A classification rate of 97.5% was achieved. The classification results compare favorably with reported global and local representation schemes. Precision vs. Recall curves indicate a strong retrieval result as compared with

Abstract. In this paper, we employ a zero-order local deformation model to model the visual variability of video streams of American sign language (ASL) words. We discuss two possible ways of combining the model with the tangent distance used to compensate for affine global transformations. The integration of the deformation model into our recognition system improves the error rate on a database of ASL words from 22.2 % to 17.2%. 1

Abstract. We introduce the use of appearance-based features in hidden Markov model emission probabilities to recognize dynamic gestures. Tangent distance and the image distortion model are used to directly model image variability in videos. No explicit hand models and no segmentation of the hand is necessary. Different appearance-based features are investigated and the invariant distance measures are systematically evaluated. The approach is evaluated for three tasks of strongly varying difficulty and and performs favorably well. We obtain promising first results on a novel database of the German finger-spelling alphabet. 1

Content--based image retrieval systems (CBIRSs) have frequently been proposed for the use in medical image databases and PACS. Still, only few systems were developed and used in a real clinical environment. It rather seems that medical professionals define their needs and computer scientists develop systems based on data sets they receive with little or no interaction between the two groups. A first study on the diagnostic use of medical image retrieval also shows an improvement in diagnostics when using CBIRSs which underlines the potential importance of this technique.

Abstract. We introduce the use of appearance-based features, and tangent distance or the image distortion model to account for image variability within the hidden Markov model emission probabilities to recognize gestures. No tracking, segmentation of the hand or shape models have to be defined. The distance measures also perform well for template matching classifiers. We obtain promising first results on a new database with the German finger-spelling alphabet. This newly recorded database is freely available for further research. 1

Abstract. We present a holistic statistical model for the automatic analysis of complex scenes. Here, holistic refers to an integrated approach that does not take local decisions about segmentation or object transformations. Starting from Bayes ’ decision rule, we develop an appearancebased approach explaining all pixels in the given scene using an explicit background model. This allows the training of object references from unsegmented data and recognition of complex scenes. We present empirical results on different databases obtaining state-of-the-art results on two databases where a comparison to other methods is possible. To obtain quantifiable results for object-based recognition, we introduce a new database with subsets of different difficulties. 1

Building effective content-based image retrieval (CBIR) systems involves the combination of image creation, storage, security, transmission, analysis, evaluation feature extraction, and feature combination in order to store and retrieve medical images effectively. This requires the involvement of a large community of experts across several fields. We have created a CBIR system called Archimedes which integrates the community together without requiring disclosure of sensitive details. Archimedes ' system design enables researchers to upload their feature sets and quickly compare the effectiveness of their methods against other stored feature sets. Additionally, research into the techniques used by radiologists is possible in Archimedes through doubleblind radiologist comparisons based on their annotations and feature markups. This research archive contains the essential technologies of secure transmission and storage, textual and feature searches, spatial searches, annotation searching, filtering of result sets, feature creation, and bulk loading of features, while creating a repository and testbed for the community. 1.