Vesna Zeljkovic

Classification algorithm of retina images of diabetic patients based on exudates detection

The chronic hyperglycemia of diabetes is associated with long-term damage, dysfunction, and failure of different organs, especially the eyes, kidneys, nerves, heart, and blood vessels. The regular examination of diabetic patients can potentially reduce the risk of vision impairment and in the last instance blindness. Early diabetic retinopathy detection enables application of laser therapy treatment in order to prevent or delay loss of vision. The diagnostics and detection of diabetic retinopathy is performed by specialized ophthalmologists manually and represents expensive procedure. Automatic exudates detection and retina images classification would be helpful for reducing diabetic retinopathy screening costs and encouraging regular examinations. We proposed the automated algorithm that applies mathematical modeling which enables light intensity levels emphasis, easier exudates detection, efficient and correct classification of retina images. The proposed algorithm is robust to various appearance changes of retinal fundus images which are usually processed in clinical environments.

Exudates and optic disk detection in retinal images of diabetic patients

Diabetic retinopathy is the progressive pathological alterations in the retinal microvasculature that very often causes blindness. Because of its clinical significance, it will be helpful to have regular cost‐effective eye screening for diabetic patients by developing algorithms to perform retinal image analysis, fundus image enhancement, and monitoring. The two cost‐effective algorithms are proposed for exudates detection and optic disk extraction aimed for retinal images classification and diagnosis assistance. They represent the effort made to offer a cost‐effective algorithm for optic disk identification, which will enable easier exudates extraction, exudates detection and retinal images classification aimed to assist ophthalmologists while making diagnoses. The proposed algorithms apply mathematical modeling, which enables light intensity levels emphasis, easier optic disk and exudates detection, efficient and correct classification of retinal images. The algorithm is robust to various appearance changes of retinal fundus images and shows very promising results. Fundus images are classified into those that are healthy and those affected by diabetes, based on the detected optic disk and exudates. The obtained results indicate that the proposed algorithm successfully and correctly classifies more than 98% of the observed retinal images because of the changes in the appearance of retinal fundus images typically encountered in clinical environments. Copyright

Multimodal classification of heart sounds attributes

Pollution and associated negative impacts on human health is one of the major concerns of the World Health Organization and healthcare providers. Current interests focus on particles suspended in air known as PM10 which significantly contribute to increased prevalence of heart disease. Specifically, the city of Mexicali was found to be one of the most polluted cities of Mexico in 2010. Cardiovascular abnormalities are often reflected in characteristic indicators of auscultation based examination. This fundamental diagnostic procedure can be significantly enhanced using low-cost detection technologies and accompanied pattern recognition for classification of associated sound attributes. Related economic issues are critical, both in Latin America and in other regions of the world, where often a limited level of specialized healthcare services are available. One of the goals of these studies was to prove initially demonstrated capabilities that the distinctive auscultatory classification indicators and diagnostic assessment can be easily implemented. In the case of heart sound signals, both the normal sounds and those representing abnormal conditions can be examined and differentiated for diagnostic purposes. The main focus of this study was to use Hidden Markov Models (HMM) for the classification and evaluation of heart sounds (HS). In particular, the application of HMM models provides greater robustness to noise and other interference such as the GMM models. The results demonstrate an enhanced quantitative evaluation, which could assist in a more accurate and economical HS assessment.

The HMM diagnostic models of respiratory sounds

Numerous studies including annual reports by Blacksmith Institute clearly document the magnitude of regional pollution and associated health risks. In particular the air pollution encompassing PM10 and smaller particles significantly contributes to the prevalence of respiratory diseases. Specifically, the city of Mexicali with a PM10 ranking of 137 in 2010 is considered as the most polluted city in Mexico largely due to the contribution of unusual environmental factors. Resulting respiratory abnormalities are often reflected in peculiar auscultatory indicators and their assessment can be accomplished using low cost technologies. These economic aspects are critical not only in Latin America but also other population centers globally considering the limited level of health services. Any classification of auscultatory indicators as reflected in lung sound (LS) characteristics needs to account for a noisy environment and the influence of heart sounds (HS). The aim of these studies was to utilize Hidden Markov Models (HMM) in light of the previously conducted assessment of lung sounds (LS) utilizing the Mixture Gaussians Models (GMM). In particular, the application of HMM models provides robustness to cope with noise and other interferences, to which the Mixture Gaussians Models (GMM) are more vulnerable. The conducted studies document that presented quantitative assessment of LS may add in more objective and economic scanning for respiratory abnormalities.

Diagnostic spectral segregation of cardiac sounds features

Globally increasing rate of cardiovascular diseases represents one of the primary health care challenges and necessitates broader population screening for earlier intervention. Present cardiovascular function assessment techniques encompass a broad range of diagnostic tools which in most cases involve high and mid level cost instrumentation. They include among others instrumentation as a part of nuclear cardiology, cardiac catheterization, cardiac ultrasound, relatively basic multiple or simple channel ECG acquisition systems, and other. These techniques and related instrumentation represent very useful and powerful tools to diagnose and/or manage particular cardiac abnormalities; however they are not suitable for broader population low cost basic supplemental cardiac screening. Therefore, these studies focused on a structured assessment of cardiac sound traits as related auscultation is a part of standard medical exams and globally accepted procedures. For these reasons a dedicated cardiac sounds assessment algorithm was developed and has been tested on heart sound database that contains 565 signals encompassing 202 normal and 363 abnormal sounds. A broad range of diverse cardiac sound indicators, versatility of database, and derived characteristics validate potency of possible clinical usefulness and potential for a low cost broader diagnostic screening. In particular, the group of abnormal heart sounds contains samples of 35 different heart abnormalities which are divided into two groups including Group I that contains 28 different heart abnormalities that are 81.81% successfully detectable by the developed algorithm and Group II that contains 7 different heart abnormalities that do not exhibit the frequency change of S1 sound and are not classified by the proposed method. Normal heart sounds in 97.03% of cases were successfully classified with the developed method which is addressing the needs for a selected cardiac screening in light of increasing prevalence of cardiovascular diseases.

Pre-Ictal phase detection algorithm based on one dimensional EEG signals and two dimensional formed images analysis

Over 50 million persons worldwide are affected by epilepsy. It can affect equally young babies as well as old people. Epilepsy is a brain disorder characterized by the neurobiological, cognitive, psychological and social consequences. It is known for sudden, unexpected transitions from normal to pathological behavioral states called epileptic seizures. Epilepsy poses a significant burden to society due to associated healthcare cost to treat and control the unpredictable and spontaneous occurrence of seizures. There is a need for a quick screening process that could help neurologists diagnose and determine the patients treatment. Electroencephalogram has been traditionally used to diagnose patients by evaluating those brain functions that might correspond to epilepsy. This research focuses on developing new classification technique and the prediction of pre-ictal states that announce epileptic seizures, from the online EEG data analysis. The idea is to place electrodes on critical regions on the patients head that would wirelessly communicate with the EEG recorder and the unit that performs online automated pre-ictal state detection based on obtained EEG signal. The patient should get timely alert about the possible seizure attack so that she/he can stop with its activities and take safety precautions.

A new approach for binary feature selection and combining classifiers

This paper explores feature selection and combining classifiers when binary features are used. The concept of Non-Reducible Descriptors (NRDs) for binary features is introduced. NRDs are descriptors of patterns that do not contain any redundant information. The underlying mathematical model for the present approach is based on learning Boolean formulas which are used to represent NRDs as conjunctions. Starting with a description of a computational procedure for the construction of all NRDs for a pattern, a two-step solution method is presented for the feature selection problem. The method computes weights of features during the construction of NRDs in the first step. The second step in the method then updates these weights based on repeated occurrences of features in the constructed NRDs. The paper then proceeds to present a new procedure for combining classifiers based on the votes computed for different classifiers. This procedure uses three different approaches for obtaining the single combined classifier, using majority, averaging, and randomized vote.

Personal access control system using moving object detection and face recognition

Real time automated personal access control system is proposed in order to detect the moving objects, localize, extract and recognize their faces in real image sequence. The described method encompasses two important issues in personal access control system that receives increased attention over years: moving object detection and face recognition. It is tested on personal access controlled area video testing. The efficiency of the described system is illustrated on four real world interior video sequences recorded in indoor/outdoor mixed environment with slight illumination changes.

Pre-ictal phase detection with SVMs

Over 50 million persons worldwide are affected by epilepsy. Epilepsy is a brain disorder known for sudden, unexpected transitions from normal to pathological behavioral states called epileptic seizures. Epilepsy poses a significant burden to society due to associated healthcare cost to treat and control the unpredictable and spontaneous occurrence of seizures. There is a need for a quick screening process that could help neurologist diagnose and determine the patients treatment. Electroencephalogram has been traditionally used to diagnose patients by evaluating those brain functions that may correspond to epilepsy. The objective of this paper is to implement a novel detection technique of pre-ictal state that announces epileptic seizures from the online EEG data analysis. Unlike most published methods, that are aimed to distinguish only the normal from the epilepsy state, in this work the pre-ictal state is introduced as a new patient status, thus differentiating three possible states: normal (healthy), pre-ictal and epileptic seizure. In this manner, the patient should get timely alert about the possible seizure attack so that she/he can stop with its activities and take safety precautions.

A framework for TV logos learning using linear inverse diffusion filters for noise removal

Different logotypes represent significant cues for video annotations. A combination of temporal and spatial segmentation methods can be used for logo extraction from various video contents. To achieve this segmentation, pixels with low variation of intensity over time are detected. Static backgrounds can become spurious parts of these logos. This paper offers a new way to use several segmentations of logos to learn new logo models from which noise has been removed. First, we group segmented logos of similar appearances into different clusters. Then, a model is learned for each cluster that has a minimum number of members. This is done by applying a linear inverse diffusion filter to all logos in each cluster. Our experiments demonstrate that this filter removes most of the noise that was added to the logo during segmentation and it successfully copes with misclassified logos that have been wrongly added to a cluster.