Areas of specialization The Institute of Computer Science conducts research in the following areas: Expert systems
The primary aim of studies is the development of the theoretical basis for designing and implementing expert systems. The validation and verification methods of knowledge databases are investigated. The design principles of support decision systems with knowledge databases verification are developed. Besides theoretical studies concerning the methods of knowledge base creation, the works on practical implementation of the support decision systems comprising knowledge verification module are also undertaken. The implementation of such systems requires an analysis of methods of knowledge representation and concluding.
The other direction of
research is an application of the rough set theory for constructing support
decision systems. With this regard the studies focus on the problem of
multi-step diagnosing based on uncertain and incomplete information. Making
up a decision requires some rough classification to be done where the steps
of the classification are connected with a real diagnosing process carried
out by experts. An example of such a complex diagnosing system is the
support decision system for children neurology, which was realised in
co-operation with the Children's Neurology Clinic of the Silesian Medical
Academy. On this system the proposed solutions based on the rough sets
theory are verified.
Heuristic and evolutionary algorithms
The heuristic and evolutionary algorithms for solving combinatorial optimization problems are studied. These problems arise in many areas of applications. The algorithms use the concepts derived from artificial intelligence, biological, mathematical, and natural and physical sciences. The ant systems in the context of the traveling salesman, bus-scheduling and delivery problems are investigated. Special attention is paid to the generative policies improving the performance of ant systems. The leader and elite strategies modelled upon the behaviour of real ants are examples of such policies. The delivery problem which reduces to the NP-complete set-partitioning problem is also solved by making use a variety of algorithms, e.g. k-optimal, tabu search, simulated annealing. All these algorithms employ a local neighbourhood search and are probabilistic in nature. The way a neighbourhood structure is defined and randomness is introduced influence significantly the performance of algorithms. Yet another direction of research we proceed is genetic programming. In this regard the influence of grammars describing the generated programs on the efficiency of genetic programming is studied.
The neural network selection for a medical diagnostic system using an artificial data set
Various experiments with a neural network selection that works as a conclusion-making unit of walk-abnormalities diagnosis have been carried out recently. The diagnostic interfaces provide the user with various tools for the disease analysis. They are having a pressure and load distribution on the foot, while taking into account the individual characteristics of the patient standing and walking. The visualisation options give the user many aims in putting the diagnosis anyhow, due to simplify the diagnostic process several methods for the data record filtering have been implemented. The discussed methods of the neural network selection and training show how to avoid difficulties with limited number of available data records, needed for the conclusion algorithms effectiveness improvement.
The first step of the disease detection concerns a static part of the data record. The user can put his first assumptions to the diagnosis; the rough estimation of the footprint abnormality, namely a class of a foot disease. In most cases a foot shape disease is visible on the footprint map (a static map), but the disease real source can not be established precisely. The final conclusion has to be searched within the dynamic part of the data record where the way of walking is analysed. The abnormality of time distribution in the cycle of walking allows the user to extract the reasons of patient troubles, underlying a nature of problems.
Experiments have been realized based on three types of neural networks: Back Propagation (BP) which uses supervised learning, hybrid Counter-Propagation (CP) which uses both supervised and unsupervised learning, ART-1 which uses unsupervised learning only. The researches showed it is possible to take advantage two neural networks for implementation of automatic conclusion-making system: BP and CP.
Researches have been realized for two classes of neurological diseases: Parkinson’s disease and hemiparesis after inchemic stroke. There is also possibility to make research for another neurological diseases, like sciatic neuralgia or ischialgia.
The interactive explorers of web presentation units
Majority of data files available in web servers contains not only various data formats. They are linked together into composition producing sequences that have to be read frame by frame – in a movie mode. The multimedia database design methods have to consider many rules that must be integrated to obtain several standards for simplifying the application development and for fast exploration of the product. The abstract presents several aspects of the design technique giving a unified composition of the project, modularity of the application and the navigation standards within the product. The developer of this kind of application meets many problems concerning the data file contents description, the internal data recognition and the key frame selection.
One of the application fields that we can notice concerns teaching packages. The Computer Aided Learning (CAL) systems have to provide the user with an individual path through the application, both in contents and in a time. This sequential structure of the application can be found in many sequential databases. The ability of selecting the individual path through the application means that several branches in the data file are available.
Design and diagnostics of digital circuits and systems
The studies concern problems of system dependability, and are connected with diagnostics of digital circuits and systems and also built-in self-test technique. The studies concentrate on the methods of designing and evaluating the efficiency of built-in diagnostics structures. Several simulation systems for determining the efficiency of structures have been developed. Furthermore, an advisory based on artificial intelligence i.e. bio-inspired and evolutionary hardware system design, which aids the design easy-, and self-testing systems is being created.
Analysis and processing of biomedical images
The project is devoted to the algorithms of analysis, processing and recognition of images applied in the identification of pathological states. By images we understand „classic” biomedical images. This group comprises X-ray, ultrasonic, thermovision images, as well as the microscopic images of tissues. In addition, two-dimensional images, so called biomedical signals, are analysed and processed. Among them are EKG and EEG signals, cardiograph records of heart action, sounds of pathological speech etc.
Hospital computer systems
The aim of hospital computer systems is to improve acquisition, transmission and processing of data generated by measuring sensors and medical apparatuses. This in turn improves the quality of medical care, decreases its cost and has a positive impact on the administrative and financial activities of hospitals. In the project the distributed system of dynamic microbiological investigations in the networked computer system of a hospital is elaborated.
Computer techniques in biotechnology
The work of biological objects can be considered as a problem of controlling a multi-level object described by a set of parameters related with each other. The research group working on the application of computer techniques in biotechnology constructs new measurement converters and medical apparatuses in order to perform complex biomedical and biotechnological procedures. The recent project was devoted to programmable control of biotechnological systems.
Computerisation of administrative processes
The subject of research is expert systems for the local administration needs. The research concentrates on general conditions of the computerisation of administrative processes. In particular they comprise the legal foundations of informatization of management processes, such as the structure and tasks of the administrative bodies and the legal rules of using computer programs and automated data bases. One of the research subjects concerns the systems of spatial information systems and personal data protection.
Main Page General Research
Publications
Education
Staff Location |
|||||