Nadia Bianchi

Multimedia document management: an anthropocentric approach

The architecture of an anthropocentric Biomedical Information Management System is proposed. The system is based on a network of computational components (agents) that users-experts in the biomedical domain, but not in computer science-can define, use, and refine to serve their own communication and documentation habits and needs. The innovation of the proposal lies in the adoption of the Participatory Design methodology, which involves users in the design of tools and their implementation, organization and integration to perform the required biomedical activities.

Situated Image Understanding in a Multiagent Framework

The paper addresses the problem of controlling situated image understanding processes. Two complementary control styles are considered and applied cooperatively, a deliberative one and a reactive one. The role of deliberative control is to account for the unpredictability of situations, by dynamically determining which strategies to pursue, based on the results obtained so far and more generally on the state of the understanding process. The role of reactive control is to account for the variability of local properties of the image by tuning operations to subimages, each one being homogeneous with respect to a given operation. A variable organization of agents is studied to face this variability. The two control modes are integrated into a unified formalism describing segmentation and interpretation activities. A feedback from high level interpretation tasks to low level segmentation tasks thus becomes possible and is exploited to recover wrong segmentations. Preliminary results in the field of liver biopsy image understanding are shown to demonstrate the potential of the approach.

Distributed Plan Construction and Execution for Medical Image Interpretation

Research in the field of medical image interpretation inspired a multi-agent planning model in which a population of agents reaches global goals pursuing local plans and producing partial results. Cooperation is achieved by collecting results, reanalysing the reached state and starting new plans; local plans are generated and executed in a distributed way. The state is composed of a directed graph, describing the currently obtained results and of an attributed string, concatenating the trace of the performed operations with the current plan. A model of agent and agent generation is presented and an example of application given in the field of liver biopsy interpretation.

A dynamical organisation for situated image interpretation

Two complementary control styles, deliberative and reactive types, are applied to account for unpredictability of situations and variability of local properties in image interpretation processes. Deliberative control is based on attributed grammars and local variability is managed by a variable organisation of agents. Feedback from interpretation to segmentation becomes possible and is exploited to recover wrong segmentation. Experimental results illustrate the potential of the approach.

Integration of Neural Networks and Rule Based Systems in the Interpretation of Liver Biopsy Images

Treatment of natural images requires, due to their complexity, to exploit high level knowledge, such as domain knowledge and heuristics, which are typically well formalized by rule based systems. However, the intrinsic variability and irregularity of objects in the image makes their characterization in terms of rules often unfeasible. Such variability and irregularity are, on the other hand, the ultimate reason for the existence of statistical methods. For these reasons, a hybrid system, exploiting characteristics of both approaches, may show better performances than purely syntactical or statistical systems in the interpretation of natural images. In this paper we present a hybrid system for image interpretation that integrates a rule based system with a Labeled Learning Vector Quantizer. The rule based system controls the interpretation process, by dynamically determining the interpretation strategy, and the Labeled Learning Vector Quantizer is exploited as classification kernel. The system has been tested on images of liver biopsies. Results on nuclei classification are here discussed.

Plastic visual tools

A plastic visual system is presented, based on a hierarchy of environments that users can define, refine and use. In each environment, experts build visual expressions specifying activities to be performed at a lower level. The mechanisms exploited to obtain these features are based on uncoupling the organisation of the interface from the definition of its actual components, objects managed by the applications from their views and views from the components of the interface which present them.<<ETX>>