Showing papers by "Avinash C. Kak published in 1999"

Content-based retrieval from medical image databases: a synergy of human interaction, machine learning and computer vision

Abstract: Content-based image retrieval (CBIR) refers to the ability to retrieve images on the basis of image content. Given a query image, the goal of a CBIR system is to search the database and return the n most visually similar images to the query image. In'this paper, we describe an approach to CBIR for medical databases that relies on human input, machine learning and computer vision. Specifically, we apply expert-level human interaction for solving that aspect of the problem which cannot yet be automated, we use computer vision for only those aspects of the problem to which it lends itself best - image characterization - and we employ machine learning algorithms to allow the system to be adapted to new clinical domains. We present empirical results for the domain of high resolution computed tomography (HRCT) of the lung. Our results illustrate the efficacy of a human-in-the-loop approach to image characterization and the ability of our approach to adapt the retrieval process to a particular clinical domain through the application of machine learning algorithms.

The customized-queries approach to CBIR using EM

Abstract: This paper makes two contributions. The first contribution is an approach called the "customized-queries" approach (CQA) to content-based image retrieval. The second is an algorithm called FSSEM that performs feature selection and clustering simultaneously. The customized queries approach first classifies a query using the features that best differentiate the major classes and then customizes the query to that class by using the features that best distinguish the images within the chosen major class. This approach is motivated by the observation that the features that are most effective in discriminating among images from different classes may not be the most effective for retrieval of visually similar images within a class. This occurs for domains in which not all pairs of images within one class have equivalent visual similarity, i.e., subclasses exists. Because we are not given subclass labels, we must simultaneously find the features that best discriminate the subclasses and at the same time find these subclasses. We use FSSEM to find these features. We apply this approach to content-based retrieval of high-resolution tomographic images of patients with lung disease and show that this approach radically improves the retrieval precision over the traditional approach that performs retrieval using a single feature vector.

Testing for human perceptual categories in a physician-in-the-loop CBIR system for medical imagery

Abstract: We address the following question: to what extent should the domain experts (in our case, physicians), be believed with regard to what they claim to see in images that allows them to recognize different types of pathology? Until recently our approach was to have a physician delineate the pathology bearing regions in the images. We then used what could be referred to as a scattershot approach to the characterization of these regions, meaning that we'd extract a very large number of features from these regions. Subsequently, we'd reduce the dimensionality of this feature space by using standard search techniques, such as the Sequential Forward Selection method. The article represents an alternative to the scattershot approach to initial feature extraction. We first describe the perceptual categories that the physicians claim to use for classifying images as belonging to different diseases. We then describe the specific low-level features that need to be extracted to determine the presence or the absence of the various perceptual categories. We subsequently show the discriminatory power of the perceptual categories by presenting retrieval results obtained when a query image is matched with the database images on the basis of the presence or the absence of the various perceptual categories.

Interactive color image segmentation editor driven by active contour model

Abstract: A new general-purpose color image segmentation editor (CISE) for the purpose of extracting a semantic object is designed, implemented and tested on a number of various natural scene images. Our editor integrates a deformable model and image statistics including intensity, color, gradient and texture. The editor starts with a coarse region segmentation which applies the Canny's operator followed by a low-complexity edge linking algorithm. This segmentation basically builds regions of smooth intensity by closing all dangling edges. Next, a topologically-based region labeling method makes full use of relationship among pixels and produces useful labeled image. Finally, to refine the extracted object of interest (O/sup 2/I), a deformable model based on energy minimization is applied by incorporating both the gradient and region criteria to the external constraint force. These processes are demonstrated through examples on natural scene color images. Experimental results suggest the efficiency and accuracy of the algorithm in its segmentation operations.