Assisted living applications are commonly understood as technical environment for disabled or elderly people providing the care in the user-specific range. We are going to present the data capture methodology and design of a home care system for medical-based surveillance and man-machine communication. The proposed system consists of the video-based subject positioning, monitoring of the heart and brain electrical activity and eye tracking. The multimodal data are automatically interpreted and translated to tokens representing subject’s status or command. The circadian repetitive status time series (behavioral patterns) are a background for learning of the subject’s habits and for automatic detection of unusual behavior or emergency. Due to mutual compatibility of methods and data redundancy, the use of unified status description vouches for high reliability of the recognition despite the use of simplified measurements methods. This surveillance system is designed for everyday use in home care, by disabled or elderly people.

The methods of Computational Intelligence (CI) including a framework of Granular Computing, open promising research avenues in the realm of processing, analysis and interpretation of biomedical signals. Similarly, they augment the existing plethora of “classic” techniques of signal processing. CI comes as a highly synergistic environment in which learning abilities, knowledge representation, and global optimization mechanisms and this essential feature is of paramount interest when processing biomedical signals. We discuss the main technologies of Computational Intelligence (namely, neural networks, fuzzy sets, and evolutionary optimization), identify their focal points and elaborate on possible limitations, and stress an overall synergistic character, which ultimately gives rise to the highly symbiotic CI environment. The direct impact of the CI technology on ECG signal processing and classification is studied with a discussion on the main directions present in the literature. The design of information granules is elaborated on; their design realized on a basis of numeric data as well as pieces of domain knowledge is considered. Examples of the CI-based ECG signal processing problems are presented. We show how the concepts and algorithms of CI augment the existing classification methods used so far in the domain of ECG signal processing. A detailed construction of granular prototypes of ECG signals being more in rapport with the diversity of signals analyzed is discussed as well. ECG signals, Computational Intelligence, neurocomputing, fuzzy sets, information granules, Granular Computing, interpretation, classification, interpretability.

Electronic passports combine classical passport booklets with the smartcard technology, biometrics and cryptography. The communication with the electronic passports is based on contactless ISO 14443 technology, designed for the communication distance of 0-10 cm. This paper is focused on the privacy aspects of the electronic passports. Weaknesses of the basic access control and extended access control are discussed. Significant emphasis is put on passport fingerprinting which may allow guessing the issuing country. Aspects of biometric data formats, skimming, eavesdropping and active authentication challenge semantics are also covered. The conclusions sum up recommendations for passport holders and issuers.

Classification plays very important role in medical diagnosis. This paper presents fuzzy clustering method dedicated to classification algorithms. It focuses on two additional sub-methods modifying obtained clustering prototypes and leading to final prototypes, which are used for creating the classifier fuzzy if-then rules. The main goal of that work was to examine a performance of the classifier which uses such rules. Commonly used including medical benchmark databases were applied. In order to validate the results, each database was represented by 100 pairs of learning and testing subsets. The obtained classification quality was better in relation to the one of the best classifiers – Lagrangian SVM and suggests that presented clustering with additional sub-methods are appropriate to application to classification algorithms.

In the paper, we discuss nondeterministic rules in decision tables, called the second type nondeterministic rules. They have a few decisions values on the right hand side but on the left hand side only one attribute that has two values. We show that these kinds of rules can be used for improving the quality of classification. It is important in rule-based diagnosis support systems, where classification error can lead to serious consequences. The well known greedy strategy to construct the new nondeterministic rules, have been proposed. Additionally, based on deterministic and nondeterministic (second type) rules, classification algorithm with polynomial computational complexity has been developed. This rule-based classifier was tested on the group of decision tables, containing medical data, from the UCI Machine Learning Repository. The reported results of experiments showing that by combining rule-based classifier based on deterministic rules with second type nondeterministic rules give us possibility to improve the classification quality.

The article presents the application of the decision tree classifier to the acute abdominal pain diagnosis. The recognition task model is based on a decision tree. In this model the decision tree structure is given by the experts. For the assumed structure of the decision tree the locally optimal strategy is considered. The problem discussed in the work shows a selection of different classifiers (parameters) to the internal nodes of the decision tree. Experiments conducted for selected medical diagnosis problem shows that the use of different parameters for k-NN classification can improve the quality of classification in comparison with the situation if it is used with the same parameter for all internal nodes of the decision tree.

This paper addresses the problem of impulsive noise cancellation in digital signal area. The myriad and meridian filters are the type of robust filters which are very useful in suppressing the impulsive type of noise. The cost functions of theses filters have very similar structure. In this paper the generalized filter based on Lp norm is presented. The proposed filter operates in a wide range of impulsive noise due to the proper adjustment of p in the Lp norm. The presented filter is applied to suppress an impulsive noise in fetal heart rate (FHR) signal. Simulation results confirm the validity of the proposed filter.

Cardiotocography is the primary method for biophysical assessment of a fetal state. It is based mainly on the recording and analysis of fetal heart rate signal (FHR). Computer systems for fetal monitoring provide a quantitative description of FHR signals, however the effective methods for their qualitative assessment are still needed. The measurements of hydronium ions concentration (pH) in newborn cord blood is considered as the objective indicator of the fetal state. Improper pH level is a symptom of acidemia being the result of fetal hypoxia. The paper proposes a two-step analysis of signals allowing for effective prediction of the acidemia risk. The first step consists in the fuzzy classification of FHR signals. The task of fuzzy inference is to indicate signals that according to the FIGO guidelines represent the fetal wellbeing. These recordings are eliminated from the further classification with Lagrangian Support Vector Machines. The proposed procedure was evaluated using data collected with computerized fetal surveillance system. The classification results confirmed the high quality of the proposed fuzzy method of fetal state evaluation.

The paper presents an approach to discrimination of textures in radiological images based on multi-aspect similarity measures composed of logical tests. There are formulated basis assumptions for similarity measures which can be composed by products of partial (single-aspect) similarity measures. On the basis of similarity measures -similarity classes are defined. Next, two types: strong and weak similarity measures are defined. It is shown that they make possible to define similarity measures based on quality objects properties as well as on their numerical parameters. As an example of application of the general concept discrimination of normal and ill (lesions affected) tissues is considered. It is illustrated by analysis of USG images of liver tissues for which morphological spectra and their statistical parameters have been calculated. It is shown that the differences between values of some pairs of corresponding parameters can be used to a construction of an effective algorithm of textures discrimination. This algorithm takes into consideration both, numerical features of the texture samples and some qualitative data concerning the patients. Conclusions are formulated at the end of the paper.

Multiple classifier systems are currently the focus of intense research. In this conceptual approach, the main effort focuses on establishing decision on the basis of a set of individual classifiers’ outputs. This approach is well known but usually most of propositions do not take exploitation cost of such a classifier under consideration. The paper deals with the problem how to take a test acquisition cost during classification task under the framework of combined approach on board. The problem is known as cost-sensitive classification and it has been usually considered for the decision tree induction. In this work we adapt mentioned above idea into choosing members of classifier ensemble and propose a method of choosing a pool of individual classifiers which take into consideration on the one hand quality of ensemble on the other hand cost of classification. Properties of mentioned concept are established during computer experiments conducted on chosen medical benchmark databases from UCI Machine Learning Repository.

Classification methods can be divided into supervised and unsupervised methods. The supervised classifier requires a training set for the classifier parameter estimation. In the case of absence of a training set, the popular classifiers (e.g. K-Nearest Neighbors) can not be used. The clustering methods are considered as unsupervised classification methods. This paper presents an idea of the unsupervised classification with the popular classifiers. The fuzzy clustering method is used to create a learning set. The learning set includes only these patterns that are the best representative of each class in the input dataset. The numerical experiment uses an artificial dataset as well as the medical datasets (PIMA, Wisconsin Breast Cancer) and illustrates the usefulness of the proposed method.

In the paper, the method of acoustic model complexity level selection for automatic speech recognition is proposed. Selection of the appropriate model complexity affects significantly the accuracy of speech recognition. For this reason the selection of the appropriate complexity level is crucial for practical speech recognition applications, where end user effort related to the implementation of speech recognition system is important. We investigated the correlation between speech recognition accuracy and two popular information criteria used in statistical model evaluation: Bayesian Information Criterion and Akaike Information Criterion computed for applied acoustic models. Experiments carried out for language models related to general medicine texts and radiology diagnostic reporting in CT and MR showed strong correlation of speech recognition accuracy and BIC criterion. Using this dependency, the procedure of Gaussian mixture count selection for acoustic model was proposed. Application of this procedure makes it possible to create the acoustic model maximizing the speech recognition accuracy without additional computational costs related to alternative cross-validation approach and without reduction of training set size, which is unavoidable in the case of cross-validation approach.

Automatic disorders recognition in speech can be very helpful for therapist while monitoring therapy progress of patients with disordered speech. This article is focused on sound repetitions. The signal is analyzed using Continuous Wavelet Transform with 16 bark scales, the result is divided into vectors and passed into Kohonen network. Finally, the Kohonen winning neuron result is put on the 3-layer perceptron. The recognition ratio was increased by about 20% by adding a modification into the Kohonen network training process as well as into CWT computation algorithm. All the analysis was performed and the results were obtained using the authors’ program “WaveBlaster”. The problem presented in this article is a part of our research work aimed at creating an automatic disordered speech recognition system.

In this paper a new model of the vocal tract is proposed. It is based on elliptical cylinders. It uses the vocal tract model based on PARCOR coefficients and midsaggital measurements of the voice tube. PARCOR coefficients were obtained from linear prediction coefficients which had been obtained by Levinson-Durbin method. Midsaggital lengths, understood as the height of a real vocal tract, were taken from X-Ray pictures, and they were averaged from the vocal tracts of a few people, who uttered the same vowels. The paper bases on Polish vowels: a,e,o,u,i,y.

The therapy of stuttering people is based on a proper selection of texts and then on a practice of their articulation by reading or narration. The texts are chosen on the basis of kind and intensity of dysfluencies appearing in a speech. Thus there is still a requirement to find effective and objective methods of analysis of dysfluent speech. Hidden Markov models are stochastic models widely used in recognition of any patterns appearing in a signal. In the work a simple monophone system based on the Hidden Markov Model Toolkit was built and tested in the context of detection and classification of phoneme repetitions – a common speech disorder in the Polish language.

The system using steerable filters for analysis suspicious lesions in mammograms is proposed. This system is based on moments and texture features. The set of well defined and classified suspicious lesions regions from mammograms database are used as a reference pattern. The similarity measure for reference pattern image and patient mammogram is found by computing the distance between their corresponding feature vectors. The Euclidean distance metric is used to finding the nearest class to patient feature vector what in result mark the automatically classify this mammograms.

The effect of pleasant and unpleasant odor on human physiological state was preliminary investigated by assessing salivary substances. A term “pleasant” (or unpleasant) is rather showing one’s subjective and thus mental state meanwhile the odor itself should give an impact to both human mind and body. This study was aimed at investigating an effect of pleasant/unpleasant odors on human body and mind. In the experiment the odor of isovaleric acid, as an unpleasant odor, and Japanese cypress, as a pleasant one, were exposed to subjects for 18 minutes. Salivary immune substance (Immunoglobulin A: IgA) and glucocorticoid (cortisol) were determined every three minutes as for an indices of the impact of the odors on human immune and endocrine secretion. As a developing result, IgA showed an increase by isovaleric odor whilst cortisol decreased, even though the concentration of that unpleasant odor was too low to be aware of. These changes in the secretion of the substances were discussed in relation with human physiological stress reaction mechanism.

Cardiovascular mortality remains a leading health and social problem in many countries throughout the world. Its main cause is related to atherosclerosis of coronary and cerebral vessels with their most severe consequences: heart attack and stroke. Therefore, it is obvious that current preventive measures include early detection of atherosclerosis process. Multi-detector computed tomography (MDCT) is one of imaging modalities allowing for noninvasive detection of atherosclerotic lesion within coronary arteries in subjects with accumulation of risk factors (smoking, high lipids, hypertension, male gender, family history) or with suspicion of coronary artery disease (CAD). In is very important that the tomographic images are taken in synchronization with cardiac cycle so that, during few heartbeats, an appropriate series of images can be recorded. Commonly, cardiac MDCT is used for visualization of cardiac and vessels morphology. Heart function can also be determined, however, this MDCT potential is only rarely applied, as current echocardiographic modalities are sufficient. Functional analysis of coronary arteries (flow, reserve) is usually approached by means of invasive procedures. We aimed at finding solution for evaluation of another kind of functional analysis of coronary arteries, namely vessel’s wall compliance by means of MDCT coronary angiography. Under the proposed procedure, on basis of serial CT images of the vessels over entire cardiac cycles, the internal area of the blood vessel is measured and its changes during various phases of heartbeat (systole, diastole) are calculated. If the vessel wall has been changed by atherosclerotic plaque, either calcified or non-calcified, then its compliance will be reduced due to its stiffness. Calculation of coronary artery compliance requires a series of measurements, which is unreliable and impractical for doing manually. One component of the method described herein involves the images being converted into binary representations and the Hough Transform then applied. The overall methodology proposed in this paper assists in the preparation of a medical diagnosis.

In this paper two multiresolution transforms (discrete 2D wavelets and complex wavelets) are compared for their capabilities to enhance local texture orientation of mammograms. The local orientation of image texture is useful feature to detect one of the typical types of abnormal findings in mammography - architectural distortions. Our research was directed to define an effective, more reliable directional model of local directional findings in mammograms. Computer-aided diagnosis was considered as a concept of accurate model application.

This work presents some conceptual improvements in assistance of acute stroke diagnosis with Stroke Monitor – computer-aided diagnosis tool developed and elaborated by Telemedicine Group from Institute of Radioelectronics, Warsaw University of Technology. Based on statistical analysis of common error sources we proposed some ideas of improvement capabilities for false positive errors reduction. Simulation and experimental verification confirmed validity of further development directions.

The paper describes decomposition of gray medical images. Pixels of the image are assigned with the variable values derived from a neighbourhood of the pixel. Then Grade Correspondence Cluster Analysis is used to order set of pixels according to their grade differentiation and to divide pixels into subsets. Subsets are visualized in separate subimages and regions are extracted on principle of spatial neighbourhood in subimage. Influence of a number of subimages is discussed. Then a new grade-spatial procedure is proposed which combines features of grade similarity and spatial neighbourhoods.

Liver diseases and more specifically viral hepatitis are at the center of interest due to their global spreading, even in the most developed countries. The range of symptoms, the complications and the course of the disease have imposed the operation of liver centers at the outpatients’ departments of hospitals, where the contribution of several specialized doctors the disease is diagnosed, prevented and treated. Many patients suffer from hepatitis without knowing it either because they manifest no symptoms or because the infection is not traced through the usual lab tests. This paper focuses on studying and proving how the systematic reading of the main liver diseases and the methods through which the doctor makes the diagnosis can help the study and analysis of a series of steps that have to be followed in order to treat the disease. Then, the use of a modern information system using Medical Informatics technologies is proposed so as both the task of diagnosis and the efforts to treat and overcome the problems related to the liver disease to be supported.

The application of new technologies in health units suggests the materialization of an advanced Electronic Health Record as a central axis in medical information management. The aim of present study is to point out the position Information and Communications Technology has in the health field and how it can play a major role in the materialization of an advanced Electronic Health Record, offering major filing capacity in small volume, while strong huge medical data, based on international standards which are directly accessible and processable.

In this paper a simple and non-expensive indirect fibrosis stage prediction method is described. Presented method is non-invasive and is based on the results of the generic blood tests. The method is based on a statistical analysis of wide range of blood tests results supported with the experience of hepatologists.

Cancer is one of the diseases which cause the highest death rate in XXI century. However, through years techniques and equipment used to fight with this disease improved there are still many opportunity to get better results. Nowadays, the attention is focused on precise cancerous cells place estimation. It is important for effective cancer treatment, but also it allows diminishing the unwanted effects as destruction of healthy cells. One of the promising techniques in obtaining better spatial and temporal resolution of the internal of the human body is the photoacoustic imaging. Combination of the acoustics, ultrasounds or microwaves by the set of ultrasonic detectors can lead to estimation of the localization of the sources of the waves. The work presents the utilization of the multiple signal classification (MUSIC) algorithm in estimation of angle and distance of microwave or photoacoustic waves sources. The technique presented in the work allows for better prediction of localization of the sources of the detected waves by the sensors array. The proposed application of this technique is detection of sentinel lymph nodes.

This article concerns the analysis of corneal endothelial image. The basic problems of binarization and segmentation of these images are discussed. Preprocessing methods are proposed, consisting of median and convolution filtration, to remove noise. An algorithm of normalization of the average brightness of the vertical and horizontal is presented. The problem of binarization is discussed. At the end the proposal of segmentation algorithm is reported.

Measurement of volumetric tomographic data, similar to other measurement techniques, suffers from several classes of artifacts, of which noise presence and the partial volume effect belong to the most prominent ones. These artifacts spoil data analysis and/or visualization, which may, for example in the case of medical imaging, lead to erroneous decisions with severe consequences. We propose a set of tools for region-based processing of volumetric data. Here, the basic entity is a spectrally homogeneous region instead of the traditional voxel. This provides for, on the one hand, higher robustness and, on the other hand, speeds up processing owing to many times smaller amount of elements to work with. Homogeneous regions are in our approach detected by the well-known segmentation by means of the watershed transform. In this paper we present algorithms for streamed computation of watershed transform, which allows for processing of very large data, region-based data smoothing and region merging based on the spectral distance. Further, we present an interactive tool for volume data segmentation and visualization which takes advantage of region hierarchies obtained by a hierarchical watershed transform.

This paper presents an automatic computer system to breast cancer diagnosis. System was designed to distinguish benign from malignant tumors based on fine needle biopsy microscope images. Studies conducted focus on two different problems, the first concern the extraction of morphometric and colorimetric parameters of nuclei from cytological images and the other concentrate on breast cancer classification. In order to extract the nuclei features, segmentation procedure that integrates results of adaptive thresholding and Gaussian mixture clustering was implemented. Next, tumors were classified using four different classification methods: k-nearest neighbors, naive Bayes, decision trees and classifiers ensemble. Diagnostic accuracy obtained for conducted experiments varies according to different classification methods and fluctuates up to 98% for quasi optimal subset of features. All computational experiments were carried out using microscope images collected from 25 benign and 25 malignant lesions cases.

Active region models are methods for automatic image segmentation. The models are able to detect shapes of irregular borders. In the present paper, the method is examined using medical images of liver changed locally by cancer cells.

Article presents fast method of the face reconstruction for forensic and archeology science developed by authors which is based on a deformation of the known skin aligned to skull. As a guide to deformation a position of the well known anatomical landmarks is used. Calculation of the new position of the vertices of a triangle mesh of the model presented is a result of the inverse distance weighted interpolation. To improve precision usage of the skin thickness values from anthropological tables is used. As a example of the available results reconstruction for archaeological investigation is presented.

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is a relatively new, promising technique for breast cancer diagnostics. A few series of images of the same body region are rapidly acquired before, during and after injection of paramagnetic contrast agent. Propagation of the contrast agent causes modification of MR signal over time. Its analysis provides information on tissue properties, including tumour status, that is not available with the regular MRI. Unintentional patient’s movements during the examination result with incorrect alignment of the consecutive image series. Their analysis is then difficult, inaccurate or even impossible. The purpose of this work is to design a registration scheme that could be applied to solve the problem in a routine manner, in standard hospital conditions. The proposed registration framework, composed of B-spline transformation, mean squares metric and LBFGSB optimizer, is able to produce satisfactory results within reasonable time.

In this paper a segmentation algorithm of carotid arteries on computed tomography angiography (CTA) images is proposed. The algorithm is based on the threshold level set approach. In the basic version, the algorithm analyzes CTA slices beginning at the brachiocephalic trunk and going towards carotid arteries. Second variant of the algorithm performs segmentation in the opposite direction, which implies that the algorithm can follow branches e.g. subclavian arteries. The localization process of the initial contour, for threshold level set method, on the first slice is based on curvature anisotropic diffusion filter, the Gaussian filter and fast marching method. The article contains segmentation results for tested sets of method parameters. Experimental results show that optimal set of parameters ensuring that the threshold level set method performs segmentation of the entire subclavian arteries, does not exist.

Information about blood pressure at the inflow and outflow connectors as well as inside a Ventricular Assist Device (VAD) supplemented with information about pressures inside the pneumatic part enables to adjust operational parameters of the VAD in the optimum manner. Practical implementation of a method that makes it possible to measure blood pressure at plurality of points is a really sophisticated task in terms of technical and technological issues. On one hand it is mandatory to assure appropriate metrological properties of the entire measurement path, on the other hand the measuring transducers must be reliably separated from blood. Internal surfaces of these VAD parts that come in direct contact with blood must be smooth and uniform; it is extremely essential due to a risk of blood coagulation on any unevenness of surfaces. The paper presets the solution for measurements of blood pressure inside the VAD, where the suggested solution meets the assumed metrological criteria as well as very stringent safety requirements.

Research was conducted to analyze the flow of a fluid within mechanical heart valve - Medtronic Hall. Physical experiment and numerical modelling were performed. The aim of the research was to determine the difference between obtained experimental and numerical data. In the experiment a dependency between static flow rate within the valve and static inlet and outlet pressure in the valve duct was examined. Moreover a dependency between static flow rate and angular valve position was also determined. Experimental data was used to perform a numerical flow analysis. The obtained flow rate values and angular positions of the valve were set to a finite-volumes-method model in order to achieve model output pressure values identical or similar to the ones obtained from the experiment. The resulting pressure values from the experiment and numerical analyses proved to be of the same order of magnitude, varying only by up to 10%. However, as far as differential pressure is concerned, numerical results were out of the range of measurement resolution. It can be assumed that numerical flow analyses quite correctly predict the real phenomenon and in view of measurement inaccuracy of used sensors authors would suggest using more accurate ones and repeating measurements for future clarification.

The problem raised in this article is the selection of the most important components from multispectral images for the purpose of skin tumor tissue detection. It occured that 21 channel spectrum makes it possible to separate healthy and tumor regions almost perfectly. The disadvantage of this method is the duration of single picture acquisition because this process requires to keep the device very stable. In the paper two approaches to the problem are presented: hill climbing strategy and some ranking methods.

The CT2TEE is an online transoesophageal echocardiography (TEE) simulator [2] developed on the AGH University of Science and Technology. As a basis for the displayed real-time simulation it uses 3D models created by hand from the CT images. The aim of this paper is to present algorithms developed specifically for efficient editing and preparation of three-dimensional models of the heart for use in the CT2TEE simulator. A detailed description of proposed algorithms, their advantages and limitations, is provided.

The paper concerns estimation of significance of differences of mutagenesis level between the wild-type strain (wt) and its derivatives which differ in DNA repair ability, namely alkA and alkB strain, devoided AlkA glycosylase and AlkB dioxygenase activity, respectively. The strains were analyzed for their ability to repair 1,N6-ethenoadenine (εA) – chloroacetaldehyde adduct to DNA. The analysis was done using classical statistical and pattern recognition methods. The obtained results confirmed that AlkB dioxygenase plays the most important role in εA repair in E. coli in the experimental modeling.