Vision-based medical expert system Оleh Berezsky, Kateryna Berezska, Yuri Batko, Grygoriy Melnyk Computer Engineering Department, Ternopil National Economic University, Lvivska Str., 11, Ternopil, 46009, UKRAINE, E- mail: ob@tneu.edu.ua, bum@tneu.edu.ua, mgm@tneu.edu.ua Abstract - This paper addresses some of the issues involved in developing a technology that supports the implementation of an hybrid medical expert systems. The technology is based on the premise that integrated solution architectures will be much more effective and highly flexible in their in their ability to successfully handle a broad base of applications with a wider scope of problem variations.. Кеу words – hybrid expert systems, medical expert system, decision support system. I. Introduction Hybrid systems [1] in artificial intelligence represent a new field of research that deals with the combination of different artificial intelligence technologies. In this paper we carry idea of incorporation of image processing algorithms and rule-based reasoning into the hybrid medical system. In the following sections we briefly describe the nature of the modules and the principles involved in their integration. Biomedical images of the pathological process acquired with the use of an brightfield microscope. On the other hand, especially when using an automated microscopic system, the number of images acquired in the inspection process is very large (on the order of 1000 images/session). This renders the human visual inspection a very time-consuming (and tiring) process. An automated image analysis system, combined with an expert system to point the user directly to the images that contain objects and give information about the degree of pathological process in a linguistic fashion, can ease very much the visual inspection task. This is purpose of the system presented here. II. Structure of the proposed system In order to perform an automated visual investigation of the specimen and to provide as result of the investigation linguistic information about the pathological process in tissue, three components need to be included in our hybrid intelligent vision system: a) an image analysis component (IAC); b) an knowledge base component (KBC); c) an decision support system component (DSSC) to diagnose the pathological process. Based on the information received from the image analysis component and on the knowledge base “learned” from human experts visual, it will provide its diagnostic about the type of pathological process in the form of some linguistic labels. Image analysis component will be responsible for: 1) the rough localization of the region of interest. The region of interest will be the only image part used for further processing and data extraction; 2) the segmentation of the region of interest into areas belonging to the object of interest (i.e. to the cells), areas belonging to the background; 3) accurate separation of the areas corresponding to the object of interest and any other areas in the region of interest (background or obstacles) and quantitative description of the object of interest; this stage will provide the data needed by the DSSC in order to decide about the type of the pathological process and give a linguistic evaluation of this status. An expert system component will be responsible for: 1) creating rules by examples (inductive learning); 2) storing If-then rules with image features as conditions. The principled block-diagram of the proposed vision- based expert system is presented in Fig. 1. Fig 1. Structure HIS III.The image analysis component Image analysis component (IAC) is responsible for extracting useful information about the object of interest from the current histology image to be processed. The difficulty in processing biomedical images comes from their particularities, namely: low visibility, variable illumination, low contrast and blurring. Some of these particularities (as the variation of illumination and blurring) are partially introduced by the acquisition system. Module of images acquisition. Input images can be loaded from a disk or captured from a video recording hardware. The system allows capturing an image from a video recording hardware in real-time mode. Video information, which is captured from a video recording hardware, is displayed in the working window of the system. To transfer an image into the working area of the program, double click of the left button mouse is used. After that the program stops displaying of video stream and transfers captured image to module of pre-processing. Module of image pre-processing. In order to remove artefacts, improve quality and perform additional processing, functions of image pre-processing are used. Among accessible functions there are: selection of image part;  down-scaling of image;  transformation from one colour base to another;  correction of image brightness; Image analysis component Knowledge Base Decision Support System Interface