2/2004 Computers in Medical Applications

Contents:

• A.Przelaskowski, The JPEG2000 Standard for Medical Image Applications - abstract | full text
• P.Kostka and E.Tkacz, Wavelet-neural Systems as Approximators of an Unknown Function - a Comparison of Biomedical Signal Classifiers - abstract | full text
• M.W.Kurzynski, Fuzzy Reasoning Applied to Multistage Diagnosis of Acute Renal Failure in Children - abstract | full text
• L.Bobrowski, Feature Selection Based on Linear Separability and a CPL Criterion Function - abstract | full text
• T.Przybyla, Breast Cancer Diagnosis via Fuzzy Clustering with Partial Supervision - abstract | full text
• T.Pander, Application of Weighted Myriad Filters to Suppress Impulsive Noise in Biomedical Signals - abstract | full text
• R.Burduk, Pattern Recognition of Sacroileitis with the Use of Multistage Logic with a Fuzzy Loss Function - abstract | full text
• M.Kretowski and J.Bezy-Wendling, Computer Modelling of Vascular Systems - abstract | full text
• L.Witkowski, An Automatic System for Calculating Basic Semen Parameters - abstract | full text
• J.Zielinski and H.Krawczyk, Achieving High Dependability of an Endoscopy Recommender System (ERS) - abstract | full text
• J.W.Grzymala-Busse, Z.S.Hippe, M.Knap and T.Mroczek, A New Algorithm for Generation of Decision Trees - abstract | full text
• P.Hoser, A Mathematical Model of the Left Ventricle Surface and a Program for Visualization and Analysis of Cardiac Ventricle Functioning - abstract | full text
• M.Kaczmarek and A.Nowakowski, Active Dynamic Thermography in Mammography - abstract | full text
• K.Krysztoforski, A.Wolczowski, R.Bedzinski and K.Helt, Recognition of Palm Finger Movements on the Basis of EMG Signals with the Application of Wavelets - abstract | full text
• K.Stanisz-Wallis, A.Izworski, A.Dembinska-Kiec, R.Tadeusiewicz and T.Lech, Assessment of Diagnostic Features in the Coronary Artery Disease (CAD) by Application of Statistical Methods and Neural Networks - abstract | full text
• M.Grabowski, K.J.Filipiak, R.Rudowski and G.Opolski, An Expert System Supporting Risk Stratification in Acute Coronary Syndromes - abstract | full text
• A.Kluza, Veterinary Toxicology Information System - abstract | full text
• A.Przelaskowski, Irreversible Medical Image Compression: Conditions of Acceptability - abstract | full text

Abstracts:

• A.Przelaskowski, The JPEG2000 Standard for Medical Image Applications

  A new standard of still image compression is characterised in the context of medical applications. Wide spectrum of JPEG2000 features is analysed with respect to its application potential to improve the performance of modern medical services (i.e. telemedicine, PACS, radiology information systems, wireless personal/home health care systems). Image data security techniques, error resilience technologies, client-side interactive Region of Interest (ROI) transmission and decoding (e.g. for teleconsultation with very large radiography exams), and storage of multiple image data sets are considered in detail. Selected tests of coders realized according to parts I and II of JPEG2000 for different modality test images are presented to evaluate the compression efficacy of this standard. Exemplary results of encoding process optimisation by wavelet transform and subband decomposition selection and screen-shots of software interfaces designed for these tests are also presented.

• P.Kostka and E.Tkacz, Wavelet-neural Systems as Approximators of an Unknown Function - a Comparison of Biomedical Signal Classifiers

  Wavelet-neural systems (WNS) presented in this work, inheriting the properties of neural networks, belong to the class of universal approximators of unknown functions, F, describing the relationship between input X ∈ R^N and output Y ∈ R^M of a process or object. Classifier structures described in this work fulfil the role of approximators of functions, which are able to assign the input signal to a particular class with a given accuracy. A performance comparison of elaborated classifier structures with preliminary time-frequency analysis in the wavelet layer has been made for different types of the neural part. A feed forward multi-layer perceptron and a neural net with radial basic functions are analysed theoretically and practically. Results included in this paper present a comparison of the learning and verification stages of a classifier, tested on the basis of non-stationary signals of heart rate variability. Despite the fact that a WNS with the Morlet basic function gives the best results for the learning phase of WNS, the other tested wavelets used in the preliminary layer, Db4, allow us to obtain the best system performance during its verification.

• M.W.Kurzynski, Fuzzy Reasoning Applied to Multistage Diagnosis of Acute Renal Failure in Children

  The paper deals with fuzzy inference systems for multistage recognition based on a decision tree scheme. Two conceptually different fuzzy methods are presented and discussed for the given learning set. The first method is developed according to the multistage approach known as the Mamdani inference engine, with rules generated from the learning set. In the second approach, we first construct a fuzzy relation between the decision set and the feature space, which is then used for decision making. Both methods were practically applied to computer-aided medical diagnosis of acute renal failure. Results of comparative experimental analysis are given.

• L.Bobrowski, Feature Selection Based on Linear Separability and a CPL Criterion Function

  Linear separability of data sets is one of the basic concepts in the theory of neural networks and pattern recognition. Data sets are often linearly separable because of their high dimensionality. Such is the case of genomic data, in which a small number of cases is represented in a space with extremely high dimensionality. An evaluation of linear separability of two data sets can be combined with feature selection and carried out through minimisation of a convex and piecewise-linear (CPL) criterion function. The perceptron criterion function belongs to the CPL family. The basis exchange algorithms allow us to find minimal values of CPL functions efficiently, even in the case of large, multidimensional data sets.

• T.Przybyla, Breast Cancer Diagnosis via Fuzzy Clustering with Partial Supervision

  A new clustering method of fuzzy c-myriad clustering with partial supervision is presented in this paper. The proposed method has been applied to breast cancer diagnosis data obtainted from the University of Wisconsin. The data set contains 699 cases of breast cancer, with each instance described by 10 features.

• T.Pander, Application of Weighted Myriad Filters to Suppress Impulsive Noise in Biomedical Signals

  Biomedical signals are commonly recorded with many kinds of noise. One of these is a waveform of the electrical activity of muscles. This "natural" distortion is usually modelled with a white Gaussian noise. In order to suppress such noise a weighted myriad filter is applied. The weighted myriad filter belongs to a class of non-linear filters and requires knowledge about noise impulsiveness. An impulsive noise can be described with α-stable distributions. One objective of this paper is to apply α-stable distribution as a model of real-life muscle noise in ECG signals. The other objective of the paper is to apply a weighted myriad filter to suppress impulsive noise in biomedical (ECG) signals. The reference filters have been the Savitzky-Golay smoothing filter and the median filter. The obtained results have shown that α-stable distributions can be applied to model muscle noise and that a weighted myriad filter with a Chebyshev weighted function can effectively suppress such noise.

• R.Burduk, Pattern Recognition of Sacroileitis with the Use of Multistage Logic with a Fuzzy Loss Function

  The article describes the problem of pattern recognition of sacroileitis. Classification is based on a scheme of multistage recognition with a fuzzy loss function dependent on the node of the decision tree. Decision rules are based on k-nearest neighbors at particular internal nodes of the decision-tree. Paper presents influence of comparison fuzzy numbers on classifications results.

• M.Kretowski and J.Bezy-Wendling, Computer Modelling of Vascular Systems

  A model of the vascular system perfusing an internal organ is presented in the paper. The system's development is driven by the increasing needs of growing tissue. The modelled network consists of 2 or 3 (in the case of the liver) vascular trees connected on the macro-cell level. Each appearance of a new macro-cell activates an angiogenic process. The geometry of newly formed vessels is determined as a result of local optimization of the bifurcation volume. The model can simulate modifications of the vascular network caused by pathological processes.

• L.Witkowski, An Automatic System for Calculating Basic Semen Parameters

  Algorithms for examination of the density and selected parameters of movement of sperm cells have been elaborated and implemented. The conducted research is a part of work on a computer system for semen analysis. The system will allow for an increase in the precision of examination thanks to an exact specification of numerical values of the chosen parameters. An additional advantage of the system is an increased time efficiency of examination. Nowadays, the basic type of examination is an estimated analysis of semen parameters by visual observation of a sample, based on a physician's subjective assessment of an image. Registration of images is impossible in visual analysis.

• J.Zielinski and H.Krawczyk, Achieving High Dependability of an Endoscopy Recommender System (ERS)

  The paper presents a strategy for achieving high dependability of a computer-based system devoted to endoscopic examination (ERS). Two levels of replication are used: hardware (3 computers) and database (redundant copies). Archivisation of documents describing patient examinations and films made during such examinations is described. The influence of the used techniques on performance and dependability of the replicated system is estimated.

• J.W.Grzymala-Busse, Z.S.Hippe, M.Knap and T.Mroczek, A New Algorithm for Generation of Decision Trees

  A new algorithm for development of quasi-optimal decision trees, based on the Bayes theorem, has been created and tested. The algorithm generates a decision tree on the basis of Bayesian belief networks, created prior to the formation of the decision tree. The efficiency of this new algorithm was compared with three other known algorithms used to develop decision trees. The data set used for the experiments was a set of cases of skin lesions, histopatolgically verified.

• P.Hoser, A Mathematical Model of the Left Ventricle Surface and a Program for Visualization and Analysis of Cardiac Ventricle Functioning

  The left heart chamber's contractibility is an important part of heart diagnostics. Ultrasonographic pictures are very often used as the imaging method, as they are widely available, inexpensive and non-invasive. However, ultrasonographic pictures are very unclear, blurred and noisy, and thus very difficult for automatic analysis. To obtain a quick and useful analysis of ventricle performance, a special mathematical model has been created. The model can be used in contour detection, visualization of the heart's motion and even in automatic surface analysis. We hope that in the future such programs could be incorporated into a general medical expert system.

• M.Kaczmarek and A.Nowakowski, Active Dynamic Thermography in Mammography

  We discuss limitations of the known methods of IR imaging in diagnostics of breast cancer. In conclusion we show that the known methods, based on simple observation of external temperature distribution, are not fully effective. Even advanced pattern recognition could not help in analysis of static images. May active dynamic thermography, known in non-destructive testing of materials, be of any help in breast cancer diagnostics? Analysis of thermal transients forced by external thermal excitation shows, even on simple models, that one may expect a visible improvement in resolution after such excitation. Applied models allow analysis of both static and active thermograms. Basing on the models one may recognise elements of the internal structure of a breast not visible in static pictures. This method, new in clinical practice, seems to be promising, but requires further studies.

• K.Krysztoforski, A.Wolczowski, R.Bedzinski and K.Helt, Recognition of Palm Finger Movements on the Basis of EMG Signals with the Application of Wavelets

  The paper describes an EMG signal analysis based on the wavelet transform, applied for the hand prosthesis control. Signal features are represented by wavelet coefficients. A cross-validation method is applied for the feature selection process. The classification algorithm uses multistage recognition. The information about finger posture provided by a data glove is recorded concurrently with forearm EMG signals. The acquired data are used to train the classification algorithm.

• K.Stanisz-Wallis, A.Izworski, A.Dembinska-Kiec, R.Tadeusiewicz and T.Lech, Assessment of Diagnostic Features in the Coronary Artery Disease (CAD) by Application of Statistical Methods and Neural Networks

  The present work is aimed at comparing the effectiveness of two different methods of risk factor assessment used for prediction of the CAD (coronary artery disease): the logistic regression method and the application of artificial neural networks. The former is widely used in medical research, while the latter is relatively rare. In the logistic regression method hierarchical analysis was employed to select the significant variables of the classification process. In the neural network approach several strategies were proposed for selection of the discriminative variables, all based on weight analysis in the constructed networks. Both methods have produced a consistent set of discriminative variables (Glu0, Ins0, Ins30, BMI, apoA1 and HDL-Ch), belonging to three groups of risk factors associated with insulin resistance, obesity and lipid disorders.

• M.Grabowski, K.J.Filipiak, R.Rudowski and G.Opolski, An Expert System Supporting Risk Stratification in Acute Coronary Syndromes

  The aim of the project was to create a computer program, an expert system (ES), which would support physicians when a management for patients with acute coronary syndrome needs to be chosen. A knowledge database was created with the support of clinical experts, based on the current management standards. Data from new patients are added to the case database. The inference engine integrates two types of reasoning rule-based and case-based. The ES gives unambiguous and objective answers, its recommendation are reliable. At present, the ES is tested in clinical practice. Strategies recommended by the ES are compared with the management applied in a clinic.

• A.Kluza, Veterinary Toxicology Information System

  A Veterinary Toxicology Information System project has been started in Poland due to the country's growing needs in the field of veterinary toxicological remote consultations. An application designed to be accessed by veterinary surgeons will help them in poisoning diagnoses. Various data mining, classification and statistical/analytical methods will be available in the system for toxicology researchers.

• A.Przelaskowski, Irreversible Medical Image Compression: Conditions of Acceptability

  Acceptance of irreversible image compression applicable to medical imagery is controversial in the medical community. The influence of this irreversible process on degradation of diagnostic image features is considered and how to preserve diagnostic accuracy. Fears, doubts, the disadvantages of the data distortion process and the advantages of safe and efficient irreversible compression for image information storage and interchange are discussed. The effects of compression on various image exams are analysed. The conclusion is that irreversible compression is not to be afraid of, but its characteristics should be well understood before implementing it in current practice.