The way to model complex shapes has a significant influence depending on the context. Handling an object can be considerably increased if a good underlying model is used. On the contrary, preponderant problems can appear if an unsuited model is associated to the object. The main criterion to discriminate existing models is to determine the balance between: their ability to control global characteristics and the possibility to handle local features of the shape. The fact is very few models are adapted both to structure and to geometrical modelling. In this paper, we first describe an overview of existing approaches. They can be classified principally in two groups: skeleton based models, used to control the global aspect of the shape, and free form models, used to control local specificities of the object. Then, trying to keep the advantages of both techniques in mind, we present an original approach based on a multi-layer model to represent a 3D object. We focus on the ability to take into account both global and local characteristics of a complex shape, on topological and morphological levels, as well as on the geometric level. To do that, the proposed model is composed of three layers. We call the boundary mesh the external layer, including a multi-resolution feature. We enhance this representation by adding an internal structure: the inner skeleton, which is topologically equivalent to the input object. In addition to that, a third layer links the structural entity and the geometrical crust, to induce an intermediary level of representation. This approach is applied to classical and medical data through a specific algorithm.

The paper presents the results of experiments concerning flow in the model of cerebral supplying arteries and the Circle of Willis (CW). Vascular phantom was prepared on the basis of anatomical specimens. The most typical artery shapes and dimensions were considered. Pressure distribution in 6 characteristic points is provided, and so are the average flow rates in the anterior, middle and posterior section of the brain. Tests were run in the conditions replicating the physiological state (i.e. when the supplying arteries were fully patent) and in pathological conditions, in which the internal carotid and vertebral arteries were occluded on one or both sides. Thus obtained results were compared with the results of computer simulations based on linear and non-linear flow models. To estimate the non-linear resistance of vascular segment two phenomenological formulae were proposed. High degree of correlation between the values obtained from experiments and those registered in non-linear computer model proves usefulness of proposed formulae. It verifies the hypothesis that nonlinearity of flow characteristics of the vessel segments is to a great extent caused by their tortuous and small length in relation to diameter. Non-linear effects are particularly pronounced in conditions of pathological occlusion of supplying vessels.

In the paper the method called CGA based on a cooperating genetic algorithm is presented. The CGA is developed for searching a set of rules describing classes in classification problems on the basis of training examples. The details of the method, such as a schema of coding (a chromosome), and a fitness function are shortly described. The method is independent of the type of attributes and it allows choosing different evaluation functions. Developed method was tested using different benchmark data sets. Next, in order to evaluate the efficiency of CGA, it was tested using the Breast Cancer data set with 10 fold cross validation technique.

While a genuine abundance of biomedical data available nowadays becomes a genuine blessing, it also posses a lot of challenges. The two fundamental and commonly occurring directions in data analysis deal with its supervised or unsupervised pursuits. Our conjecture is that in the area of biomedical data processing and understanding where we encounter a genuine diversity of patterns, problem descriptions and design objectives, this type of dichotomy is neither ideal nor the most productive. In particular, the limitations of such taxonomy become profoundly evident in the context of unsupervised learning. Clustering (being usually regarded as a synonym of unsupervised data analysis) is aimed at determining a structure in a data set by optimizing a given partition criterion. In this sense, a structure emerges (becomes formed) without a direct intervention of the user. While the underlying concept looks appealing, there are numerous sources of domain knowledge that could be effectively incorporated into clustering mechanisms and subsequently help navigate throughout large data spaces. In unsupervised learning, this unified treatment of data and domain knowledge leads to the general concept of what could be coined as knowledge-based clustering. In this study, we discuss the underlying principles of this paradigm and present its various methodological and algorithmic facets. In particular, we elaborate on the main issues of incorporating domain knowledge into the clustering environment such as (a) partial labelling, (b) referential labelling (including proximity and entropy constraints), (c) usage of conditional (navigational) variables, (d) exploitation of external structure. Presented are also concepts of stepwise clustering in which the structure of data is revealed via a series of refinements of existing domain granular information.

The paper shows several aspects of the gait data record analysis describing neurological diseases. The diagnosis of the gait abnormalities concerns interferences level of the patient physiological records. The disease source and level can be classified by the relevant interference functions. These functions were used for artificial records creation to multiply the necessary set of data needed for neural network training.

Almost unlimited possibilities of sharing neuroinformatics resources, opened by the Internet, create an almost unlimited number of issues. Growing amount of available data, combined with the lack of reliable and large enough metainformation resources, limits the proliferation and reliability of this media. In this paper we propose a solution, which may help in an efficient sharing of neuroinformatics resources, by means of a network of vortals dedicated to particular and well defined topics. These vortals are responsible for collection of high quality resources in their particular fields. They are interconnected in a way transparent to the user, using a low level interface for interchanging queries. For a user this means that a query entered in one of the vortals will return relevant results found also in the other vortals of the Network. We also describe technical details and pilot implementation; metainformation is based upon Open Archives/ Dublin Core standards, and interchange of queries on XML/SOAP.

In the paper, evaluation of two approaches to modelling of hemodialysis is performed. Results obtained by regression are compared to those generated by neural models. Differences in the modelling quality are small. Both models shown the same qualitative dependencies between analyzed parameters.

In this article was written attempt to use the eye movement to biometrics identification. There are few words about theory of biometrics identification. After that there is described the way of recording of data. It is made by system "Ober2". To process collected data the algorithms of artificial neural network are used. For this need, dedicated application was written. Functionality of the application, in article was described, as well as the examples - results of working of the artificial neural networks, for chosen criterions of researches. The plans for future researches were placed at the end of the article.

The problem of medical teleconsultations with intelligent computer system rather than with a human expert is analyzed. System for content-based retrieval of images is described and presented as a use case of telemedical tool. Selected features, crucial for retrieval quality, are introduced including: synthesis of parametric images, regions of interest detection and extraction, definition of content-based features, generation of descriptors, query algebra, system architecture and performance. Additionally, electronic business pattern is proposed to generalize teleconsultation services like content-based retrieval systems.

To-date research in the area of applied medical artificial intelligence systems suggests that it is necessary to focus further on the characteristic requirements of this research field. One of those requirements is related to the need for effective analysis of multidimensional heterogeneous data sets, which poses particular difficulties when considering AI-suggested solutions. Recent works point to the possibility of extending the activation function of a perception to the time domain, thus significantly enhancing the capabilities of neural networks. This change results in the ability to dynamically tune the size of the decision space under consideration, which stems from continuous adaptation of the interneuron connection architecture to the data being classified. Such adaptation reflects the importance of individual decision attributes for the patterns being classified, as defined by the Sigma-if network during its training phase. These characteristics enable effective employment of such networks in solving classification problems, which emerge in medical sciences. The described approach is also a novel, interesting area of neural network research. This article discusses selected aspects of construction as well as training of Sigma-if networks, based on a sample problem of classifying Arabic numeral images.

The essence of the recruitment and selection in sports depends on determining an aptitude vector of the candidate to sport training. For this reason, the recruitment process may be optimized by determining possibly large amount of information on the sport level of the candidate with as small as possible number of tested characteristics, using a mathematical model based on neural networks. The main aim of this work was verification of the usefulness of neural models in optimization of the process of recruitment process, both at sprint distance of 50 m and typically endurance distance of 800 m. The material for the investigation was a group of 80 young swimmers in the youngster and junior category from the Silesian macro-region. For the purpose of verification of the usefulness of neural models, statistical analysis were made of the measurement results of young swimmers and two neural models were developed - for sprint distance (50 m) and endurance distance (800 m). The developed models, based on the architecture of perception networks, has shown capability of generalization and prediction, which has enabled to reach conclusions on practical possibility of using neural networks in optimization of the recruitment and selection process for sport swimming.

Scheduling doctors duties in a hospital are complicated and time-consuming tasks. The person responsible for creating a duty timetable is facing one major problem when allocating doctors to time periods: the agreement between several constraining (and often mutually excluding) requirements must be found. In this paper a solution methodology for the monthly duty assignment of doctors is presented. The typical problem is described in detail, along with specific hospital environment, from which datasets for experiments have been taken. A hybrid approach that utilizes strengths of a few artificial intelligence techniques was used to solve the problem. In particular, a population of initial solutions is generated heuristically and then improved using evolutionary algorithm. Experimental results are presented along with a discussion on the computational efficiency, operational acceptability and quality of the solutions.

The article discusses the main problems connected to the specificity of medical aspects, especially as concerns the quality and means of selection of data and tools used for constructing classification systems. Special attention is devoted to the risks inherent in direct application of classical knowledge extraction algorithms (such as the algorithms for constructing decision trees) to medical data. The article describes some attempts at solving emerging problems and points to the need for analysis of classifiers with regard to more than just their potential redundancy and mutual exclusion. The article also proposes two functions, useful for analysing rule sets with focus on data semantics.

This work presents the possibility of measuring heart action parameters in the alpine conditions as well as the measuring system, which helps to measure and transmit EKG signal with the aid of GPRS net. The GPRS transmission makes the estimation of heart action on - line possible. Thanks to it, it is possible to monitor the reaction of human organism being on a certain level of height. GPRS net application enables the measuring in the whole world. The charge for using this net is only dependent on the quantity of transmitted data which needs the signal compression. This work also mentions the possibility of compression use, which takes the advantage of temporal and frequency decomposition.

The value of each state variable (physical condition, sport scores) at the end of certain period is a function of general state of the athlete, his scores described at the beginning of this period of time and adapted training, that is implemented intensity of separate training means (controlling variables) in the analyzed time period. The aim of this work was to determine optimal values of control in swimmer training, targeted at achieving the best score in 25 m and 800 m swimming. For this purpose a model was developed, beginning with its shape (it has been defined, what is state variable and what is control, how the state variables and control variables influence the increase) to determining numerical values of all parameters. The basis for the development of the detailed model was pedagogic experiment conducted on the group of 14-year old boys. Conclusions were drawn on the practical possibility of using a mathematical model in control of sport training.

In this paper the problem of nonparametric impulsive noise removal in multichannel images is addressed. The proposed filter class is based on the nonparametric estimation of the density probability function in a sliding filter window. The obtained results show good noise removal capabilities and excellent structure preserving properties of the new impulsive noise removal technique.

A new approach in the analysis of fundus images performed in frequency domain is described. It is related to specific features of the power spectrum and refers to the notion of spectral fractal dimension as well. Preliminary calculations have been carried out for sample series of diagnosed fundus eye images. The results and proposals for further improvements of the method are discussed.

Airways remodelling is currently described as a process occurring before asthma becomes clinically manifest, which is confirmed by biopsy studies. The aim of this study was to test and validate image analysis methods to describe the changes such as peripheral airways remodelling in HRCT readings. Different methods of airways extraction from HRCT images were investigated including: manual identification of an airway region major axes on original and scaled images (using different interpolation techniques like pixel resize, bilinear interpolation and cubic convolution), manual extraction of the density profile through the major axes of an airway region, semi-automatic method using active contours and the Hough transform. Methods were tested with original images and artificially modified images by blurring and noise addition (Gaussian, Laplacian and salt-and-pepper). Results suggest that popular image magnification using cubic convolution is not suitable for accurate estimation of shape properties of small regions. Smart pixel resizing enables to delineate a region of inner and outer borders with subpixel accuracy reducing the total error of the wall thickness estimation. Additionally smoothing must be reduced to the minimum in the case of an active contour application.

Recent investigations in neuropsychology and psychology of vision have proven that human eye does not get all the information from the surrounding world in the same degree. There are three classes of signals received by human brain. The more important one is the information about features such as corners, junctions, ends of lines, etc. Straight lines and edges are the second in the hierarchy of importance. And the last ones are textures they support the less important information about objects. Basing on these results, in image processing, theory of intrinsic dimensionality and related to it theory of feature extractors have been established. In the paper a survey of approaches that are used for construction of feature extractors based on intrinsic dimensionality have been presented. To carry out experiments the approach based on geometrical wavelets has been chosen and the software prepared by the first author has been used. Experiments presented in the paper have been performed on relatively complex images that had been faces" images. They confirmed that the information about the basic elements of faces (eyes, nose, lips, etc.) might be properly extracted from the face with the usage of the feature extractor. Moreover, the experiments have shown that in this way one could obtain the smallest possible amount of information, which was enough that human eyes yet have seen the face. Very promising results of experiments suggest that it is possible to use the proposed approach to face identification and recognition. Also some possible medical applications have been suggested.

In this paper we present classification of the thermal images in order to discriminate healthy and pathological cases during breast cancer screening. Different image features and approaches for data reduction and classification have been used to distinguish healthy breast one with malignant tumour. We use image histogram and co-occurrence matrix to get thermal signatures and analyze symmetry between left and right side. The most promised method was based on wavelet transformation and nonlinear neural network classifier. The proposed approach was used in the pilot investigations in the medical centre which is permanently using thermograph for breast cancer screening, as an adjacent method for other classical diagnostic method, such as mammography.

This paper reports on our experiences using datasets from the Visible Human Project in different biomedical applications. Introduced 1994 by the US National Library of Medicine the digitized multimodal anatomical datasets of the Visible Man have challenged the worldwide scientific community. A significant response to this challenge from several interdisciplinary research teams has emerged as a new area of research. This area requires close interaction and collaboration among anatomists, radiologists, computer scientists, mathematicians, engineers and physicians. The digitized volumetric images of the human body have been applied not only for the computer-aided exploration of the human gross anatomy, but also as structural input for the therapy planning and simulation systems. The importance of such virtual patient model is becoming increasingly recognized in modern medicine. To effectively use these specific datasets a sophisticated framework consisting of image processing, computer graphics and mathematical modelling methods is required. In this work various aspects of the developed framework are presented and discussed. Some preliminary results of our biomedical simulations are presented.

Segmentation and visualisation of anatomical regions of the brain are fundamental problems in medical image analysis. In this paper, we present a fuzzy-logic segmentation system that is capable of segmenting magnetic resonance (MR) images of a human brain. The presented method consists of two main stages: histogram thresholding and pixel classification using a rule-based fuzzy logic inference. After the segmentation is complete, attributes of different tissue classes may be determined (e.g., volumes), or the classes may be visualised as spatial objects. The implemented system provides many advanced 3D imaging tools.

In this paper we introduce the f3d format for storage of volumetric data together with a suite of tools for processing, segmentation and visualization of such data. Both the format and tools were developed for a highly variable and rapidly evolving academic environment, where new data processing and visualization tasks emerge very often. The tools address all the steps of a volume visualization pipeline: starting with import of external formats, over preprocessing, filtering, segmentation to interactive visualization.

This paper presents a methodology that addresses important issues concerned with the optimization of the misregistration measures in the volumetric medical data registration problem. Our registration framework uses robust simulated annealing method to handle multiple local minima of the cost function. Our efforts have been centred on obtaining a reliable, efficient and generally applicable method to solve such optimization problems. This has been accomplished through developing an adaptive cooling schedule for the simulated annealing method. The proposed method is very reliable for the estimation of the global minimum in the optimization of objective functions with highly differentiated search space landscapes. We present the detailed description of the method as well as discussion of its advantages and disadvantages.