DEVELOPMENT OF A PARTIAL SUPERVISION STRATEGY TO AUGMENT A NEAREST NEIGHBOUR CLUSTERING ALGORITHM FOR BIOMEDICAL DATA CLASSIFICATION

Sameh A. Salem, Nancy M. Salem and Asoke K. Nandi

In this paper, a partial supervision strategy for a recently developed clustering algorithm NNCA (Salem et al., 2006), Nearest Neighbour Clustering Algorithm, is proposed. The proposed method (NNCA-PS) offers classification capability with smaller amount of a priori knowledge, where a small number of data objects from the entire dataset are used as labelled objects to guide the clustering process towards a better search space. Results from the proposed supervision method indicate its robustness in classification compared with other classifiers.

A REGION BASED METHODOLOGY FOR FACIAL EXPRESSION RECOGNITION

Anastasios C. Koutlas and Dimitrios I. Fotiadis

Facial expression recognition is an active research field which accommodates the need of interaction between humans and machines in a broad field of subjects. This work investigates the performance of a multi-scale and multi-orientation Gabor Filter Bank constructed in such a way to avoid redundant information. A region based approach is employed using different neighbourhood size at the locations of 34 fiducial points. Furthermore, a reduced set of 19 fiducial points is used to model the face geometry. The use of Principal Component Analysis (PCA) is evaluated. The proposed methodology is evaluated for the classification of the 6 basic emotions proposed by Ekman considering neutral expression as the seventh emotion.

BIOSIGNAL-BASED COMPUTING BY AHL INDUCED SYNTHETIC GENE REGULATORY NETWORKS - From an in vivo Flip-Flop Implementation to Programmable Computing Agents

T. Hinze, T. Lenser, N. Matsumaru, P. Dittrich and S. Hayat

Gene regulatory networks (GRNs) form naturally predefined and optimised computational units envisioned to act as biohardware able to solve hard computational problems efficiently. This interplay of GRNs via signalling pathways allows the consideration as well as implementation of interconnection-free and fault tolerant programmable computing agents. It has been quantitatively shown in an in vivo study that a reporter gene encoding the green fluorescent protein (gfp) can be switched between high and low expression states, thus mimicking a NAND gate and a RS flip-flop. This was accomplished by incorporating the N-acyl homoserine lactone (AHL) sensing lux operon from Vibrio fischeri along with a toggle switch in Escherichia coli. gfp expression was quantified using flow cytometry. The computational capacity of this approach is extendable by coupling several logic gates and flip-flops. We demonstrate its feasibility by designing a finite automaton capable of solving a knapsack problem instance.

IMAGE SEGMENTATION TO EVALUATE ISLETS OF LANGHERANS

C. Grimaudo, D. Tegolo, C. Valenti and F. Bertuzzi

This contribution deals with an unsupervised system to process digital photomicrographs in order to locate and analyze islets of Langherans in human pancreases. The experiment has been conducted on real data and, though we are still going to complete the evaluation of the whole method, we expect to define a set of proper features (e.g. area, perimeter, fractal dimension, shape complexity, texture and entropy) useful for a fast and reliable counting of healthy cells. In particular, this research aims to measure the advisability of a possible implantation in patients affected by type 1 diabetes mellitus.

TRADITIONAL AVERAGING, WEIGHTED AVERAGING, AND ERPSUB FOR ERP DENOISING IN EEG DATA - A Comparison of the Convergence Properties

Andriy Ivannikov, Tommi Kärkkäinen, Tapani Ristaniemi and Heikki Lyytinen

In this article we compare the convergence rates of the three methods applied in ElectroEncephaloGraphy research for ERP denoising: traditional averaging, weighted averaging and ERPSUB. We derive the weighted averaging procedure based on maximizing SNR and show thereby that SNR criterion is equivalent to the originally proposed mean-square error criterion in the sense of the weighted averaging problem solving. Moreover, in order to characterize fully the performance of the selected methods we compare also noise reduction rates.

NOISE REDUCTION AND VOICE SEPARATION ALGORITHMS APPLIED TOWOLF POPULATION COUNTING

B. Dugnol, C. Fernández, G. Galiano and J. Velasco

We use signal and image theory based algorithms to produce estimations of the number of wolves emitting howls or barks in a given field recording as an individuals counting alternative to the traditional trace collecting methodologies. We proceed in two steps. Firstly, we clean and enhance the signal by using PDE based image processing algorithms applied to the signal spectrogram. Secondly, assuming that the wolves chorus may be modelled as an addition of nonlinear chirps, we use the quadratic energy distribution corresponding to the Chirplet Transform of the signal to produce estimates of the corresponding instantaneous frequencies, chirp-rates and amplitudes at each instant of the recording. We finally establish suitable criteria to decide how such estimates are connected in time.

BIOMIMETICS AND PROPORTIONAL NOISE IN MOTOR CONTROL

Christopher M. Harris

Proportional noise, in which the standard deviation of signal noise is proportional to signal mean, is a fundamental constraint on human motor performance but why it occurs is unknown. We show that for neural networks with binary thresholded units, channel capacity is maximised with a recruitment strategy that produces PN. The size principle also emerges, in agreement with observation. We therefore argue that Fitt’s law, speed-accuracy trade-off, and the minimum variance trajectories (including minimum jerk trajectories for limiting brief movements), which are observed in most human point-to-point movements, have evolved as optimal strategies resulting from maximising channel capacity. We conclude that biomimicry of minimum variance and minimum jerk trajectories in robotics is probably only of aesthetic value when using standard technology. In contrast, biomimicry using emergent neuromorphic technology in which networks are built from stochastic silicon ‘neurons’ with thresholds, is functional biomimetics and optimization of channel capacity will produce behaviours that are human-like.

A VOCAL TRACT VISUALISATION TOOL FOR A COMPUTER-BASED SPEECH TRAINING AID FOR HEARING-IMPAIRED INDIVIDUALS

Abdulhussain E. Mahdi

This paper describes a computer-based software tool for visualisation of the vocal-tract, during speech articulation, by means of a mid-sagittal view of the human head. The vocal tract graphics are generated by estimating both the area functions and the formant frequencies from the acoustic speech signal. First, it is assumed that the speech production process is an autoregressive model. Using a linear prediction analysis, the vocal tract area functions and the first three formants are estimated. The estimated area functions are then mapped to corresponding mid-sagittal distances and displayed as 2D vocal tract lateral graphics. The mapping process is based on a simple numerical algorithm and an accurate reference grid derived from x-rays for the pronunciation of a number English vowels uttered by different speakers. To compensate for possible errors in the estimated area functions due to variation in vocal tract length between speakers, the first two sectional distances are determined by the three formants. Experimental results show high correlation with x-ray data and the PARAFAC analysis. The tool also displays other speech parameters that are closely related to the production of intelligible speech and hence would be useful as a visual feedback aid for speech training of hearing–impaired individuals.

IDENTIFICATION OF HAND MOVEMENTS BASED ON MMG AND EMG SIGNALS

Pawel Prociow, Andrzej Wolczowski, Tito G. Amaral, Octávio P. Dias and Joaquim Filipe

This paper proposes a methodology that analysis and classifies the EMG and MMG signals using neural networks to control prosthetic members. Finger motions discrimination is the key problem in this study. Thus the emphasis is put on myoelectric signal processing approaches in this paper. The EMG and MMG signals classification system was established using the LVQ neural network. The experimental results show a promising performance in classification of motions based on both EMG and MMG patterns.

BIO-INSPIRED DATA AND SIGNALS CELLULAR SYSTEMS

André Stauffer, Daniel Mange and Joël Rossier

Living organisms are endowed with three structural principles: multicellular architecture, cellular division, and cellular differentiation. Implemented in digital according to these principles, our data and signals cellular systems present self-organizing mechanisms like configuration, cloning, cicatrization, and regeneration. These mechanisms are made of simple processes such as growth, load, branching, repair, reset, and kill. The data processed in the self-organizing mechanisms and the signals triggering their underlying processes constitute the core of this paper.

APPLICATION OF WALSH TRANSFORM BASED METHOD ON TRACHEAL BREATH SOUND SIGNAL SEGEMENTATION

Jin Feng, Farook Sattar and Moe Pwint

This paper proposes a robust segmentation method for differentiating consecutive inspiratory/expiratory episodes of different types of tracheal breath sounds. This has been done by applying minimal Walsh basis functions to transform the original input respiratory sound signals. Decision module is then applied to differentiate transformed signal into respiration segments and gap segments. The segmentation results are improved through a refinement scheme by new evaluation algorithm which is based on the duration of the segment. The results of the experiments, which have been carried out on various types of tracheal breath sounds, show the robustness and effectiveness of the proposed segmentation method.

A NEW METHOD FOR DETECTION OF BRAIN STEM IN TRANSCRANIAL ULTRASOUND IMAGES

Josef Schreiber, Eduard Sojka, Lacezar Licev, Petra Sknourilova, David Skoloudik and Jan Gaura

Transcranial sonography is to date only method able to detect structural damage of brain tissue in Parkinson’s disease patients. The problem is that the images provided by this method often suffer from a very poor quality what makes the final diagnosis strongly dependent on experience of examinating medical doctor. Our objective is to create a method that should help to minimize the physician’s subjectivity in the final diagnosis and should provide more exact information about the processed ultrasound images. The method itself is divided into two phases. In a first one, we try to locate the position of a minimal window, containing the brain stem, in an analyzed image. In a second phase, we locate and measure the echogenic substantia nigra area.

ANALYSIS OF DIFFERENCES BETWEEN SPECT IMAGES OF THE LEFT AND RIGHT CEREBRAL HEMISPHERES IN PATIENTS WITH EPILEPTIC SYMPTOMS

Elżbieta Olejarczyk and Małgorzata Przytulska

The aim of his work was examination of asymmetries in activity of the left and right cerebral hemispheres as well as localization and contouring of the regions of reduced or increased activity on the basis of single photon emission computer tomography (SPECT) images. The mean and standard deviation of normalized intensities inside the contoured areas of images, entropy based on intensity histograms and Chen’s fractal dimension were calculated.

A NEW METHOD FOR ICG CHARACTERISTIC POINT DETECTION

Maria Rizzi, Matteo D'Aloia and Beniamino Castagnolo

Impedance Cardiography is a cost-effective, non-invasive technique particularly useful in measuring cardiac functions. It evaluates systolic time intervals and stroke volume measuring thorax bioimpedance. In this paper, adopting the time-frequency analysis method, a new design has been developed to study the first derivative of impedance cardiography signal. The application of parallel wavelet filter banks has been investigated and a new method for ICG signal characteristic point detection has been developed. Test results show the improvement of the method in sensitivity and the feasibility of an easy implementation by design tools. Moreover, the algorithm noise immunity has been investigated.

MOTION ESTIMATION IN MEDICAL IMAGE SEQUENCES USING INVERSE POLYNOMIAL INTERPOLATION

Saleh Al-Takrouri and Andrey Savkin

In this paper, we propose a new method for motion estimation between two successive frames in medical image sequences and videos where the problem is defined in terms of pixel correspondence. The method is based on solving the problem of inverse polynomial interpolation and the solution is presented in the form of an iterative formula that numerically estimates the horizontal and vertical displacements of pixels between the two images. Examples are provided to show the performance of the proposed method.

PHASE SEGMENTATION OF NOISY RESPIRATORY SOUND SIGNALS USING GENETIC APPROACH

Feng Jin, Farook Sattar and Moe Pwint

In this paper, a new approach to automatically segment noisy respiratory sound signals is proposed. Segmentation is formulated as an optimization problem and the boundaries of the signal segments are detected using a genetic algorithm (GA). As the estimated number of segments present in a segmenting signal is initially obtained, a multi-population GA is employed to determine the locations of segment boundaries. The segmentation results are found through the generations of GA by introducing a new evaluation function, which is based on the sample entropy and a heterogeneity measure. Illustrative results for respiratory sound signals contaminated by loud heartbeats and other high level noises show that the proposed genetic segmentation method is quite accurate and threshold independent to find the noisy respiratory segments as well as the pause segments under different noisy conditions.

EFFECTIVENESS FOR A SLEEPINESS TEST OF PUPIL SIZE ESTIMATION DURING BLINK

Minoru Nakayama, Keiko Yamamoto and Fumio Kobayashi

Pupillary response has been used for an index of sleepiness, but the validity of the index is not clear. In this paper, the influence of blinks on the Pupillary Unrest Index (PUI) and the Power Spectrum Density (PSD) for the frequency range $f<0.8Hz$, as indices of pupil's instability during a sleepiness test, was examined. To estimate pupil size during blink, a procedure for collecting the clinical data was developed using Support Vector Regression (SVR). The values of PUI increased with experimental time, and the values and deviations of PUI for experimental observation were larger than the ones with SVR estimation. The blink time also increased with experimental time, and there were significant correlation relationships between the value of PUI and blink time. The mean PSD also correlated significantly with blink time. The relationship between pupillary indices and a subjective sleepiness index was not significant, as it was not in other previous works. These results provide evidence that pupillary indices were significantly affected by blink, and they did not reflect sleepiness correctly.

AUTOMATIC SEGMENTATION OF CAPILLARY NON-PERFUSION IN RETINAL ANGIOGRAMS

Amit Agarwal, Jayanthi Sivaswamy, Alka Rani and Taraprasad Das

Capillary Non-Perfusion (CNP) is a condition in diabetic retinopathy where blood ceases to flow to certain parts of the retina, potentially leading to blindness. This paper presents a solution for automatically detecting and segmenting CNP regions from fundus fluorescein angiograms (FFAs). CNPs are modeled as valleys, and a novel multi resolution technique for trough-based valley detection is presented. The proposed algorithm has been tested on 40 images and validated against expert-marked ground truth. Obtained results are presented as a receiver operating characteristic (ROC) curve. The area under this curve is 0.842 and the distance of ROC from the ideal point (0,1) is 0.31.

ECG SIGNAL DENOISING - Using Wavelet in Besov Spaces

Shi Zhao, Yiding Wang and Hong Yang

This paper proposes a novel technique to eliminate the noise in practical electrocardiogram (ECG) signals. Using wavelet bases to reduce the noise is a state-of-the-art denoising technique, which is first presented by Donoho and Johnstone. Traditional algorithms discuss wavelets in spaces. Compared to them, the proposed technique projects the ECG signals onto Besov spaces, which is a more sophisticated smoothness space, in order to determine the threshold of shrinkage function. In addition, instead of using linear shrinkage function, the proposed algorithm uses nonlinear hyper shrinkage function, which is proposed by S. Poornachandra. Combining the two techniques, we obtain a significant improvement over conventional wavelet denoising algorithm.

ELASTIC IMAGE WARPING USING A NEW RADIAL BASIC FUNCTION WITH COMPACT SUPPORT

Zhixiong Zhang and Xuan Yang

Thin plate spline (TPS) and compact support radial basis functions (CSRBF) are well-known and successful tools in medical image elastic registration base on landmark. TPS minimizes the bending energy of the whole image. However, in real application, such scheme would deform the image globally when deformation is local. Although CSRBF can limit the effect of the deformation locally, it cost more bending energy which means more information was lost. A new radial basic function named ‘Compact Support Thin Plate Spline Radial Basic Function’ (CSTPF) has been proposed in this paper. It costs less bending energy than CSRBF in deforming image locally and its global deformation effect is similar to TPS. Numerous experimental results show that CSTPF performs outstanding in both global and local image deformation.

TWO-STAGE CLUSTERING OF A HUMAN BRAIN TUMOUR DATASET USING MANIFOLD LEARNING MODELS

Raúl Cruz-Barbosa and Alfredo Vellido

This paper analyzes, through clustering and visualization, Magnetic Resonance spectra corresponding to a complex multi-center human brain tumour dataset. Clustering is performed as a two-stage process, in which the models used in the first stage are variants of Generative Topographic Mapping (GTM), belonging to the Manifold Learning family. In semi-supervised settings, class information can be added to refine the clustering process. Class information-enriched variants of GTM are used in this study to obtain a primary cluster description of the data. The number of clusters used by GTM is usually large for visualization purposes and does not necessarily correspond to the overall class structure. Consequently, in a second stage, clusters are agglomerated using the K-means algorithm with different initialization strategies, some of them defined ad hoc for the GTM models. We aim to evaluate whether the use of class information influences brain tumour cluster-wise class separability in the final result of the two-stage clustering process and under what circumstances this may be the case. We also explore the existence of atypical cases in the dataset and resort to a robust variant of GTM that detects outliers while effectively minimizing their negative impact in the clustering process.

TREMOR CHARACTERIZATION - Algorithms for the Study of Tremor Time Series

E. Rocon, A. F. Ruiz, J. C. Moreno, J. L. Pons, J. A. Miranda and A. Barrientos

A great deal of effort has been devoted in the past decades in the generic area of tremor management. Specific topics of modelling for objective classification of pathological tremor out of kinematics and physiological data, compensatory technologies and evaluation rating tools are just a few examples of application field. This paper introduces the work developed by the authors in the study of tremor time series. First, it introduces a novel technique for the study of tremor. The technique presented is a high-resolution technique that solves most of limitations of the Fourier Analysis (the standard technique to the study of tremor time series). This technique was used for the study of tremorous movement in joints of the upper limb. After, some conclusions about tremor behaviour in upper limb based on the technique introduces are presented. Furthermore, an algorithm able to estimated in real-time the voluntary and the tremorous movement was presented. This algorithm was validated in two contexts with successful results. Finally, some conclusions and future work are given.

ACOUSTIC INDICES OF CARDIAC FUNCTIONALITY

Guy Amit, Jonathan Lessick, Noam Gavriely and Nathan Intrator

The mechanical processes of the cardiac cycle generate vibratory and acoustic signals that are received on the chest wall. We describe signal processing and feature extraction methods utilizing these signals for continuous non-invasive monitoring of systolic cardiac functionality. Vibro-acoustic heart signals were acquired from eleven subjects during a routine pharmacological stress echocardiography test. Principal component analysis, applied to the joint time-frequency distribution of the first heart sound (S1), revealed a pattern of an increase in the spectral energy and the frequency bandwidth of the signal associated with the increase of cardiac contractility during the stress test. Novel acoustic indexes of S1 that compactly describe this pattern showed good linear correlation with reference indexes of systolic functionality estimated by strain-echocardiography. The acoustic indexes may therefore be used to improve monitoring and diagnosis of cardiac systolic dysfunction.

ANALYSIS OF FOCUSES OF ATTENTION DISTRIBUTION FOR A NOVEL FACE RECOGNITION SYSTEM

C. Spampinato, M. Nicotra and A. Travaglianti

In this paper we propose an automated approach to recognize human faces based on the analysis of the distribution of the focuses of attention (FOAs) that reproduces the ability of the humans in the interpretation of visual scenes. The analysis of the FOAs (distribution and position), carried out by an efficient and source light independent visual attention module, allows us to integrate the face features (e.g., eyes, nose, mouth shape) and the holistic features (the relations between the various parts of the face). Moreover, a remarkable approach has been developed for skin recognition based on the shifting of the Hue plane in the HSL color space.

REGISTRATION AND RETRIEVAL OF ELONGATED STRUCTURES IN MEDICAL IMAGES

Alexei Manso Correa Machado and Christiano Augusto Caldas Teixeira

This work aims at proposing a set of methods to describe, register and retrieve images of elongated structures from a database based on their shape content. We propose a registration algorithm that jointly takes into account the gross shape of the structure and the shape of its boundary, resulting in anatomically consistent deformations. The method determines a medial axis that represents the full extent of the structure with no branches. Registration follows the linear elasticity model and is implemented through dynamic programming. Discriminative anatomic features are computed from the results of registration and used as variables in a content-based image retrieval system. A case study on the morphology of the corpus callosum in the chromosome 22q11.2 deletion syndrome illustrates the effectiveness of the method and corroborates the hypothesis that retrieval systems may also act as knowledge discovery tools.

NONLINEAR MODELING OF CARDIOVASCULAR RESPONSE TO EXERCISE

Lu Wang, Steven W. Su, Gregory S. H. Chan, Branko G. Celler, Teddy M. Cheng and Andrey V. Savkin

This study experimentally investigates the relationships between central cardiovascular variables and oxygen uptake based on nonlinear analysis and modeling. Ten healthy subjects were studied using cycle-ergometry exercise tests with constant workloads ranging from 25 Watt to 125 Watt. Breath by breath gas exchange, heart rate, cardiac output, stroke volume and blood pressure were measured at each stage. The modeling results proved that the nonlinear modeling method (Support Vector Regression) outperforms traditional regression method (reducing Estimation Error between 59% and 80%, reducing Testing Error between 53% and 72%) and is the ideal approach in the modeling of physiological data, especially with small training data set.

NONLINEAR MODELLING AND CONTROL OF HEART RATE RESPONSE TO TREADMILLWALKING EXERCISE

Teddy M. Cheng, Andrey V. Savkin, Branko G. Celler, Steven W. Su and Lu Wnag

In this study, a nonlinear system was developed for the modelling of the heart rate response to treadmill walking exercise. The model is a feedback interconnected system which can represent the neural response and peripheral local response to exercise. The parameters of the model were identified from an experimental study which involved 6 healthy adult male subjects, each completed 3 sets of walking exercise at different speeds. The proposed model will be useful in explaining the cardiovascular response to exercise. Based on the model, a 2-degree-of-freedom controller was developed for the regulation of the heart rate response during exercise. The controller consists of a piecewise LQ and $H_{\infty}$ controllers. Simulation results showed that the proposed controller had the ability to regulate heart rate at a given target, indicating that the controller can play an important role in the design of exercise protocols for individuals.

BREAST CANCER DETECTION USING GENETIC PROGRAMMING

Hong Guo, Qing Zhang and Asoke K. Nandi

Breast cancer diagnosis have been investigated by different machine learning methods. This paper proposes a new method for breast cancer diagnosis using a single feature generated by Genetic Programming(GP). GP as an evolutionary mechanism that provides a training structure to generate features. The presented approach is experimentally compared with some kernel feature extraction methods: The Kernel Principal Component Analysis (KPCA) and Kernel Generalised Discriminant Analysis (KGDA). Results demonstrate the capability of this method to transform information from high dimensional feature space into one dimensional space for breast cancer diagnosis.

BREAST CANCER DIAGNOSIS AND PROGNOSIS USING DIFFERENT KERNEL-BASED CLASSIFIERS

Tingting Mu and Asoke Nandi

The medical applications of several advanced, kernel-based, nonlinear classifiers to breast cancer diagnosis and prognosis are studied and compared in this paper. The pairwise Rayleigh quotient (PRQ) classifier and kernel Fisher’s discriminative analysis (KFDA) seek one discriminant boundary based on the scatter measurements. The support vector machines (SVMs) seek one discriminant boundary based on the maximal margin rule. The strict 2-surface proximal (S2SP) classifier and multisurface proximal SVMs (MPSVMs) learn two proximal hyperplanes by optimizing two Rayleigh quotients. The Radial basis function (RBF) kernel is employed to incorporate the nonlinearity. Studies are conducted with the Wisconsin diagnosis and prognosis breast cancer (WDBC and WPBC) datasets generated from fine needle aspiration (FNA) samples by image processing. Comparative analysis are developed on the classification accuracies, computing times, and sensitivities to regularization parameters for the above kernel-based classifiers.

AN EFFICIENT METHOD FOR VESSEL WIDTH MEASUREMENT ON COLOR RETINAL IMAGES

Alauddin Bhuiyan, Baikunth Nath, Joselito Chua and Kotagiri Ramamohanarao

Vessel diameter is an important factor for indicating retinal microvascular signs. In automated retinal image analysis, the measurement of vascular width is a complicated process as most of the vessels are few pixels wide. In this paper, we propose a new technique to measure the retinal blood vessel diameter which can be used to detect arteriolar narrowing, arteriovenous (AV) nicking, branching coefficients, etc. to diagnose related diseases. First, we apply the Adaptive Region Growing (ARG) segmentation technique to obtain the edges of the blood vessels. Following that we apply the unsupervised texture classification method to segment the blood vessels from where we obtain the vessel centreline. Then we utilize the edge image and vessel centreline image to obtain the potential pixels pairs which pass through a centreline pixel. We apply a rotational invariant mask to search the pixel pairs from the edge image. From those pixels we calculate the shortest distance pair which will be the vessel width for that cross-section. We evaluate our technique with manually measured width for different vessels' cross-sectional area which shows that our technique is very accurate.

MODEL ORDER ESTIMATION FOR INDEPENDENT COMPONENT ANALYSIS OF EPOCHED EEG SIGNALS

Peter Mondrup Rasmussen,Morten Mørup, Lars Kai Hansen and Sidse M. Arnfred

In analysis of multi-channel event related EEG signals indepedent component analysis (ICA) has become a widely used tool to attempt to separate the data into neural activity, physiological and non-physiological artifacts. High density elctrode systems offer an opportunity to estimate a corresponding large number of independent components (ICs). However, too large a number of ICs leads to overfitting of the ICA model, which can have a major impact on the model validity. Consequently, finding the optimal number of components in the ICA model is an important problem. In this paper we present a method for model order selection, based on a probabilistic framework. The proposed method is a modification of the Molgedey Schuster (MS) algorithm to epoched, i.e. event related data. Thus, the contribution of the present paper can be summarized as follows: 1) We advocate MS as a low complexity ICA alternative for EEG. 2) We define an epoch based likelihood function for estimation of a principled unbiased 'test error'. 3) Based on the unbiased test error measure we perform model order selection for ICA of EEG. Applied to a 64 channel EEG data set we were able to determine an optimum order of the ICA model and to extract 22 ICs related to the neurophysiological stimulus responses as well as ICs related to physiological- and non-physiological noise. Furthermore, highly relevant high frequency responce information was captured by the ICA model.

USE OF CEPSTRUM-BASED PARAMETERS FOR AUTOMATIC PATHOLOGY DETECTION ON SPEECH - Analysis of Performance and Theoretical Justification

Rubén Fraile, Juan Ignacio Godino-Llorente, Nicolás Sáenz-Lechón, Víctor Osma-Ruiz and Pedro Gómez-Vilda

The majority of speech signal analysis procedures for automatic pathology detection mostly rely on parameters extracted from time-domain processing. Moreover, calculation of these parameters often requires prior pitch period estimation; therefore, their validity heavily depends on the robustness of pitch detection. Within this paper, an alternative approach based on cepstral-domain processing is presented which has the advantage of not requiring pitch estimation, thus providing a gain in both simplicity and robustness. While the proposed scheme is similar to solutions based on Mel-frequency cepstral parameters, already present in literature, it has an easier physical interpretation while achieving similar performance standards.

BIOSIGNAL ACQUISITION DEVICE - A Novel Topology for Wearable Signal Acquisition Devices

Luca Maggi, Luca Piccini, Sergio Parini, Giuseppe Andreoni and Guido Panfili

The here presented work illustrates a novel circuit topology for the conditioning of biomedical signals. The system is composed of an amplification chain and relies on a double feedback path which assure the stability of the system whichever the amplification block gain and the order of the low-pass filter are. During the normal operation the offset recovery circuit has a linear transfer function, when it detects a saturation of the amplifier, it automatically switches to the fast recovery mode and restores the baseline in few milliseconds. The proposed configuration has been developed in order to make wearable biosignal acquisition devices more robust, simpler and smaller. Thanks to the used AC coupling method, very low high-pass cut-off frequencies, can be achieved even using small valued passive components with advantages in terms of circuit bulkiness. The noise rejection filter between the pre-amplification and the amplification stages eliminates the out-of-band noise before the amplification reducing the possibility of having clipping noise and minimizing the dynamic power consumption. The presented topology is currently used in a prototypal EEG acquisition device in a Brain Computer Interface (BCI) system, and in a commercial polygraph which will be soon certificated for clinical use.

MICROGLIA MODELLING AND ANALYSIS USING L-SYSTEMS GRAMMAR

Herbert F. Jelinek and Audrey Karperien

Medical image analysis requires in the first instance information on the extent of normal variation in a biological system in order to identify pathological changes. MicroMod is an L-system based modelling software package available through the World Wide Web that allows construction of branching structures such as neurons and glia. In addition MicroMod includes analystical software to analyse complex structures such as fractal analysis and lacunarity. MicroMod consists of three options with subroutines for constructing branching structures in a deterministic or probabilistic manner. The fractal dimensions of microglia visualised using histochmical techniques with modelled glia using MicroMod showed good agreement (1.423 and 1.425 respectively). An analysis of simulated microglia by fractal analysis indicates that changes in the length of sub-branches relative to the parent branch with the number of sprouts remaining the same and manipulating the scale of sub to parent branch diameter and the number of new branches per branch affected the fractal dimension and lacunarity. The results indicate that MicroMod provides a useful adjunct to neuroscience research and provides a platform for understanding complex changes in structure associated with normal function and disease processes.

STATISTICAL SIGNIFICANCE IN OMIC DATA ANALYSES - Alternative/Complementary Method for Efficient Automatic Identification of Statistically Significant Tests in High Throughput Biological Studies

Christine Nardini, Luca Benini and Michael D. Kuo

The post-Genomic Era is characterized by the proliferation of high-throughput platforms that allow the parallel study of a complete body of molecules in one single run of experiments (omic approach). Analysis and integration of omic data represent one of the most challenging frontiers for all the disciplines related to Systems Biology. From the computational perspective this requires, among others, the massive use of automated approaches in several steps of the complex analysis pipeline, often consisting of cascades of statistical tests. In this frame, the identification of statistical significance has been one of the early challenges in the handling of omic data and remains a critical step due to the multiple hypotheses testing issue, given the large number of hypotheses examined at one time. Two main approaches are currently used: p-values based on random permutation approaches and the False Discovery Rate. Both give meaningful and important results, however they suffer respectively from being computationally heavy -due to the large number of data that has to be generated-, or extremely flexible with respect to the definition of the significance threshold, leading to difficulties in standardization. We present here a complementary/alternative approach to these current ones and discuss performances and limitations.

PRINCIPAL COMPONENT ANALYSIS OF THE P-WAVE

Federica Censi, Giovanni Calcagnini, Pietro Bartolini, Chiara Ricci, Renato Pietro Ricci and Massimo Santini

Aim of this study is to perform the principal component analysis (PCA) of the P-wave in patients prone to atrial fibrillation (AF). Eighteen patients affected by paroxysmal AF and implanted with pacemakers were studied. Two 5-minute ECG recordings were performed: during spontaneous (SR) and paced rhythm (PR). ECG signals were acquired using a 32-lead system (2048 Hz, 24 bit, 0-400 Hz bandwidth). For each patient, PCA of the averaged P-waves extracted in any of the 32 leads has been performed. We computed PCA parameters related to the dipolar (using the first 3 PCs) and not dipolar (from the 4th to the 32nd PCs) components of the P-wave. The number of PCs according to the latent root criterion ranges between 2 and 3 during SR and between 2 and 4 during PR. PCA parameters related to the 3 largest PCs, and describing the dipolar component of the P-wave, did not significantly differ during SR and PR. The not dipolar components during SR were significantly lower than during PR (PCAres%: 0.03±0.06 vs 0.12±0.21, p=0.001; PCAres [mV4]: 0.10±0.14 vs 0.49±0.73, p=0.001). Factor analysis showed that on average all leads contributes to the first principal component. These findings encourage the use of PCA to obtain crucial quantitative information from surface ECG P-wave.

SPECTRAL AND CROSS-SPECTRAL ANALYSIS OF CONDUCTANCE CATHETER SIGNALS - New Indexes for Quantification of Mechanical Dyssinchrony

Sergio Valsecchi, Luigi Padeletti, Giovanni Battista Perego, Federica Censi, Pietro Bartolini and Jan J. Schreuder

We hereby present novel indexes to quantify ventricular mechanical dyssynchrony by using spectral and cross-spectral analysis of conductance catheter volume signals. Conductance catheter is a volume measurement technique based on conductance measurement: the intraventricular volume, i.e. the time-varying volume of blood contained within the heart cavity, is estimated by measuring the electrical conductance of the blood employing a multi-pole catheter. Five segmental volume signals (SVi, i=1,…5) can be acquired; total volume (TV) is estimated as the instantaneous sum of the segmental volumes. We implemented classical time-domain dyssynchrony indexes already utilized in conductance catheter signals analysis, and new frequency-domain indexes. Study population consisted of 15 heart failure (HF) patients with left bundle branch block and 12 patients with preserved left ventricular (LV) function. We found that spectral measures seem to out-perform classical time-domain parameters in differentiating atrial HF patients from no-HF group. These findings encourage the use of spectral analysis to obtain crucial quantitative information from conductance catheter signals.

EVOLUTIONARY COMPUTATION APPROACH TO ECG SIGNAL CLASSIFICATION

Farid Melgani and Yakoub Bazi

In this paper, we propose a novel classification system for ECG signals based on particle swarm optimization (PSO). The main objective of this system is to optimize the performance of the support vector machine (SVM) classifier in terms of accuracy by automatically: i) searching for the best subset of features where to carry out the classification task; and ii) solving the SVM model selection issue. Experiments conducted on the basis of ECG data from the MIT-BIH arrhythmia database to classify five kinds of abnormal waveforms and normal beats confirm the effectiveness of the proposed PSO-SVM classification system.

COMPARATIVE STUDY OF SEVERAL NOVEL ACOUSTIC FEATURES FOR SPEAKER RECOGNITION

Vladimir Pervouchine, Graham Leedham, Haishan Zhong, David Cho and Haizhou Li

Finding good features that represent speaker identity is an important problem in speaker recognition area. Recently a number of new and novel acoustic features have been proposed for speaker recognition. The researchers use different data sets and sometimes different classifiers to evaluate the features and compare them to the baselines such as MFCC or LPCC. However, due to different experimental conditions direct comparison of those features to each other is difficult or impossible. This paper presents a study of five new acoustic features recently proposed. The feature extraction has been performed on the same data (NIST~2001~SRE), and the same UBM-GMM classifier has been used. The results are presented as DET curves with equal error ratios indicated. Also, an SVM-based combination of GMM scores produced on different features has been made in hope that classifier fusion can result in higher speaker recognition accuracy. The results for different features as well as for their combinations are directly comparable to each other and to those obtained with the baseline MFCC features.

COMBINING NOVEL ACOUSTIC FEATURES USING SVM TO DETECT SPEAKER CHANGING POINTS

Haishan Zhong, David Cho, Vladimir Pervouchine and Graham Leedham

Automatic speaker change point detection segments different speakers from continuous speech according to speaker characteristics. This is often a necessary step before applying speaker verification or identification systems. Among the features to represent a speaker in the speaker change point detection systems acoustic features are commonly used. Commonly used features are Mel Frequency Cepstral Coefficients (MFCC) and Linear Prediction Cepstral Coefficients (LPCC). However, the features are affected by speech content, environment, type of recording device, etc. So far, no features have been discovered, which values depend only on the speaker. In this paper four novel feature types proposed in recent major journals and conference papers for speaker verification problem, are applied to the problem of speaker change point detection. The features are also used to form a combination scheme via SVM classifier. The results shows that the proposed scheme improves the performance of speaker changing point detection as compared to the system that uses MFCC features. It was also found that some of the novel features of low dimensionality give comparable speaker change point detection accuracy to the high-dimensional MFCC features.

POSSIBILITY OF MENTAL HEALTH SELF-CHECKS USING DIVERGENCE PROPERTIES OF PULSE WAVES

Mayumi Oyama-Higa and Tiejun Miao

We conducted a nonlinear analysis of fingertip pulse waves and found that the Lyapunov exponent referencing the “divergence” of attractor trajectory is an effective method for determining mental health in humans. In particular, we showed that this method is very effective for the early detection of dementia and depression, as well as in the detection of mental changes in healthy persons. In contrast, current measurement methods to determine mental health are subjective in most cases and are neither objective nor simple in terms of time and cost. The development of an apparatus allowing easy measurement for many users is therefore necessary. We illustrate the possibility of mental health self-checks using pulse wave divergence based on a series of examples in previous studies. In addition, we developed software to express the fluctuation of the Lyapunov exponent using time series data from multiple measurements. If changes in mental status can be assessed by studying the fluctuation factor of the Lyapunov exponent, we will be closer to effectively evaluating and controlling mental health problems. And, we developed an easy-to-use economical device, a PC mouse with an integrated sensor for measuring the pulse waves.

IDENTIFICATION OF TIME-VARYING T-WAVE ALTERNANS FROM 20-MINUTE ECG RECORDINGS - Issues Related to TWA Magnitude Threshold and Length of ECG Time Series

Laura Burattini, Wojciech Zareba and Roberto Burattini

Aim of this study was the assessment of a T-wave alternans (TWA) identification procedure based on application of an adaptive match filter (AMF) method, recently developed by ourselves, to a 20-minute digital ECG recording (ECG20). Three-lead ECG20 tracings from 20 patients who survived an acute myocardial infarction (AMI-group) and 20 healthy subjects (H-group) were analysed. The AMI-group showed, on average, increased levels of TWA (P<0.05). Considering that noise may cause false positive TWA detection, a threshold (THRTWA) was defined for TWA magnitude (TWAM) as the mean TWAM +2SD over the H-group. TWAM exceeding this threshold identified a TWA-positive subject (TWA+) as one at increased risk of sudden cardiac death. Eight (40%) AMI-patients vs. zero H-subjects were detected as TWA+. This result meets clinical expectation. TWA manifested as a non stationary phenomenon that could even be missed in all TWA+ subjects if our AMF (as well as any other technique) was applied to a single short-term 128-beat ECG series, as usually done in previous reports. In conclusion, our AMF-based TWA identification technique, applied to 20-minute ECG recordings, yields a good compromise between reliability of time-varying TWA identification and computational efforts.

NETWORK TOMOGRAPHY-BASED TRACKING FOR INTRACELLULAR TRAFFIC ANALYSIS IN FLUORESCENCE MICROSCOPY IMAGING

Thierry Pécot, Charles Kervrann and Patrick Bouthemy

Determination of the sub-cellular localization and dynamics of any proteins is an important step towards the understanding of multi-molecular complexes in a cellular context. Green Fluorescent Protein (GFP)-tagging and time-lapse fluorescence microscopy allows to acquire multidimensional data on rapid cellular activities, and then make possible the analysis of proteins of interest. Consequently, novel techniques of image analysis are needed to quantify dynamics of biological processes observed in such image sequences. In biological trafficking analysis, the previous tracking methods do not manage when many small and poorly distinguishable objects interact. Nevertheless, an another way of tracking that usually consists in determining the full trajectories of all the objects, can be more relevant. General information about the traffic like the regions of origin and destination of the moving objects represent interesting features for analysis. In this paper, we propose to estimate the paths (regions of origin and destination) used by the objects of interest, and the proportions of moving objects for each path. This can be accomplished by exploiting the recent advances in Network Tomography (NT) commonly used in network communications. This idea is demonstrated on real image sequences for the Rab6 protein, a GTPase involved in the regulation of intracellular membrane trafficking.

A HYBRID METHOD BASED ON FUZZY INFERENCE AND NON-LINEAR OSCILLATORS FOR REAL-TIME CONTROL OF GAIT

J. C. Moreno, J. L. Pons, E. Rocon and Y. Demiris

Robust generation of motor commands for real-time control of locomotion with artificial means is crucial for human safety. This paper addresses the combination of fuzzy inference for determination of rules with a non linear oscillator system, as generators of motor commands for the control of human leg joints during walking, by means of external gait compensators, e.g. exoskeletons, functional electrical stimulation or hybrid systems. The response of the proposed method is evaluated for variations in stride frequency and step length. The testing during gait conditions is performed considering inertial sensing as feedback in a simulation study. The reference data considered is obtained in multiple experiments with healthy subjects walking with a controllable exoskeleton designed to compensate quadriceps weakness. A model of the operation of the knee joint compensation provided by the exoskeleton is obtained as reference to evaluate the method based on real data. The results demonstrate the benefits of both incorporating a) the fuzzy inference system in cyclical decision making for generation of motor commands and b) the dynamic adaptation of the timing parameters of the external compensator provided by the van der Pol oscillator.

A IMAGE PROCESSING METHOD FOR COMPARISON OF MULTIPLE RADIOGRAPHS

Chen Sheng, Li Li and Wang Pei

Portable chest radiography is the most commonly ordered radiographic test in the intensive care unit (ICU). In the ICU, a succession of portable images is usually taken over a period of time to monitor the progress of a patient’s condition. A prompt diagnosis of any changes in the conditions of these ICU patients allows clinicians to provide immediate attention and treatments that are required to prevent the conditions from worsening and which could result in a treat to the patient’s life. However, because of differences in X-ray exposure setting, patient and apparatus positioning, scattering, and grid application, for example, differences in image quality from on image to the next taken at different times can be significant. The differences in image quality make it difficult for clinicians to compare images to detect subtle changes. This paper presents an image-rendering method that reduces the variability in image appearance and enhances the diagnostic quality of these images. Use of the presented method allows clinicians to detect subtle pathological changes from one image to the next, thus improving the quality of patient management in the ICU.

AUTOMATED DETECTION OF SUPPORTING DEVICE POSITIONING IN RADIOGRAPHY

Chen Sheng, Li Li and Ying Jun

Portable X-ray radiographs are heavily used in the ICU for detecting significant or unexpected conditions requiring immediate changes in patient management. One concern for effective patient management relates to the ability to detect the proper positioning of tubes that have been inserted into the patient. These include, for example, endo-tracheal tubes (ET), feeding tubes (FT), naso-gastric tubes (NT), and other tubes. Proper tube positioning can help to ensure delivery or disposal of liquids and air/gases to and from the patient during a treatment procedure. Improper tube positioning can cause patient discomfort, render a treatment ineffective, or can even be life-threatening. However, because the poor image quality in portable AP X-ray images due to the variability in patients, apparatus positioning, and X-ray exposure, it is often difficult for clinicians to visually detect the position of tube tips. Thus, there is a need for detecting and identifying tube position and type to assist clinicians. The purpose of this paper is to present a computer-aided method for automated detection of tubes and identification of tube types. Use of this method may allow clinicians to detect the tube tips more easily and accurately, thus improving the quality of patient management in the ICU.

INFLUENCES OF DIGITAL BAND-PASS FILTERING ON THE BCG WAVEFORM

Mikko Koivuluoma, Laurentiu Barna, Alpo Värri, Teemu Koivistoinen, Tiit Kööbi and Alpo Värri

Ballistocardiography is a non-invasive technique for the assessment of cardiac function. The BCG signals usually have two main components: the heart originated component and the respiratory originated component. The frequency bands of these components overlaps, and hereby complete separation of these two components is not possible. In this study, we used several band pass filters to remove the respiratory, and tried to estimate the optimal lower cut-off frequency for this band pass filter. The optimal band pass filter should have very small effect to the heart originated BCG. We found that the optimal lower cout-off frequency is about 1.3 Hz.

BALLISTOCARDIOGRAPHIC ARTIFACT REMOVAL FROM SIMULTANEOUS EEG/FMRI RECORDING BY MEANS OF CANONICAL CORRELATION ANALYSIS

S. Assecondi, P. Van Hese, H. Hallez, Y. D'Asseler, I. Lemahieu, A. M. Bianchi and P. Boon

The electroencephalogram (EEG) is a standard technique to record and study the brain activity with a high temporal resolution. Blood oxygenation level dependent functional magnetic resonance imaging (BOLD fMRI) is a non-invasive imaging method that allows the localization of activated brain regions with a high spatial resolution. The co-recording of these two complementary modalities can give new insights into how the brain functions. However, the interaction between the strong electromagnetic field (3T) of the MR scanner and the currents recorded by the electrodes placed on the scalp generates artifacts that obscure the EEG and diminish its readability. In this work we used canonical correlation analysis (CCA) in order to remove the ballistocardiographic artifact (BCGa). CCA is applied to two consecutive windows in order to take into account both spatial and temporal information. We showed that users can easily remove the artifact through a graphical user interface by adjusting the number of components to be removed according to visual inspection of the signal and its power spectrum.

ON-CHIP FLUORESCENCE LIFETIME EXTRACTION USING SYNCHRONOUS GATING SCHEME - Theoretical Error Analysis and Practical Implementation

Day-Uei Li, Bruce Rae, David Renshaw, Robert Henderson and Eleanor Bonnist

A synchronous gating technique was proposed for fluorescent photon collecting. The two-gate rapid lifetime determination (RLD) technique was applied to implement on-chip fluorescence lifetime extraction. Compared with all available iterative least square method (LSM) or maximum likelihood estimation (MLE) based general purpose FLIM analysis software, our chips offer direct calculation of lifetime based on the photon counts stored on the on-chip memory and deliver faster analysis for higher possibility of real-time applications, such as clinical diagnosis. The cost of our chips is much less than available solutions, since we don’t need any data fitting software and photon counting card. Theoretical error analysis of the two- and multi-gate RLDs were derived for comparison. And we applied a two-gate RLD scheme based on the analysis suggested. The performance of the chips were tested on a single-exponential Rhodamine B obtained from our SPAD detector using 468nm laser diode as light sources with optimized gate width. Moreover, a multi-exponential pipelined two-gate RLD (PL-RLD-2) FLIM was also proposed and tested on a four-exponential decays DNA sample containing a single adenine analogue 2-aminopurine.

MOUSE CONTROL THROUGH ELECTROMYOGRAPHY - Using Biosignals Towards New User Interface Paradigms

Vasco Vinhas and Antonio Gomes

Recent technologic breakthroughs have enabled the usage of minimal invasive biometric hardware devices that no longer interfere with the audience immersion feeling. The usage of EMG to extend traditional mouse-oriented user interfaces is a proof-of-concept prototype integrated in a wider horizon project. A subset of the main project's architecture was reused, specially the communication middleware, as a stable development platform. An originally intended EEG hardware was adapted to perform EMG and therefore detect muscular activity. It was chosen, as a practical proof-of-concept, that it was desired to detect winking as a triggering device to perform a given traditional user interface action. The described application achieved extremely positive records with hit rates of around 90%. The volume of false positives and undetected desired actions are considered negligible due to both system development stage and application contextualization – non critical systems. The success and acceptance levels of the project are really encouraging not only to the enhancement of the proposed application but also to the global system continuous development.

DO MOBILE PHONES AFFECT SLEEP? - Investigating Effects of Mobile Phone Exposure on Human Sleep EEG

Andrew Wood, Sarah Loughran, Rodney Croft, Con Stough and Bruce Thompson

This paper will summarize the results of a human volunteer study on the effects on sleep parameters of exposure to RF emissions from a mobile phone handset for 30min prior to going to sleep. A cohort of 55 volunteers were tested over 4 nights in a double-blind design. The significant outcomes were: Rapid Eye Movement (REM) sleep latency reduced by 16%; EEG alpha power enhanced by 8% during 1st non-REM period. These results are compared for overall internal consistency and with studies from other laboratories. Part of the program of the Australian Centre for Radiofrequency Bioeffects Research extending these studies is described.

A NOVEL TEMPLATE HUMAN FACE MODEL WITH TEXTURING

Ken Yano and Koichi Harada

We present a method to fit a template face model to 3D scan face. We first normalize the size and align the orientation then fit the model roughly by scattered interpolation method. Secondly we run the optimization method based on Allen's work. We are able to generate face models which have "poin-to-point" correspondence among them. We also suggest a way to transfer any facial texture image over this fitted model.

ANT COLONY INSPIRED METAHEURISTICS IN BIOLOGICAL SIGNAL PROCESSING - Hybrid Ant Colony and Evolutionary Approach

Miroslav Bursa, Michal Huptych and Lenka Lhotska

Nature inspired metaheuristics have interesting stochastic properties which make them suitable for use in data mining, data clustering and other application areas, because they often produce more robust solutions. This paper presents an application of clustering method inspired by the behavior of real ants in the nature in biomedical signal processing. The main aim of our study was to design and develop a combination of feature extraction and classification methods for automatic recognition of significant structure in biological signal recordings. The method would speed up and increase objectivity of identification of important classes and may be used for online classification and can be also used as a hint in the expert classification. We have obtained significant results in electrocardiogram and electroencephalogram recordings, which justify the use of such methods method.

ON THE FUTILITY OF INTERPRETING OVER-REPRESENTATION OF MOTIFS IN GENOMIC SEQUENCES AS FUNCTIONAL SIGNALS

Nikola Stojanovic

Locating signals for the initiation of gene expression in DNA sequences is an important unsolved problem in genetics. Over more than two decades researchers have applied a large variety of sophisticated computational techniques in order to address it, but only with moderate success. In this paper we investigate the reasons for the relatively poor performance of the current models, and outline some possible directions for future work in this field.

INVESTIGATION OF ICA ALGORITHMS FOR FEATURE EXTRACTION OF EEG SIGNALS IN DISCRIMINATION OF ALZHEIMER DISEASE

Jordi Solé-Casals, François Vialatte, Zhe Chen and Andrzej Cichocki

In this paper we present a quantitative comparisons of different independent component analysis (ICA) algorithms in order to investigate their potential use in preprocessing (such as noise reduction and feature extraction) the electroencephalogram (EEG) data for early detection of Alzhemier disease (AD) or discrimination between AD (or mild cognitive impairment, MCI) and age-match control subjects.

USING WAVELET TRANSFORM FOR FEATURE EXTRACTION FROM EEG SIGNAL

Lenka Lhotska, Vaclav Gerla, Jiri Bukartyk, Vladimir Krajca and Svojmil Petranek

Manual evaluation of long-term EEG recordings is very tedious, time consuming, and subjective process. The aims of automated processing are on one side to ease the work of medical doctors and on the other side to make the evaluation more objective. This paper addresses the problem of computer-assisted sleep staging. It describes ongoing research in this area. The proposed solution comprises several consecutive steps, namely EEG signal pre-processing, feature extraction, feature normalization, and application of decision trees for classification. The work is focused on the feature extraction step that is regarded as the most important one in the classification process.

DYNAMICAL PROPERTY OF PERIODIC OSCILLATIONS OBSERVED IN A COUPLED NEURAL OSCILLATOR NETWORK FOR IMAGE SEGMENTATION

Tetsuya Yoshinaga and Keníchi Fujimoto

We consider image segmentation using the LEGION (Locally-Excitatory Globally-Inhibitory Oscillator Network), and investigate dynamical properties of a modified LEGION, described by noise-free or deterministic continuous ordinary differential equations. We clarify a phenomenon of image segmentation corresponds to the appearance of a synchronized periodic solution, and the ability of segmentation depends on its symmetric properties. We study bifurcations of periodic solutions by using a computational method based on the qualitative dynamical system theory.

ARAFAC CLASSIFICATION OF LAMB CARCASS SOFT TISSUES IN COMPUTER TOMOGRAPHY (CT) IMAGE STACKS

Jørgen Kongsro

Computer Tomography is shown to be an efficient and cost-effective tool for classification and segmentation of soft tissues in animal carcasses. By using 15 fixed anatomical sites based on vertebra columns, 120 lamb carcasses were CT scanned in Norway during autumn of 2005. Frequency distributions of CT values (HU [-200,200]) of soft tissues from each image were obtained. This yielded a 3-way data set (120 samples * 400 CT values * 15 anatomical sites). The classification of the soft tissues was done by multi way Parallel Factor Analysis (PARAFAC), which resulted in 3 components or soft tissues classified from the images; fat, marbled and lean muscle tissue.

BIOPHYSICAL MODEL OF A MUSCLE FATIGUE PROCESS INVOLVING Ca2+ RELEASE DYNAMICS UPON THE HIGH FREQUENCY ELECTRICAL STIMULATION

Piotr Kaczmarek

The aim of this study is to create a model which enables to explain the muscle fibre contraction due to various stimulation programs. The model accounts for $Ca^{2+}$ release dynamics both as a result of an action potential and of a stimulus shape, duration and frequency. It has been assumed that the stimulus can directly activate the voltage-dependent receptors (dihydropiridine receptors) responsible for a $Ca^{2+}$ release. The stimulation programs consisted of standard stimulation trains made of low and middle frequency square pulses. High frequency modulating harmonic signals have been tested to investigate the fibre fatigue effect. It has been observed that fatigue effect factors depend on the selected stimulation program. The results reveal that the fatigue effect could be minimized by changing the shape and frequency of the stimulation waveform. Such the model could be useful for a preliminary selection and optimization of the stimulus shape and the stimulation trains, thus reducing the number of in vivo experiments.

AUTOMATIC DETECTION OF IN VITRO CAPILLARY TUBE NETWORK IN A MATRIGEL ANALYSIS

Eric Brassart, Cyril Drocourt, Jacques Rochette, Michel Slama and Carole Amant

Angiogenesis, the formation of new capillary blood vessels from pre-existing vessel, has become an important area of scientific research. Numerous in vivo and in vitro angiogenesis assays have been developed in order to test molecules designed to cure deregulated angiogenesis. But unlike most animal models, most in vitro angiogenesis models are not yet automatically analysed and conclusion and data quantification depend on the observer’s analysis. In our study, we will develop a new automatic in vitro matrigel angiogenesis analysis allowing tube length and the number of tubes per cell islets as well as cell islet and tubule mapping to be determined, percentage of vascularisation area, the determination of ratio of tubule length per number of cells in cell islet and, ratio length/width per tubule determination. This new method will also take image noise into account. Our method uses classical imaging quantification. For the first image processing we used image segmentation (Sobel type edge detection) and artefact erasing (morphologic operator). Subsequent image processing used Snakes: Active contour models in order to precisely detect cells or cell islets. We suggest that this new automated image analysis method for quantification of in vitro angiogenesis will give the researcher vascular specific quantified data that will help in the comparison of samples.

A SUPERVISED LEARNING APPROACH BASED ON THE CONTINUOUS WAVELET TRANSFORM FOR R SPIKE DETECTION IN ECG

G. de Lannoy, A. de Decker and M. Verleysen

One of the most important tasks in automatic annotation of the ECG is the detection of the R spike. The wavelet transform is a widely used tool for R spike detection. The time-frequency decomposition is indeed a powerful tool to analyze non-stationary signals. Still, current methods use consecutive wavelet scales in an a priori restricted range and may therefore lack adaptivity. This paper introduces a supervised learning algorithm which learns the optimal scales for each dataset using the annotations provided by physicians on a small training set. For each record, this method allows a specific set of non consecutive scales to be selected, based on the record characteristics. The selected scales are then used on the original long-term ECG signal recording and a hard thresholding rule is applied on the derivative of the wavelet coefficients to label the R spikes. This algorithm has been tested on the MIT-BIH arrhythmia database and obtains an average sensitivity rate of 99.7% and average positive predictivity rate of 99.7%.

ROBUST CENTROID-BASED CLUSTERING USING DERIVATIVES OF PEARSON CORRELATION

Marc Strickert, Nese Sreenivasulu, Thomas Villmann and Barbara Hammer

Modern high-throughput facilities provide the basis of -omics research by delivering extensive biomedical data sets. Mass spectra, multi-channel chromatograms, or cDNA arrays are such data sources of interest for which accurate analysis is desired. Centroid-based clustering provides helpful data abstraction by representing sets of similar data vectors by characteristic prototypes, placed in high-density regions of the data space. This way, specific modes can be detected, for example, in gene expression profiles or in lists containing protein and metabolite abundances. Despite their widespread use, k-means and self-organizing maps (SOM) often only produce suboptimum results in centroid computation: the final clusters are strongly dependent on the initialization and they do not quantize data as accurately as possible, particularly, if other than the Euclidean distance is chosen for data comparison. Neural gas (NG) is a mathematically rigorous clustering method that optimizes the centroid positions by minimizing their quantization errors. Originally formulated for Euclidean distance, in this work NG is mathematically generalized to give accurate and robust results for the Pearson correlation similarity measure. The benefits of the new NG for correlation (NG-C) are demonstrated for sets of gene expression data and mass spectra.

A PROBABILISTIC TRACKING APPROACH TO ROOT MEASUREMENT IN IMAGES - Particle Filter Tracking is used to Measure Roots, via a Probabilistic Graph

Andrew French, Malcolm Bennett, Caroline Howells, Dhaval Patel and Tony Pridmore

This paper introduces a new methodology to aid the tracing and measurement of lines in digital images. The techniques in this paper have specifically been applied to the labour intensive process of measuring roots in digital images. Current manual methods can be slow and error prone, and so we propose a semi-automatic way to trace the root image and measure the corresponding length in the image plane. This is achieved using a particle filter tracker, normally applied to object tracking though time, to trace along a root in an image. The samples the particle filter generates are used to build a probabilistic graph across the root location in the image, and this is traversed to produce a final estimate of length. The software is compared to real-world and artificial length data. Extensions of the algorithm are noted, including the automatic detection of the end of the root, and the detection of multiple growth modes using a mixed state particle filter.

FEASABILITY OF YEAST AND BACTERIA IDENTIFICATION USING UV-VIS-SWNIR DIFUSIVE REFLECTANCE SPECTROSCOPY

J. S. Silva, R. C. Martins, A. A. Vicente and J. A. Teixeira

UV-VIS spectroscopy is a powerfull qualitative and quantitative technique used in analytical chemistry, which gives information about electronic transitions of electrons in molecular orbitals. As in UV-VIS spectra there is no direct information on characteristic organic groups, vibrational spectroscopy (e.g. infrared) has been preferred for biological applications. In this research, we try to use state-of-the-art fiber optics probes to obtain UV-VIS-SWNIR diffusive reflectance measurements of yeasts and bacteria colonies on plate count agar in the region of 200-1200nm; in order to discriminate the following microorganisms: i) yeasts: Saccharomyces cerevisiae, Saccharomyces bayanus, Candida albicans, Yarrowia lipolytica; and ii) bacteria: Micrococcus luteus, Pseudomonas fluorescens, Escherichia coli, Bacillus cereus. Spectroscopy results show that UV-VIS-SWNIR has great potential for identifying microorganisms on plate count agar. Scattering artifacts of both colonies and plate count agar can be significantly removed using a robust mean scattering algorithm, allowing also better discriminations between the scores obtained by singular value decomposition. Hierarchical clustering analysis of UV-VIS and VIS-SWNIR decomposed spectral scores lead to the conclusion that the use of VIS-SWNIR light source produces higher discrimination ratios for all the studied microorganisms, presenting great potential for developing biotechnology applications.

ENHANCED ANALYSIS OF UTERINE ACTIVTY USING SURFACE ELECTROMYOGRAPHY

A. Herzog, L. Reicke, M. Kröger, C. Sohn and H. Maul

This contribution presents a new approach for the enhanced analysis of uterine surface electromyography (EMG). First, a pulse detection separates the pulses, which contain the essential information about the uterine contractibility, from the flat line sections during relaxation. The functionality of this semi-automatic algorithm is controlled by two comprehensible parameters. Subsequently, the mean frequency, which serves as a criterion for imminent delivery, is evaluated from the extracted pulses. Although the pulse detection reduces the deviation of the mean frequency significantly, the results are still sensitive to parameter variations in the pulse detection. A stochastic analysis based on the Karhunen-Loève transform (KLT) derives generalised patterns, the eigenforms, from the pulse ensemble. The mean frequency of the first eigenform is less sensitive to parameter variations. Additionally, the correlation between the eigenforms of the left and right surface electrode can serve as a criterion for the measurement's quality.

BIOMIMETIC FLOW IMAGING WITH AN ARTIFICIAL FISH LATERAL LINE

Nam Nguyen, Douglas Jones, Saunvit Pandya, Yingchen Yang, Nannan Chen, Craig Tucker and Chang Liu

Recent studies have discovered that almost all fish possess a flow-sensing system along their body, called the lateral line, that allows them to perform various behaviours such as schooling, preying, and obstacle or predator avoidance. Inspired from this, our group has built artificial lateral lines from newly-developed flow sensors using Micro-Electro-Mechanical Systems (MEMS) technology. To make our lateral line a functional sensory system, we develop an adaptive beamforming algorithm (applying Capon’s method) that provides our lateral line with the capability of imaging the locations of oscillating dipoles in a 3D underwater environment. To help our sensor arrays adapt to the environment for better performance, we introduce a self-calibration algorithm that significantly improves the image accuracy. Finally, we derive the Cramer-Rao Lower Bound (CRLB) that represents the fundamental perfomance limit of our system and provides guidance in optimizing artificial lateral-line systems.

MULTIPLE SCALE NEURAL ARCHITECTURE FOR RECOGNISING COLOURED AND TEXTURED SCENES

Francisco Javier Díaz-Pernas, Míriam Antón-Rodríguez, Víctor Iván Serna-González José Fernando Díez-Higuera and Mario Martínez-Zarzuela

A dynamic multiple scale neural model for recognise colour images of textured scenes is proposed. This model combines colour and textural information to recognise coloured textures through the operation of two main components: segmentation component formed by the Colour Opponent System (COS) and the Chromatic Segmentation System (CSS); and recognition component formed by pattern generation stages and Fuzzy ARTMAP neural network. Firstly, the COS module transforms the RGB chromatic input signals into a bio-inspired codification system (L, M, S and luminance signals), and then it generates the opponent channels (black-white, L-M and S-(L+M)). The CSS module incorporates contour extraction, double opponency mechanisms and diffusion processes in order to generate coherent enhancing regions in colour image segmentation. These colour region enhancements along with the local textural features of the scene constitute the recognition pattern to be sent into the Fuzzy ARTMAP network. The structure of the CSS architecture is based on BCS/FCS systems, thus, maintaining their essential qualities such as illusory contours extraction, perceptual grouping and discounting the illuminant. But base models have been extended to allow colour stimuli processing in order to obtain general purpose architecture for image segmentation with later applications on computer vision and object recognition. Some comparative testing with other models is included here in order to prove the recognition capabilities of this neural architecture.

AUTOMATIC COUINAUD LIVER AND VEINS SEGMENTATION FROM CT IMAGES

Dário A. B. Oliveira, Raul Q. Feitosa and Mauro M. Correia

This paper presents an algorithm to segment the liver structures on computed tomography (CT) images according to the Couinaud orientation. Our method firstly separates the liver from the rest of the image. Then it segments the vessels inside the liver area using a region growing technique combined with hysteresis thresholding. It separates the vessels in segments without any bifurcation, and using heuristics based on anatomy, it classifies all vessel segments as hepatic or portal vein. Finally, the method estimates the planes that best fit each of the three branches of the segmented hepatic veins and the plane that best fits the portal vein. These planes define the subdivision of the liver in the Couinaud segments. An experimental evaluation based on real CT images demonstrated that the outcome of the proposed method is generally consistent with a visual segmentation.

MULTI-CHANNEL BIOSIGNAL ANALYSIS FOR AUTOMATIC EMOTION RECOGNITION

Jonghwa Kim and Elisabeth André

This paper investigates the potential of physiological signals as a reliable channel for automatic recognition of user's emotial state. For the emotion recognition, little attention has been paid so far to physiological signals compared to audio-visual emotion channels such as facial expression or speech. All essential stages of automatic recognition system using biosignals are discussed, from recording physiological dataset up to feature-based multiclass classification. Four-channel biosensors are used to measure electromyogram, electrocardiogram, skin conductivity and respiration changes. A wide range of physiological features from various analysis domains, including time/frequency, entropy, geometric analysis, subband spectra, multiscale entropy, etc., is proposed in order to search the best emotion-relevant features and to correlate them with emotional states. The best features extracted are specified in detail and their effectiveness is proven by emotion recognition results.

BIOSIGNALS ANALYSIS AND ITS APPLICATION IN A PERFORMANCE SETTING - Towards the Development of an Emotional-Imaging Generator

Mitchel Benovoy, Jeremy R. Cooperstock and Jordan Deitcher

The study of automatic emotional awareness of human subjects by computerized systems is a promising avenue of research in human-computer interaction with profound implications in media arts and theatrical performance. A novel emotion elicitation paradigm focused on self-generated stimuli is applied here for a heightened degree of confidence in collected physiological data. This is coupled with biosignal acquisition (electrocardiogram, blood volume pulse, galvanic skin response, respiration, phalange temperature) for determination of emotional state using signal processing and pattern recognition techniques involving sequential feature selection, Fisher dimensionality reduction and linear discriminant analysis. Discrete emotions significant to Russell’s arousal/valence circumplex are classified with an average recognition rate of 90%.

BIO-INSPIRED IMAGE PROCESSING FOR VISION AIDS

C. Morillas, F. Pelayo, J. P. Cobos, A. Prieto and S. Romero

We present in this paper a system conceived to perform a bioinspired image processing and different output encoding schemes, oriented to the development of visual aids for the blind or for