Search

The therapeutic efficacy of medication and treatment strategies is based on the diagnosis and staging of liver diseases. According to recent researches, the Collagen Proportional Area (CPA) is a reliable metric to assess fibrosis in liver tissues. Several image processing techniques are used for the analysis of liver biopsy images, providing objective assessment for the severity of the disease. In current work a novel classification via clustering algorithm is proposed, based on K-means, which is used for image segmentation of liver biopsies. More specifically, supervised learning is employed to insert constraints on centroids movement. Furthermore, feature weighting is utilized for the classification process. At first, a hypercube is extracted and feature weights are computed, for each class, using a training set of liver biopsy images. Classification via clustering follows, initializing a centroid for each class within the respective hypercube, which, during the iterations of the clustering, is allowed to move only inside the hypercube. The weighted Euclidean distance is used as similarity criterion to the clusters. 93 liver biopsy images are employed to evaluate the proposed approach. The classification results along with CPA values are computed.

Touch is a fundamental aspect of human interaction with the surrounding environment. It affects individuals' development in different manners and figures prominently in everyday operations such as the sense of presence, object recognition, performing actions, non-verbal communication and emotional state. In recent years there has been a growth of interest in researching the electro physiological activity of the brain originating from haptic stimulation. In the present preliminary experiment, we performed a classification process of extracted EEG features acquired from four healthy participants' EEG data when they actively touched different natural textures. Each participant was asked to use their fingertips and calmly rub for one minute, each of the three different textured materials (smooth, rough and water surface). EEG recordings were acquired and processed. Next, time and frequency-based features were extracted and used as inputto four classifiers to correctly identify each different texture. The results obtained show a classification performance of 63% with C4.5 algorithm and 76% with Random Forests and 10-fold cross-validation.

In this paper, it is discussed the control of a double inverted pendulum (DIP) on a cart with the use of a Fuzzy Logic Controller (FLC). The DIP is a classic problem in mechanics and the control procedure to stabilize around equilibrium position constitutes a classic criterion for the various control methods in Control Systems Theory and Robotics e.g., can serve as a representative model for the computation and stability of walking robots. To access the efficacy of the proposed method, a linear quadratic regulator (LQR) was used as a benchmark. The study showed that the proposed FLC can provide considerably better performance than LQR under different initial conditions of the system. This performance is remarkably significant in terms of the system outputs peak values reduction. The balancing and velocity tracking performance of the proposed controller was studied via simulations. The simulation result shows the effectiveness of our control scheme.

This article introduces a method that can automatically classify the hemiplegia type (right or left side of the body is paralyzed) between healthy and non-healthy subjects. The proposed method utilizes the data taken from the accelerometer sensor of the RehaGait mobile gait analysis system. These data undergo a pre-processing and feature extraction stage before being sent as input to a scaled conjugate gradient backpropagation (SCG-BP) trained neural network. The proposed system is tested using a custom-created dataset containing 10 healthy and 20 patients suffering from hemiplegia (right or left). The experimental part of the system utilized 7 sensors placed on the left and right foot, the left and right shank, the left and right thigh, and the hip of each subject. Each sensor captured a 3-dimensional (3D) signal from 3 different device types: accelerometer, magnetometer, and gyroscope. The system utilized and split into 2-second windows only the accelerometer data, achieving a classification accuracy of 87.71%.

In this paper, a hybrid learning approach, which combines the extreme learning machine (ELM) with a genetic algorithm (GA), is proposed. The utilization of this hybrid algorithm enables the creation of heterogeneous single layer neural networks (SLNNs) with better generalization ability than traditional ELM in terms of lower mean square error (MSE) for regression problems or higher accuracy for classification problems. The architecture of this method is not limited to traditional linear neurons, where each input participates equally to the neuron’s activation, but is extended to support higher order neurons which affect the network’s generalization ability. Initially, the proposed heterogeneous hybrid extreme learning machine (He-HyELM) algorithm creates a number of custom created neurons with different structure, which are used for the creation of homogeneous SLNNs. These networks are trained with ELM and an application specific GA evolves them into heterogeneous networks according to a fitness criterion utilizing the uniform crossover operator for the recombination process. After the completion of the evolution process, the network with the best fitness is selected as the most optimal. Experimental results demonstrate that the proposed learning algorithm can get better results than traditional ELM, homogeneous hybrid extreme learning machine (Ho-HyELM) and optimally pruned extreme learning machine (OP-ELM) for homogeneous and heterogeneous SLNNs.

Alzheimer’s Disease (AD) is a neurodegenerative disorder and the most common type of dementia with a great prevalence in western countries. The diagnosis of AD and its progression is performed through a variety of clinical procedures including neuropsychological and physical examination, Electroencephalographic (EEG) recording, brain imaging and blood analysis. During the last decades, analysis of the electrophysiological dynamics in AD patients has gained great research interest, as an alternative and cost-effective approach. This paper summarizes recent publications focusing on (a) AD detection and (b) the correlation of quantitative EEG features with AD progression, as it is estimated by Mini Mental State Examination (MMSE) score. A total of 49 experimental studies published from 2009 until 2020, which apply machine learning algorithms on resting state EEG recordings from AD patients, are reviewed. Results of each experimental study are presented and compared. The majority of the studies focus on AD detection incorporating Support Vector Machines, while deep learning techniques have not yet been applied on large EEG datasets. Promising conclusions for future studies are presented.

The purpose of fetal monitoring during childbirth is the early recognition of any pathological conditions to guide a clinician in early intervention to avoid any complication in the health of the fetus. Non-Invasive Fetal Electrocardiography (NIFECG) represents an alternative fetal monitoring technique. The fetal ECG (fECG) derived from maternal thoracic and abdominal ECG recordings, provides an alternative to typical embryo monitoring means. In addition, it allows for long-term and ambulatory registrations that broaden the diagnostic capabilities for assessing the fetal health. However, in real situations, clear fECG is difficult to extract because it is usually overwhelmed by the dominant maternal ECG and other contaminated noise such as baseline wander and high-frequency interference. In this paper, a novel integrated adaptive methodology based on the combination of blind source separation, empirical mode decomposition, wavelet shrinkage denoising and correlation analysis, for the non-invasive extraction and processing of the FECG, is proposed. The methodology has been evaluated using both real and simulated recordings, and the obtained results indicate it efficiently.

This paper presents a hybrid algorithm for the creation of heterogeneous single layer neural networks (SLNNs). The proposed self-adaptive heterogeneous hybrid extreme learning machine (SA-He-HyELM) trains a series of SLNNs with different neuron types in the hidden layer utilizing the extreme learning machine (ELM) algorithm. These networks are evolved into heterogeneous networks (networks having different combinations of hidden neurons) with the help of a modified genetic algorithm (GA). The algorithm is able to handle two architecturally different neuron types: traditional low order (linear) units and higher order units with different transfer functions. The GA is fully self-adaptive and uses one novel hybrid crossover operator along with a self-adaptive mutation operator in order to retain ELM’s simplicity and minimize the number of parameters need tuning. The experimental part of the current paper involves testing SA-He-HyELM with traditional ELM and other three ELM-based methods. The experimental part utilized a series of regression and classification experiments on relatively large datasets. In all cases the proposed method managed to get lower MSE or higher classification accuracy when compared to the aforementioned methods.

A methodology for prediction of pre-term births is presented in this paper. The methodology is based on the analysis of EHG signals and data mining techniques. Initially, spectral and non-linear characteristics of the EHG are extracted, forming a pattern that is used to train a classifier to discriminate between term and pre-term cases. The method has been tested using a benchmark EHG database, and the obtained results indicate its effectiveness in accurate pre-term/term labour prediction.

Drug inefficiency in patients with refractory seizures renders epilepsy a life-threatening and challenging brain disorder and stresses the need for accurate seizure detection and prediction methods and more personalized closed-loop treatment systems. In this paper, a multicenter methodology for automated seizure detection based on Discrete Wavelet Transform (DWT) is presented. A decomposition of 5 levels is applied in each EEG segment and five features are extracted from the wavelet coefficients. The extracted feature vector is used to train a Random Forest classifier and discriminate between ictal and interictal data. EEG recordings from the database of University of Bonn and the database of the University Hospital of Freiburg were employed, in an attempt to test the efficiency and robustness of the method. Classification results in both databases are significant, reaching accuracy above 95% and confirming the robustness of the methodology. Sensitivity and False Positive Rate for the Freiburg database reached 99.74% and 0.21/h respectively.

Background & Aims Liver biopsy is the reference standard for staging and grading nonalcoholic fatty liver disease (NAFLD), but histologic scoring systems are semiquantitative with marked interobserver and intraobserver variation. We used machine learning to develop fully automated software for quantification of steatosis, inflammation, ballooning, and fibrosis in biopsy specimens from patients with NAFLD and validated the technology in a separate group of patients. Methods We collected data from 246 consecutive patients with biopsy-proven NAFLD and followed up in London from January 2010 through December 2016. Biopsy specimens from the first 100 patients were used to derive the algorithm and biopsy specimens from the following 146 were used to validate it. Biopsy specimens were scored independently by pathologists using the Nonalcoholic Steatohepatitis Clinical Research Network criteria and digitalized. Areas of steatosis, inflammation, ballooning, and fibrosis were annotated on biopsy specimens by 2 hepatobiliary histopathologists to facilitate machine learning. Images of biopsies from the derivation and validation sets then were analyzed by the algorithm to compute percentages of fat, inflammation, ballooning, and fibrosis, as well as the collagen proportionate area, and compared with findings from pathologists’ manual annotations and conventional scoring systems. Results In the derivation group, results from manual annotation and the software had an interclass correlation coefficient (ICC) of 0.97 for steatosis (95% CI, 0.95–0.99; P < .001); ICC of 0.96 for inflammation (95% CI, 0.9–0.98; P < .001); ICC of 0.94 for ballooning (95% CI, 0.87–0.98; P

Objective characterization of pain intensity is necessary under certain clinical conditions. The portable electroencephalogram (EEG) is a cost-effective assessment tool and lately, new methods using efficient analysis of related dynamic changes in brain activity in the EEG recordings proved that these can reflect the dynamic changes of pain intensity. In this paper, a novel method for automated assessment of pain intensity using EEG data is presented. EEG recordings from twenty-two (22) healthy volunteers are recorded with the Emotiv EPOC+ using the Cold Pressor Test (CPT) protocol. The relative power of each brain band's energy for each channel is extracted and the stochastic forest algorithm is employed for discrimination across five classes, depicting the pain intensity. Obtained results in terms of classification accuracy reached high levels (72.7%), which renders the proposed method suitable for automated pain detection and quantification of its intensity.

Nonalcoholic fatty liver disease (NAFLD) presents a wide range of pathological conditions, varying from nonalcoholic steatohepatitis (NASH) to cirrhosis and hepatocellular carcinoma (HCC). Their prevalence is characterized by increased fat accumulation and hepatocellular ballooning. They have become a cause of concern among physicians and engineers, as significant implications tend to occur regarding their accurate diagnosis and treatment. Although magnetic resonance, ultrasonography and other noninvasive methods can reveal the presence of NAFLD, image quantitative interpretation through histology has become the gold standard in clinical examinations. The proposed work introduces a fully automated diagnostic tool, taking into account the high discrimination capability of histological findings in liver biopsy images. The developed methodology is based on deep supervised learning and image analysis techniques, with the determination of an efficient convolutional neural network (CNN) architecture, performing eventually a classification accuracy of 95%.

Alzheimer&rsquo, s Disease (AD) is a neurogenerative disorder and the most common type of dementia with a rapidly increasing world prevalence. In this paper, the ability of several statistical and spectral features to detect AD from electroencephalographic (EEG) recordings is evaluated. For this purpose, clinical EEG recordings from 14 patients with AD (8 with mild AD and 6 with moderate AD) and 10 healthy, age-matched individuals are analyzed. The EEG signals are initially segmented in nonoverlapping epochs of different lengths ranging from 5 s to 12 s. Then, a group of statistical and spectral features calculated for each EEG rhythm (&delta, &theta, &alpha, &beta, and &gamma, ) are extracted, forming the feature vector that trained and tested a Random Forests classifier. Six classification problems are addressed, including the discrimination from whole-brain dynamics and separately from specific brain regions in order to highlight any alterations of the cortical regions. The results indicated a high accuracy ranging from 88.79% to 96.78% for whole-brain classification. Also, the classification accuracy was higher at the posterior and central regions than at the frontal area and the right side of temporal lobe for all classification problems.

Background The spectral information of the EEG signal with respect to epilepsy is examined in this study. Method In order to assess the impact of the alternative definitions of the frequency sub-bands that are analysed, a number of spectral thresholds are defined and the respective frequency sub-band combinations are generated. For each of these frequency sub-band combination, the EEG signal is analysed and a vector of spectral characteristics is defined. Based on this feature vector, a classification schema is used to measure the appropriateness of the specific frequency sub-band combination, in terms of epileptic EEG classification accuracy. Results The obtained results indicate that additional frequency band analysis is beneficial towards epilepsy detection. Conclusions This work includes the first systematic assessment of the impact of the frequency sub-bands to the epileptic EEG classification accuracy, and the obtained results revealed several frequency sub-band combinations that achieve high classification accuracy and have never been reported in the literature before.

Discrimination of eye movements and visual states is a flourishing field of research and there is an urgent need for non-manual EEG-based wheelchair control and navigation systems. This paper presents a novel system that utilizes a brain–computer interface (BCI) to capture electroencephalographic (EEG) signals from human subjects while eye movement and subsequently classify them into six categories by applying a random forests (RF) classification algorithm. RF is an ensemble learning method that constructs a series of decision trees where each tree gives a class prediction, and the class with the highest number of class predictions becomes the model’s prediction. The categories of the proposed random forests brain–computer interface (RF-BCI) are defined according to the position of the subject’s eyes: open, closed, left, right, up, and down. The purpose of RF-BCI is to be utilized as an EEG-based control system for driving an electromechanical wheelchair (rehabilitation device). The proposed approach has been tested using a dataset containing 219 records taken from 10 different patients. The BCI implemented the EPOC Flex head cap system, which includes 32 saline felt sensors for capturing the subjects’ EEG signals. Each sensor caught four different brain waves (delta, theta, alpha, and beta) per second. Then, these signals were split in 4-second windows resulting in 512 samples per record and the band energy was extracted for each EEG rhythm. The proposed system was compared with naïve Bayes, Bayes Network, k-nearest neighbors (K-NN), multilayer perceptron (MLP), support vector machine (SVM), J48-C4.5 decision tree, and Bagging classification algorithms. The experimental results showed that the RF algorithm outperformed compared to the other approaches and high levels of accuracy (85.39%) for a 6-class classification are obtained. This method exploits high spatial information acquired from the Emotiv EPOC Flex wearable EEG recording device and examines successfully the potential of this device to be used for BCI wheelchair technology.

Nonalcoholic fatty liver disease (NAFLD) is responsible for a wide range of pathological disorders. It is characterized by the prevalence of steatosis, which results in excessive accumulation of triglyceride in the liver tissue. At high rates, it can lead to a partial or total occlusion of the organ. In contrast, nonalcoholic steatohepatitis (NASH) is a progressive form of NAFLD, with the inclusion of hepatocellular injury and inflammation histological diseases. Since there is no approved pharmacotherapeutic solution for both conditions, physicians and engineers are constantly in search for fast and accurate diagnostic methods. The proposed work introduces a fully automated classification approach, taking into consideration the high discrimination capability of four histological tissue alterations. The proposed work utilizes a deep supervised learning method, with a convolutional neural network (CNN) architecture achieving a classification accuracy of 95%. The classification capability of the new CNN model is compared with a pre-trained AlexNet model, a visual geometry group (VGG)-16 deep architecture and a conventional multilayer perceptron (MLP) artificial neural network. The results show that the constructed model can achieve better classification accuracy than VGG-16 (94%) and MLP (90.3%), while AlexNet emerges as the most efficient classifier (97%).

The complexity of epilepsy created a fertile ground for further research in automated methods, attempting to help the epileptologists’ task. Over the past years, great breakthroughs have emerged in computer-aided analysis. Furthermore, the advent of Brain Computer Interface (BCI) systems has facilitated significantly the automated seizure analysis. In this study, an evaluation of the window size in automated seizure detection is proposed. The EEG signals from the University of Bonn was employed and segmented into 24 epochs of different window lengths with 50% overlap each time. Statistical and spectral features were extracted in the OpenViBE scenario and were used to train four different classifiers. Results in terms of accuracy were above 80% for the Decision Tree classifier. Also, results indicated that different window sizes provide small variations in classification accuracy.

Classification-via-clustering (CvC) is a widely used method, using a clustering procedure to perform classification tasks. In this paper, a novel K-Means-based CvC algorithm is presented, analysed and evaluated. Two additional techniques are employed to reduce the effects of the limitations of K-Means. A hypercube of constraints is defined for each centroid and weights are acquired for each attribute of each class, for the use of a weighted Euclidean distance as a similarity criterion in the clustering procedure. Experiments are made with 42 well–known classification datasets. The experimental results demonstrate that the proposed algorithm outperforms CvC with simple K-Means.

Epilepsy is a complex neurological disorder recognized by abnormal synchronization of cerebral neurons, named seizures. During the last decades, significant progress has been done in automated detection and prediction of seizures, aiming to develop personalized closed-loop intervention systems. In this paper, a methodology for automated seizure detection based on Discrete Wavelet Transform (DWT) is presented. Twenty-one intracranial ictal recordings acquired from the database of University Hospital of Freiburg are firstly segmented in 2 s epochs. Then, a five-level decomposition is applied in each segment and five features are extracted from the wavelet coefficients. The extracted feature vector is used to train a Support Vector Machines (SVM) classifier. Average sensitivity and specificity reached above 93% and 99% respectively.

In this work, an innovative classification algorithmic technique through sequential pattern mining was developed to predict the secondary structure of proteins. A basic algorithm was selected for the extraction of the sequential patterns and another algorithm was developed which employs these patterns for protein structure prediction. In the matter of predicting protein structures and scoring sequential patterns, several methodologies has been implemented that theoretically and experimentally overcome the disadvantages of existing algorithms.

Diagnosis and staging of liver diseases are essential for the therapeutic efficacy of medication and treatment strategies. Measuring the Collagen Proportional Area (CPA) in liver biopsies recently becomes an effective tool for the assessment of fibrosis in liver tissues. State of the art image processing techniques are employed to analyze biopsy images, providing objective assessment of diseases severity. In current work a novel modification of K-means clustering is proposed for image segmentation of liver biopsies. More specifically, supervised restriction of centroids movement is utilized. In the first stage, a training set of images are employed to extract a hypercube for each class. Then, one centroid is initialized inside each hypercube and during the iterations of the clustering is allowed to move only inside the hypercube. For the evaluation of the proposed method 8 liver biopsy images are employed and classification results along with CPA values are computed for each image.

Epilepsy is a chronic neurological disorder characterized by recurrent, sudden discharges of cerebral neurons, called seizures. Seizures are not always clearly defined and have extremely varied morphologies. Neurophysiologists are not always able to discriminate seizures, especially in long-term EEG datasets. Affecting 1% of the worlds population with 1/3 of the epileptic patients not corresponding to anti-epileptic medication, epilepsy is constantly under the microscope and systems for automated detection of seizures are thoroughly examined. In this paper, a method for automated detection of epileptic activity is presented. The Discrete Wavelet Transform (DWT) is used to decompose the EEG recordings in several subbands and five features are extracted from the wavelet coefficients creating a set of features. The extracted feature vector is used to train a Support Vector Machine (SVM) classifier. Five classification problems are addressed, reaching high levels of overall accuracy ranging from 87% to 100%.

A state-of-the-art method based on a grammatical evolution approach is utilized in this study to classify EEG signals. The method is able to construct nonlinear mappings of the original features in order to improve their effectiveness when used as input into artificial intelligence techniques. Several features are initially extracted from the EEG signals which are subsequently used to create the non-linear mappings. Then, a classification stage is applied, using multi-layer perceptron (MLP) and radial basis functions (RBF), to categorize the EEG signals. The proposed method is evaluated using a benchmark epileptic EEG dataset and promising results are reported.

Methodology for collagen proportional area extraction from liver biopsy images.Fully automated, machine-learning based image analysis.Robust methodology, without any threshold-based decisions.Processing low resolutions images without sophisticated equipment.Low processing time. Background and objectiveCollagen proportional area (CPA) extraction in liver biopsy images provides the degree of fibrosis expansion in liver tissue, which is the most characteristic histological alteration in hepatitis C virus (HCV). Assessment of the fibrotic tissue is currently based on semiquantitative staging scores such as Ishak and Metavir. Since its introduction as a fibrotic tissue assessment technique, CPA calculation based on image analysis techniques has proven to be more accurate than semiquantitative scores. However, CPA has yet to reach everyday clinical practice, since the lack of standardized and robust methods for computerized image analysis for CPA assessment have proven to be a major limitation. MethodsThe current work introduces a three-stage fully automated methodology for CPA extraction based on machine learning techniques. Specifically, clustering algorithms have been employed for background-tissue separation, as well as for fibrosis detection in liver tissue regions, in the first and the third stage of the methodology, respectively. Due to the existence of several types of tissue regions in the image (such as blood clots, muscle tissue, structural collagen, etc.), classification algorithms have been employed to identify liver tissue regions and exclude all other non-liver tissue regions from CPA computation. ResultsFor the evaluation of the methodology, 79 liver biopsy images have been employed, obtaining 1.31% mean absolute CPA error, with 0.923 concordance correlation coefficient. ConclusionsThe proposed methodology is designed to (i) avoid manual threshold-based and region selection processes, widely used in similar approaches presented in the literature, and (ii) minimize CPA calculation time.

A novel three-stage methodology for the detection of fetal heart rate (fHR) from multivariate abdominal ECG recordings is introduced. In the first stage, the maternal R-peaks and fiducial points (maternal QRS onset and offset) are detected, using band-pass filtering and phase space analysis. The maternal fiducial points are used to eliminate the maternal QRS complexes from the abdominal ECG recordings. In the second stage, two denoising procedures are applied to enhance the fetal QRS complexes. The phase space characteristics are employed to identify fetal heart beats not overlapping with the maternal QRSs, which are eliminated in the first stage. The extraction of the fHR is accomplished in the third stage, using a histogram-based technique in order to identify the location of the fetal heart beats that overlap with the maternal QRSs. The methodology is evaluated on simulated multichannel ECG signals, generated by a recently proposed model with various SNRs, and on real signals, recorded from pregnant women in various weeks during gestation. In both cases, the obtained results indicate high performance; in the simulated ECGs, the accuracy ranges from 72.78% to 98.61%, depending on the employed SNR, while in the real recordings, the average accuracy is 95.45%. The proposed methodology is advantageous since it copes with the existence of noise from various sources while it is applicable in multichannel abdominal recordings.

In this paper we propose a generic methodology for the automated generation of fuzzy models. The methodology is realized in three stages. Initially, a crisp model is created and in the second stage it is transformed to a fuzzy one. In the third stage, all parameters entering the fuzzy model are optimized. The proposed methodology is novel and generic since it can integrate alternative techniques in each of its stages. A specific realization of this methodology is implemented, using decision trees for the creation of the crisp model, the sigmoid function, the min-max operators and the maximum defuzzifier, for the transformation of the crisp model into a fuzzy one, and four different optimization strategies, including global and local optimization techniques, as well as, hybrid approaches. The proposed methodology presents several advantages and novelties: the transformation of the crisp model to the respective fuzzy one is straightforward ensuring its full automated nature and it introduces a set of parameters, expressing the importance of each fuzzy rule. The above realization is extensively evaluated using several benchmark data sets from the UCI machine learning repository and the obtained results indicate its high efficiency.

We present a methodology for sequence classification, which employs sequential pattern mining and optimization, in a two-stage process. In the first stage, a sequence classification model is defined, based on a set of sequential patterns and two sets of weights are introduced, one for the patterns and one for classes. In the second stage, an optimization technique is employed to estimate the weight values and achieve optimal classification accuracy. Extensive evaluation of the methodology is carried out, by varying the number of sequences, the number of patterns and the number of classes and it is compared with similar sequence classification approaches.

A fuzzy rule-based decision support system (DSS) is presented for the diagnosis of coronary artery disease (CAD). The system is automatically generated from an initial annotated dataset, using a four stage methodology: 1) induction of a decision tree from the data; 2) extraction of a set of rules from the decision tree, in disjunctive normal form and formulation of a crisp model; 3) transformation of the crisp set of rules into a fuzzy model; and 4) optimization of the parameters of the fuzzy model. The dataset used for the DSS generation and evaluation consists of 199 subjects, each one characterized by 19 features, including demographic and history data, as well as laboratory examinations. Tenfold cross validation is employed, and the average sensitivity and specificity obtained is 62% and 54%, respectively, using the set of rules extracted from the decision tree (first and second stages), while the average sensitivity and specificity increase to 80% and 65%, respectively, when the fuzzification and optimization stages are used. The system offers several advantages since it is automatically generated, it provides CAD diagnosis based on easily and noninvasively acquired features, and is able to provide interpretation for the decisions made.

In this paper we present a novel methodology for sequence classification, based on sequential pattern mining and optimization algorithms. The proposed methodology automatically generates a sequence classification model, based on a two stage process. In the first stage, a sequential pattern mining algorithm is applied to a set of sequences and the sequential patterns are extracted. Then, the score of every pattern with respect to each sequence is calculated using a scoring function and the score of each class under consideration is estimated by summing the specific pattern scores. Each score is updated, multiplied by a weight and the output of the first stage is the classification confusion matrix of the sequences. In the second stage an optimization technique, aims to finding a set of weights which minimize an objective function, defined using the classification confusion matrix. The set of the extracted sequential patterns and the optimal weights of the classes comprise the sequence classification model. Extensive evaluation of the methodology was carried out in the protein classification domain, by varying the number of training and test sequences, the number of patterns and the number of classes. The methodology is compared with other similar sequence classification approaches. The proposed methodology exhibits several advantages, such as automated weight assignment to classes using optimization techniques and knowledge discovery in the domain of application.

We have developed an automatic arrhythmia detection system, which is based on heart rate features only. Initially, the RR interval duration signal is extracted from ECG recordings and segmented into small intervals. The analysis is based on both time and time-frequency (t-f) features. Time domain measurements are extracted and several combinations between the obtained features are used for the training of a set of neural networks. Short time Fourier transform and several time-frequency distributions (TFD) are used in the t-f analysis. The features obtained are used for the training of a set of neural networks, one for each distribution. The proposed approach is tested using the MIT-BIH arrhythmia database and satisfactory results are obtained for both sensitivity and specificity (87.5 and 89.5%, respectively, for time domain analysis and 90 and 93%, respectively, for t-f domain analysis).

This paper explores an approach to study entropy differentiations of heart’s activities estimation in Low Frequency (LF) and High Frequency (HF) bands. Dataset composed of 34 ECGs, obtained from healthy and diabetic rats under normal and exercise living conditions. RR intervals extracted efficiently in order to create Heart Rate (HR) time series. Continuous Wavelet Transform (CWT) has been used, as the most appropriate approach, to evaluate the effects of exercise on healthy and diabetic HR variability (HRV). Statistical analysis performed taking into account both wavelet entropy in the low and the high frequency selected bands and the corresponding index LF/HF of the wavelet coefficients. Our results show that wavelet entropy measure based on CWT decomposition can capture significant differences between the specific frequency regions that are intrinsically related to the structure of the RR signal. According to our analysis, diabetic rats living under exercise conditions appear to have a reduced LF/HF entropy ratio compared to healthy population.

In this paper, we introduce a new suction detection approach based on online learning of a Gaussian Mixture Model (GMM) with constrained parameters to model the reduction in pump flow signals baseline during suction events. A novel three-step methodology is employed: i) signal windowing, ii) GMM based classification and iii) GMM parameter adaptation. More specifically, the first 5 second segment is used for the parameter initialization and the consequent 1 second windows are classified and used for model adaptation. The proposed approach has been tested in simulation (pump flow) signals and satisfactory results have been obtained.

The SensorART project focus on the management of heart failure (HF) patients which are treated with implantable ventricular assist devices (VADs). This work presents the way that crisp models are transformed into fuzzy in the weaning module, which is one of the core modules of the specialist's decision support system (DSS) in SensorART. The weaning module is a DSS that supports the medical expert on the weaning and remove VAD from the patient decision. Weaning module has been developed following a "mixture of experts" philosophy, with the experts being fuzzy knowledge-based models, automatically generated from initial crisp knowledge-based set of rules and criteria for weaning.

The aim of this study is to detect freezing of gait (FoG) events in patients suffering from Parkinson's disease (PD) using signals received from wearable sensors (six accelerometers and two gyroscopes) placed on the patients' body. For this purpose, an automated methodology has been developed which consists of four stages. In the first stage, missing values due to signal loss or degradation are replaced and then (second stage) low frequency components of the raw signal are removed. In the third stage, the entropy of the raw signal is calculated. Finally (fourth stage), four classification algorithms have been tested (Naive Bayes, Random Forests, Decision Trees and Random Tree) in order to detect the FoG events. The methodology has been evaluated using several different configurations of sensors in order to conclude to the set of sensors which can produce optimal FoG episode detection. Signals recorded from five healthy subjects, five patients with PD who presented the symptom of FoG and six patients who suffered from PD but they do not present FoG events. The signals included 93 FoG events with 405.6s total duration. The results indicate that the proposed methodology is able to detect FoG events with 81.94% sensitivity, 98.74% specificity, 96.11% accuracy and 98.6% area under curve (AUC) using the signals from all sensors and the Random Forests classification algorithm.

ECG is one of the most common signals used in medical practice due to its noninvasive nature and the information it contains. Several systems and various automated approaches have been developed that use computer technology to provide ECG diagnosis. These systems detect abnormalities and other features in the ECG signal and produce a decision which helps the physician when performing diagnosis. ECG decision support systems can serve as a diagnostic tool for specific cardiac anomalies such as myocardial ischaemia and arrhythmia.

ObjectiveIn this study, a methodology is presented for an automated levodopa-induced dyskinesia (LID) assessment in patients suffering from Parkinson's disease (PD) under real-life conditions. Methods and MaterialThe methodology is based on the analysis of signals recorded from several accelerometers and gyroscopes, which are placed on the subjects' body while they were performing a series of standardised motor tasks as well as voluntary movements. Sixteen subjects were enrolled in the study. The recordings were analysed in order to extract several features and, based on these features, a classification technique was used for LID assessment, i.e. detection of LID symptoms and classification of their severity. ResultsThe results were compared with the clinical annotation of the signals, provided by two expert neurologists. The analysis was performed related to the number and topology of sensors used; several different experimental settings were evaluated while a 10-fold stratified cross validation technique was employed in all cases. Moreover, several different classification techniques were examined. The ability of the methodology to be generalised was also evaluated using leave-one-patient-out cross validation. The sensitivity and positive predictive values (average for all LID severities) were 80.35% and 76.84%, respectively. ConclusionsThe proposed methodology can be applied in real-life conditions since it can perform LID assessment in recordings which include various PD symptoms (such as tremor, dyskinesia and freezing of gait) of several motor tasks and random voluntary movements.

An automated methodology for Levodopa-induced dyskinesia (LID) assessment is presented in this paper. The methodology is based on the analysis of the signals recorded from accelerometers and gyroscopes, which are placed on certain positions on the subject's body. The obtained signals are analyzed and several features are extracted. Based on these features a classification technique is used for LID detection and classification of its severity. The method has been evaluated using a group of 10 subjects. Results are presented related to each individual sensor as well as for various sensor combinations. The obtained results indicate high classification ability (93.73% classification accuracy). Conf Proc IEEE Eng Med Biol Soc

In this work a decision support tool for optimal administration of Levodopa (L-Dopa) in patients with Parkinson's Disease (PD) is presented. The so called Medication Change Proposer (MCP) is part of the PERFORM system, which monitors patient's motion status using a set of wearable sensors, assesses PD symptoms and their severity through dedicated symptoms detection modules (On-Off detection and Levodopa-Induced Dyskinesia detection), and proposes treatment plans to alleviate the symptoms. The methodology is based on the analysis of the outputs of the symptoms detection modules, taking into account the uncertainty (confidence) propagated by their respective classification algorithms. The evaluation of our methodology is performed using artificial follow-up data produced by a specially built simulation environment. The obtained results demonstrate that the MCP outputs compare well with the experts' decisions, indicating high performance.

In this work, an efficient method for spot addressing in images, which are generated by the scanning of hexagonal structured microarrays, is proposed. Initially, the blocks of the image are separated using the projections of the image. Next, all the blocks of the image are processed separately for the detection of each spot. The spot addressing procedure begins with the detection of the high intensity objects, which are probably the spots of the image. Next, the Growing Concentric Hexagon algorithm, which uses the properties of the hexagonal grid, is introduced for the detection of the non-hybridized spots. Finally, the Voronoi diagram is applied to the centers of the detected spots for the gridding of the image. The method is evaluated using spots generated from the scanning of the Beadchip of Illumina, which is used for the detection of Single Nucleotide Polymorphisms in the human genome, and uses hexagonal structure for the location of the spots. For the evaluation, the detected centers for each of the spot in the image are compared to the centers of the annotation, obtaining up to 98% accuracy for the spot addressing procedure.

In this paper we propose an automated method for Levodopa-induced dyskinesia (LID) detection and classification of its severity. The method is based on the analysis of the signals recorded from accelerometers which are placed on certain positions on the patient’s body. The signals are analyzed using a moving window and several features are extracted. Based on these features a decision tree is used to detect if LID symptoms occur and classify them related to their severity. The method has been evaluated using a group of patients and the obtained results indicate high classification ability (95% classification accuracy). Furthermore, extensive evaluation has been done in order to determine the optimal positioning of the sensors and the selection of the classification algorithm.

In this paper we propose a novel approach for the simplification of a fuzzy model. Initially, we employ a methodology for the automated generation of fuzzy models based on decision trees. The methodology is realized in three stages. Initially, a crisp model is created from a decision tree, induced from the data. Then, the crisp model it is transformed to a fuzzy one. Finally, in the third stage, all parameters entering the fuzzy model are optimized. The simplification technique is based on the pruning of the initial decision tree. The proposed approach is applied for diabetes diagnosis and the obtained results indicate its efficiency and effectiveness.

This article addresses the decision support regarding cardiovascular diseases, using computer-based methods, focusing on the coronary artery disease (CAD) diagnosis and on the prediction of clinical restenosis in patients undergoing angioplasty. Methods reported in the literature are reviewed with respect to (i) the medical information that are employing in order to reach the diagnosis and (ii) the data analysis techniques used for the creation of the CDSSs. In what concerns medical information, easily and noninvasively-obtained data present several advantages compared to other types of data, while data analysis techniques that are characterized by transparency regarding their decisions are more suitable for medical decision making. A recently developed approach that complies with the above requirements is presented. The approach is based on data mining and fuzzy modelling. Using this approach, one CDSS has been developed for each of the two cardiovascular problems mentioned above. These CDSSs are extensively evaluated and comments about the discovered knowledge are provided by medical experts. The later is of great importance in designing and evaluating CDSSs, since it allows them to be integrated into real clinical environments.

This paper introduces an automated methodology for the extraction of fetal heart rate from cutaneous potential abdominal electrocardiogram (abdECG) recordings. A three-stage methodology is proposed. Having the initial recording, which consists of a small number of abdECG leads in the first stage, the maternal R-peaks and fiducial points (QRS onset and offset) are detected using time-frequency (t-f) analysis and medical knowledge. Then, the maternal QRS complexes are eliminated. In the second stage, the positions of the candidate fetal R-peaks are located using complex wavelets and matching theory techniques. In the third stage, the fetal R-peaks, which overlap with the maternal QRS complexes (eliminated in the first stage) are found using two approaches: a heuristic algorithm technique and a histogram-based technique. The fetal R-peaks detected are used to calculate the fetal heart rate. The methodology is validated using a dataset of eight short and ten long-duration recordings, obtained between the 20th and the 41st week of gestation, and the obtained accuracy is 97.47%. The proposed methodology is advantageous, since it is based on the analysis of few abdominal leads in contrast to other proposed methods, which need a large number of leads.

A methodology for the automated development of fuzzy expert systems is presented. The idea is to start with a crisp model described by crisp rules and then transform them into a set of fuzzy rules, thus creating a fuzzy model. The adjustment of the model's parameters is performed via a stochastic global optimization procedure. The proposed methodology is tested by applying it to problems related to cardiovascular diseases, such as automated arrhythmic beat classification and automated ischemic beat classification, which, besides being well-known benchmarks, are of particular interest due to their obvious medical diagnostic importance. For both problems, the initial set of rules was determined by expert cardiologists, and the MIT-BIH arrhythmia database and the European ST-T database are used for optimizing the fuzzy model's parameters and evaluating the fuzzy expert system. In both cases, the results indicate an escalation of the performance from the simple initial crisp model to the more sophisticated fuzzy models, proving the scientific added value of the proposed framework. Also, the ability to interpret the decisions of the created fuzzy expert systems is a major advantage compared to "black box" approaches, such as neural networks and other techniques.

In this work, we propose a method for the automated expert system creation. The method is based on the integration of global knowledge (i.e. knowledge from the field experts) and local knowledge (i.e. knowledge derived from the available data) in a single inference engine. Starting from an initial set of rules (expert's knowledge) and an annotated dataset, data mining is performed to the dataset and a second set of rules is acquired. Both of them are integrated into a single set of rules. Fuzzy modeling is then applied to the rules, transforming them into a fuzzy model, and finally, an optimization technique is used to tune the fuzzy model's parameters. The method is applied to a medical domain problem, the cardiac arrhythmic beat classification and satisfactory results have been obtained. The method experiences several advantages compared to approaches based solely on expert's knowledge or mined knowledge while the ability to interpret the decisions made from the created fuzzy expert system is a major advantage compared to ";black box"; approaches.

In this paper we present a methodology for biosequence classification, which employs sequential pattern mining and optimization algorithms. In the first stage, a sequential pattern mining algorithm is applied to a set of biological sequences and the sequential patterns are extracted. Then, the score of each pattern with respect to each sequence is calculated using a scoring function and the score of each class under consideration is estimated. The scores of the patterns and classes are updated, multiplied by a weight. In the second stage an optimization technique is employed to calculate the weight values to achieve the optimal classification accuracy. The methodology is applied in the protein class and fold prediction problem. Extensive evaluation is carried out, using a dataset obtained from the Protein Data Bank.

In this work, three different methodologies for fuzzy expert systems creation are compared: a well-known neuro- fuzzy approach, a knowledge-based approach and a novel methodology, based on rule-extraction. The adaptive neuro- fuzzy information system (ANFIS) is used to automatically generate a fuzzy expert system. In the knowledge-based approach and the rule-extraction methodology, the idea is to start with a model described by crisp rules, provided by medical experts in the first case or extracted using data mining techniques in the second, and then to transform them into a set of fuzzy rules, creating a fuzzy model. In either case, the adjustment of the model's parameters is performed via a stochastic global optimization procedure. All three approaches are applied to a medical domain problem, the cardiac arrhythmic beat classification. The ability to interpret the decisions made from the created fuzzy expert systems is a major advantage compared to other "black box" approaches.