Classification plays an important role in many fields of life, including medical diagnosis support. In the paper, fuzzy clustering algorithm dedicated to classification methods is proposed. Its goal is to find pairs of prototypes located near boundaries of both classes of objects. The minimization procedure of the proposed criterion function is described. The algorithm for determining the value of the clustering parameter is also presented. Presented results (synthetic dataset) confirm correctness of clustering – most of final prototypes, determined based on obtained pairs, are located between boundary of two classes.

The main aim of the article is to present the modifications of inference algorithms based on information extracted from large rule sets. The article introduces the conception of discovering the knowledge about rules saved in rule bases. It also describes the cluster analysis and decision units conception for this task and presents the optimization of forward and backward inference algorithms as well as selected experimental results.

The paper compares performance of Baldwin’s fuzzy reasoning based on a fuzzy truth value with the fastest available solutions. The analysis is important in order to locate areas where improvement of the first is the most significant. Potential fast approach based on the fuzzy truth value would be very interesting for many users applying fuzzy systems to solve problems involved with complex knowledge bases. Particularly, all research considering an analysis of genes employing DNA microarrays. Such methods very often generate rules with thousands of atomic premises. The most valuable advantage of Baldwin’s reasoning is preserving a fuzzy relation between a fact and a premise in the inference process, where other solutions, especially those commonly used, usually reduce it to only one value. Obtaining the method which, from computation time point of view, is comparable with common approaches but offers more advanced process of fuzzy reasoning, would be a significant achievement. The goal of this analysis is to prepare the future research considering development of Baldwin’s method, which computational complexity is comparable to simple, fast and widely used solutions like systems based on the approach of Mamdani and Assilan or Larsen.

In empirical sciences, among others – in medicine, domain data - stored in different repositories - are the most important source of domain information. There is a great number of methods, including semantic data integration, that enable to acquire domain knowledge from such data and express it in a convenient form. In the paper we propose a model for rules with uncertainty (2-uncertain rules) that can be obtained from somewhat heterogeneous data, written in a common format of tuples. The rules are uncertain implications, with complex premises and single conclusions, and two specific reliability factors. In addition, we propose functions for propagating uncertainty through reasoning chains in Rule-Based Systems (RBSs) with such rules in their knowledge base.

In the paper some examples of reasoning with 2-uncertain rules are presented. First of all, they will illustrate the method for designing 2-uncertain rules from medical aggregate data. The obtained rules compose the knowledge base of a medical Rule-Based System (RBS) aiding medical diagnosis and treatment. For each obtained rule two determined factors of rules’ reliability – global and internal ones – will rank it in the designed RBS. Furthermore, the presented examples will realize the influence of the reliability factors on the process of uncertain reasoning.

Large volumes of data are generated during DNA microarrays experiments. Database management systems (DBMS) are increasingly applied to these data, providing optimum processing and management from multiple microarray experiments. In this study, freely accessible DBMS software versions were compared (Microsoft SQL Server 2008 Express Edition, Oracle Database 10g Express Edition, DB2 Express-C 9.7.2, MySQL 5.1, and PostgreSQL 9.0). We examined them in the context of possible Entity-Attribute-Value (EAV) application as an optimal organization method for microarray data. It was confirmed in the comparative analysis of component data processing methods, consistent with the EAV model, that efficient methods for microarray data analysis are available in Microsoft SQL Server 2008 Express Edition and PostgreSQL 9.0 systems. Also, DNA microarray data processing was confirmed to be more efficient with Microsoft SQL Server 2008 Express Edition as compared with PostgreSQL 9.0. The EAV method was also shown to be suitable for use with open-source versions of DBMS software as an optimum storage model for DNA microarray data. In terms of data processing methods and performance, the Microsoft SQL Server 2008 Express Edition proved to be the best among compared database systems.

In medical decision making (e.g., classification) we expect that decision will be made effectively and reliably. Decision making systems with their ability to learn automatically seem to be very appropriate for performing such tasks. Decision trees provide high classification accuracy with simple representation of gathered knowledge. Those advantages cause that decision trees have been widely used in different areas of medical decision making. In this paper we present characteristic of univariate and multivariate decision tree. We apply those classifiers to the problem of acute abdominal pain diagnosis.

The paper presents an algorithm for induction of decision list from survival data. The algorithm uses a survival tree as the inner learner which is repeatedly executed in order to select the best rule at each iteration. The effectiveness of the algorithm was empirical tested for two implementations of survival trees on 15 benchmark datasets. The results show that proposed algorithm for survival decision list construction is able to induce more compact models than corresponding survival tree without the loss of the accuracy of predictions.

The two kinds of classifier based on the k-NN rule, the standard and the parallel version, were used for recognition of severity of ALS disease. In case of the second classifier version, feature selection was done separately for each pair of classes. The error rate, estimated by the leave one out method, was used as a criterion as for determination the optimum values of k’s as well as for feature selection. All features selected in this manner were used in the standard and in the parallel classifier based on k-NN rule. Furthermore, only for the verification purpose, the linear classifier was applied. For this kind of classifier the error rates were calculated by use the training set also as a testing one. The linear classifier was trained by the error correction algorithm with a modified stop condition. The data set concerned with the healthy subjects and patients with amyotrophic lateral sclerosis (ALS). The set of several biomarkers such as erythropoietin, matrix metalloproteinases and their tissue inhibitors measured in serum and cerebrospinal fluid (CSF) were treated as features. It was shown that CSF biomarkers were very sensitive for the ALS progress.

In the case of biomedical signals with a quasi-cyclic character, such as electrocardiographic signals, the high resolution electrocardiograms or electrical potentials recorded from the nervous system of patients (estimating brain activity evoked by a known stimulus), as a method of averaging in the time domain may be used for noise attenuation. In this paper there is presented input data partitioning applied to a few different methods of weighted averaging. This procedure usually leads to improve the quality of the resulting averaged signal, especially when fuzzy partitioning is used. Below it is presented the computational study of weighted averaging with traditional (sharp) and fuzzy partition of the input data in the presence of non-stationary noise. The performance of presented methods is experimentally evaluated for analytical signal of EN 60601-2-51 (2003), namely ANE20000 ECG record.

Many digital images, especially in biomedical fields, contain some disturbances. The image analysis depends on quality of the images that is why reduction or elimination (if it is possible) the disturbances is the key issue. There are many methods of improvement in the quality of the images and thus improve the quality of the image analysis, among them one of the simplest method is low-pass filtering such as arithmetic mean or its generalization, weighted mean. The basic problem of the weighted mean is the proper selection of the weights. This can be done using adaptive algorithms. This paper presents several such algorithms which are modifications of the existing weighted averaging methods created originally for noise reduction in electrocardiographic signal. The description of the new filtering methods and a few results of its application are also presented with comparison to existing arithmetic average filtering.

In the paper two dynamic ensemble selection (DES) systems are proposed. Both systems are based on a probabilistic model and utilize the concept of Randomized Reference Classifier (RRC) to determine the competence function of base classifiers. In the first system in the selection procedure of base classifiers the dynamic threshold of competence is applied. In the second DES system, selected classifiers are combined using weighted majority voting rule with continuous-valued outputs, where the weights are equal to the class-dependent competences. The performance of proposed MCSs were tested and compared against DES system with better-than-random selection rule using eleven databases taken from the UCI Machine Learning Repository. The experimental results clearly show the effectiveness of the proposed methods.

The article presents the method of building compact language model for speech recognition in devices with limited amount of memory. Most popularly used bigram word-based language models allow for highly accurate speech recognition but need large amount of memory to store, mainly due to the big number of word bigrams. The method proposed here ranks bigrams according to their importance in speech recognition and replaces explicit estimation of less important bigrams probabilities by probabilities derived from the class-based model. The class-based model is created by assigning words appearing in the corpus to classes corresponding to syntactic properties of words. The classes represent various combinations of part of speech inflectional features like number, case, tense, person etc. In order to maximally reduce the amount of memory necessary to store class-based model, a method that reduces the number of part-of-speech classes has been applied, that merges the classes appearing in stochastically similar contexts in the corpus. The experiments carried out with selected domains of medical speech show that the method allows for 75% reduction of model size without significant loss of speech recognition accuracy.

Contemporary medicine should provide high quality diagnostic services while at the same time remaining as comfortable as possible for a patient. Therefore novel non-invasive disease recognition methods are becoming one of the key issues in the health services domain. Analysis of data from such examinations opens an interdisciplinary bridge between the medical research and artificial intelligence. The paper presents application of machine learning techniques to biomedical data coming from indirect examination method of the liver fibrosis stage. Presented approach is based on a common set of non-invasive blood test results. The performance of four different compound machine learning algorithms, namely Bagging, Boosting, Random Forest and Random Subspaces, is examined and grid search method is used to find the best setting of their parameters. Extensive experimental investigations, carried out on a dataset collected by authors, show that automatic methods achieve a satisfactory level of the fibrosis level recognition and may be used as a real-time medical decision support system for this task.

This paper presents our new proposition of human body poses and gesture description methodology for Natural User Interfaces. Our approach is based on forward chaining inferring schema performed on the set of rules that are defined with formal LALR grammar. The set of rules is called Gesture Description Language (GDL) script while automated reasoning module with heap-like memory is a GDL interpreter. We have also implemented and tested our initial GDL specification and we have obtained very promising early experiments results.

Automatic disorder recognition in speech can be very helpful for the therapist while monitoring therapy progress of the patients with disordered speech. In this article we focus on prolongations. We analyze the signal using Continuous Wavelet Transform with 18 bark scales, we divide the result into vectors (using windowing) and then we pass such vectors into Kohonen network. Quite large search analysis was performed (5 variables were checked) during which, recognition above 90% was achieved. All the analysis was performed and the results were obtained using the authors’ program – “WaveBlaster”. It is very important that the recognition ratio above 90% was obtained by a fully automatic algorithm (without a teacher) from the continuous speech. The presented problem is part of our research aimed at creating an automatic prolongation recognition system.