|
EU key project: The research and development of the modern technology (project approved, 1st January 2008 – 31th December 2011)
U¦ theme of the project: Decision support – the new generation systems The development of the practical new generation systems that is the goal of this subproject requires conducting wide-ranging basic research, mainly in the following areas:
A universal system for decision support with built-in tools for the processes of complex conclusions and knowledge base self-organization as well as new architectures and paradigms will be created. The task is to design and create a software realization of a system of a unique architecture. Within this task, it is planned to create the concept of the self-organizing structure of a hierarchical knowledge base, methods for its verification and to develop modified conclusion algorithms for complex knowledge bases (the concept of hierarchical rule clusters, decision units, Petri nets, etc.). It is also intended to use and develop rough sets with the employment of decision patterns, rules and partial reducts for acquisition knowledge from incomplete information. For the graph analysis of knowledge representation, ant colony algorithms flow graphs and cognitive networks are to be used. A self-organizing hierarchical knowledge base with integrated methods for the verification of knowledge base correctness is to be developed. The knowledge gathered in the system will be acquired with different methods and from various sources (from experts, from large databases, from distributed network databases). Acquiring and managing incomplete knowledge (rough sets, neural networks, cognitive networks) is also planned. There will be conducted the optimization of the contents of the hierarchical knowledge base and conclusion processes under the conditions of using knowledge from heterogeneous sources. The implementation and realisation of the support decision system will be the final stage of the researches. Useful versions of the system will operate on real data referring to:
Complex knowledge bases: the structure and the inference processes
MSHE promotor’s grant No.: N 206 1531 33, 2007-2008, 25.000 PLN
The primary aim of this project is to analyze the problem of complex knowledge bases in decision support systems. The knowledge base with large number of rules, which have complex inner structure, we called complex knowledge base. It is necessary to use data-mining techniques for searching dependences in data and clustering similar rules. Cluster Analysis is a method that makes possible to built hierarchy of rules (hierarchical knowledge bases). Efficient methods of searching such hierarchical structures help to decrease the time of rules interpreter in decision support system. It gives an optimization of the inference processes in these systems. The very important issue of the project is to analyze the both: forward and backward chaining techniques of inference processes on hierarchical complex knowledge bases. With the theoretical researches the practical implementation is connected. The result of the project will be the application of decision support system with complex knowledge bases with researches realized for real data sets (medical or economy diagnosis systems).
Appliance of rough sets theory in data analysis of complex medical data
MSHE promotor’s grant No.: N 206 1531 33, 2007-2008, 25.000 zł Nowadays capabilities of generating and collecting data within computer systems are growing rapidly. Gathered data are stored in huge and powerful database systems. Common for database systems is that they do not offer enough possibilities to analyze data they manage. Additionally a process of discovering natural phenomena or complex system by finding formulas that fit the collected, empirical data reaches its limits when the complexity of the natural processes increases. Therefore, to analyze such data an alternative approach is needed. Several mathematical methods including neural nets, inductive learning fuzzy sets and rough sets were proposed to model data in a form of decision tables or rules. Pawlak's Rough Sets Theory for handling imprecision and uncertainty in data has a main advantage over the other techniques, which is the possibility to analyze data without any preliminary information what favors its usage in medical decision systems. Such decision systems can be used to improve medical care, uncover new relationships among data and reveal new patterns that can help to identify diseases or suggest more effective treatments. At present most research efforts in machine learning is directed toward the invention of new algorithms and much less into gaining experience in applying those to important practical applications. Therefore, the goal of our research project is both to propose optimal algorithms for data analysis and to develop a complete decision system, which joins advantages of several data mining techniques and helps doctors and clinicians to understand relations between complex medical data. Data used in our research are obtained from Electrocardiology Department of Silesian Medical Academy in Katowice - the leading Electrocardiology Department in Poland specializing in hospitalization of severe heart and arrhythmia diseases. This data describes more then 4500 patients hospitalized between 2003 and 2006 in this department. To extract a human understandable model from original clinical free-text reports several stages are needed including: transformation to binary attributes, joining and selecting of important attributes, inducing decision rules, reducing and joining of the generated decision rules, validation and data visualization. Algorithms proposed within our research project will be continuously presented at international journals and conferences and finally described by G. Ilczuk as a Ph.D. thesis.
Model of spatial distribution of bone density, which allows computer analysis of mechanical properties of bone
MSHE grant No.: N518 3018 33, 2007-2008, 76.000 PLN The one of the main action included in Biomedical Engineering is performing of the computer simulation in solving medical and biological problems. This area is developing rapidly mainly thanks to progress in technology. Few years ago calculations of the numerical models could be done only in the specialized computing centers. Today past simulations can be performed within one hour on an average home personal computer. Fast technology development allows to construct very complicated models which are the basis of advanced simulations. Such simulations can be performed based on the novel numerical methods. It has big influence on the acceleration also other fields of science, which are based on the simulations methods. The basis of the simulation is well constructed model. In the case of simulation of the bone tissue the finite element modelling is a very useful and popular method. Many units is still trying to study the best methods of the model construction. There are many different methods of the model construction including different way to define properties of the tissue. The main aim of the project is combing in one act the knoledge from biomedical engineering and computer science toward to design software toolkit which can provide computer added construction of the numerical model of bone tissue. Consutrcted model will provide to perform the simulation with using spatial distribution of bone density (SDBD). Complex analisis of the properies of modeled bone tissue will be possible. The analisis will include the densitoemtry and mechanical parameters as well. Based on the SDBD model (which will be constructed as a results of this project) searching of the new solution to treat civilisation deseases will be possible. Activities performed during describing engineering investigations will be a very important support for clinical studies like better performing of diagnostic procedures (also monitoring of treatment) of the bone tissue deseases egg. Osteoporsis, osteoartrosis, and also bone cancer. The possitive results of the project can be predicted with high probability. The next step will include performing modern software of the tomography scanners.
The methods of image analysis and processing applied to determining a microtubule inclination angle
MSHE grant No.: N 518 005 31/0336, 2006-2007, 55.000 PLN
The aim of the project is to develop a computer-based method for the analysis of microtubule orientation based on digital image processing techniques. The project will cover the following:
Using the Hough transformation in image comparison and classification method
MSHE promotor’s grant No.: 3 T11C 070 28, 2005, 31.800,00 PLN In this work the new method of image comparison and classification has been stated. Proposed method allows to extract image feature by means of modified Hough transformation and by binary contour parameterization algorithm. In dissertation two approaches of classifying objects have been presented. The first approach means assigning individual line segment to appropriate area of section. The second approach is based on image and pattern analysis, where line sections constitute the decision rules. The object belongs to a given category if similarity measure is determined. A new similarity measure definition has also been presented and compared with well-known other measures. In investigation the vehicle and medical (x-ray) images database has been prepared. Additionally, in computations the Internet signature image database has been used. Credible verification of proposed method enable selection of mentioned images. Proposed and verified classification method was compared with other methods in the literature, such as: measurement of geometrical property, overlap test, analyse of Fourier spectral coefficients. The classification effectiveness factors as well as computational complexity have also been analysed.
Geometrical wavelets and their generalizations in digital image coding and processing
MSHE promotor’s grant No.: 3 T11A 017 28, 2005, 22.490,00 PLN In the work the generalization of geometrical wavelets, namely wedgelets and beamlets to the ones based on second-degree curves has been proposed. It has been proved that the use of generalized wedgelets improves effectiveness of image coding. The carried out experiments have also confirmed the fact that the use of the new wavelets both in image coding and noisy images processing provides better results in comparison to other known methods. Additionally, the new application of geometrical wavelets to feature extraction of different meaning to HVS has been pointed out. Basing on the recent research from neuropsychology and psychology of vision it has been noted that it is a strong need for building of a feature extractor which can work in geometrical multiresolution way. So in the work the beamlets- and wedgelets-based operators have been proposed and defined which, as shown, effectively filter different features from binary and grayscale images, respectively.
Methods of disease classification on the basis of gait characteristic features
KBN promotor’s grant No.: 3 T11E 055 26, 2004-2005, 24.000,00 PLN An aim of the project is elaboration of the neurological disease classification method on the basis of gait disturbance features and patient’s balance disturbance. Second main goals of the project are implementation of selected classification methods and clinical verification of the methods. During researches the diagnostic measures were defined on the basis PSW data record. The diagnostic measures determine a set of input data of automatic conclusion-making unit. Computer software package, based on artificial intelligence, serves as automatic conclusion-making unit about neurological diseases on the basis of appointed diagnostic measures. Researches are realized for two classes of neurological diseases: Parkinson’s disease and hemi paresis after inchemic stroke. Till now, medicine doesn’t have the objective method of diseases classification, in particular Parkinson’s disease. There isn’t also possibility of making a diagnosis of Parkinson’s disease in early stage. The medical diagnosis is made not before than in visibly symptom of the disease, that means really bad state of patient’s health. This state of health means an impossibility of return to patient’s normal functionality.
Automatic programming
KBN grant No.: 7 T11C 021 21, 2001-2003, 150.000,00 PLN The subject of theoretical and experimental research of the project includes the methods of automatic creation of computer programs, or for short, the methods of automatic programming. These methods make possible to obtain a desired program without a tedions task of writing such a program. It is achieved by specifying first the goals, which are to be realized by the program. Then, based on this specification, the program is constructed automatically. In the project the approximation problems, which consist in a choice of an optimum function from some class of functions are considered. Approximation problems are encountered in analysis of numerical data, modeling physical phenomena, analysis of statistical observations etc. The aim of the project is to determine the usefulness of two selected methods of automatic programming, i.e. genetic programming and ant colony programming for solving the approximation problems.
Dynamic verification of knowledge databases during the process of their design
KBN promotor’s grant No.: 7 T11C 020 20, 2001-2002, 25.000,00 PLN Artificial intelligence (AI) techniques have enabled the construction of complex software systems, which are capable of solving difficult real-world problems for which more conventional programming techniques are insufficiently powerful. One of the most successful AI software technology has been knowledge-based systems (KBS), in which a relatively-simple inference engine uses a large domain-specific knowledge base to search for solutions to ill-structured problems. One impediment to the growing use of AI-based software is the lack of effective methods for ensuring the quality and reliability of this type of system. This has lead to a great deal of interest in techniques for the verification and validation of AI-based software. Much of this interest has focussed upon KBS. Current project concerns the theoretical and practical issues of verification of knowledge bases. Main goal of the project is developing own verification method and implementation of verification software. We present a dynamic verification method based on the decision unit concept. The aim of the system is to validate, on the basis of dynamic verification method, a set of rules saved in the knowledge base. We assume that this system operates on a set of rules previously specified by an expert. The system has been designed so that the user can decide what types of anomalies are to be checked and corrected.
|
|||||||||||||||||||||||||||||||||||||