Catálogo de publicaciones - libros

In this paper a novel approach to the impulsive noise removal in color images is presented. The proposed technique employs the switching scheme based on the impulse detection mechanism using the so called concept. Compared to the vector median filter, the proposed technique consistently yields better results in suppressing both the random-valued and fixed-valued impulsive noise. The main advantage of the proposed noise detection framework is its enormous computational speed, which enables efficient filtering of large images in real-time applications.

This paper presents new idea of remote control over a wearable general purpose health monitor based on biosignals. The autonomy of wearable devices is usually an opposite requisite to the interpretation intelligence requiring computation power. The idea of device programmability is applied to continuous optimization of resources use towards the best diagnosis quality. The proposed concept of programmable recorder assumes high flexibility of the remote device, however only certain aspects of adaptation were implemented up to today. The application of such monitors extends beyond the traditional long term ECG recording and covers the area of sports, endurance, stress, pregnancy and elderly surveillance in various combinations.

Methodological aspects of quantitative morphological analysis of digital fundus eye images are described, both in the spatial and frequency domains. They refer to the notion of fractal dimension as a measure of image texture complexity thus providing an additional descriptor of the image, which may be useful for medical data classifications.

In this article we present a new method for recognition of subtle microcalcifications in high-resolution digital mammograms. The identification of suspicious regions in mammogram images is carried out using a method based on the 2D discrete wavelet transform. For classification of regions we use the SVC algorithm. Initially, a number of statistical features of mammogram regions was employed to achieve high classification accuracy. Using a novel clustering technique we performed feature selection and identified 18 highly descriptive features. Our method can achieve sensitivity as high as 0.97 while maintaining specificity above 0.90.

The gait characteristic features can be used for various diseases source and level diagnosis. The interferences of a physiological data record caused by any characteristic abnormal data spectrum (of the disease) corresponds to the disease source and level. The disease source can easy be classified by the relevant periodical functions. The interferences characteristics were analysed then matched to the well defined mathematical expressions for doing several manipulations on the data set. Knowing a range of the data records disorders the physiological data record can be modified due to multiply the data set of the defined disease class. This way the efficient amount of the data for the neural network training can be obtained. The paper shows several aspects of the gait data record analysis concerning two neurological diseases.

Recent developments of the construction of the syntactic pattern recognition-based expert System for Auditory Brainstem Response Analysis (SABRA) are described. The paper includes a brief characterization of problems related to the the use of computer science in ABR analysis, an outline of the concept of the SABRA system and a short description of GDPLL() and ACLL() grammars as the solution for fundamental problems that arose during the design of the system.

In the current study two various segmentation methods have been implemented sequentially in computer-aided approach to the assessment of skeletal maturity, where correct location of borders of anatomical structures is a crucial step. The first segmentation stage, based on the Gibbs random fields technique, correctly segments out a bony structure and roughly outlines the edges of cartilage while the second one using the active contour strategy, smoothes them and prepares for the feature extraction stage. A synthetic region of interest has been designed to test and adjust weights of the snake energy functional. These weights have afterwards been applied to a real image data. In comparison to our previous works we observe a significant improvement of boundary location mainly when cartilage is included in the epiphyses.

In this paper we discuss computer-aided diagnosing and classification of melanoid skin lesions. The main goal of our research was to elaborate and to promote via Internet a new skin lesion diagnostic computer system. Its functionality and structure is described briefly in this report. In the current version of the system, five learning models are implemented to simultaneously supply five independent, partial results. Then, a special evaluation and voting algorithm is applied to select the correct class (concept) of the diagnosed skin lesion.

Fetal heart rate (FHR) is the main source of information on the fetal state. FHR is characterized by two components: the basal fetal heart rate (baseline) and the variability of FHR — mainly accelerations (A) and decelerations (D). A correctly determined FHR baseline is a precondition for correct identification of the acceleration and deceleration patterns. Even a small inaccuracy in the FHR baseline curve may significantly distort the detection of A/D, which may subsequently lead to false interpretation of clinical symptoms. It is necessary to develop a method and criteria, which would allow evaluation of the baseline algorithms. The algorithms could be evaluated either by means of direct comparison of baselines or by a comparison of A/D patterns. The paper presents a method which can be used to determine matched A/D patterns recognized using two various baselines in the meaning of their time location. The method uses the coefficients of inconsistency between these patterns.

The analysis of fetal heart rate (FHR) trace is not based on evaluation of the baseline itself, but on the episodes which have been detected using this baseline - the acceleration (A) and deceleration (D) patterns. This implies the following approach: for a given trace the baseline is estimated automatically according to the algorithm selected, and once again manually by the clinical expert. For each baseline the A/D patterns are identified automatically by the fixed recognition algorithm. In the next step, the baseline interference procedure is used, which ensures constant difference between the primary baseline and the baseline with the interference superimposed. This difference corresponds to the average expertŠs inconsistency. The goal was to check the correlation between the inconsistencies of baselines and A/D episodes. The obtained results gave a proof that two baselines (from two different automated estimation methods) must be compared indirectly, solely on the basis of the A/D recognized when using them.