Search

Purpose: The CIGMA study investigated a novel human polyclonal antibody preparation (trimodulin) containing ~ 23% immunoglobulin (Ig) M, ~ 21% IgA, and ~ 56% IgG as add-on therapy for patients with severe community-acquired pneumonia (sCAP). Methods: In this double-blind, phase II study (NCT01420744), 160 patients with sCAP requiring invasive mechanical ventilation were randomized (1:1) to trimodulin (42 mg IgM/kg/day) or placebo for five consecutive days. Primary endpoint was ventilator-free days (VFDs). Secondary endpoints included 28-day all-cause and pneumonia-related mortality. Safety and tolerability were monitored. Exploratory post hoc analyses were performed in subsets stratified by baseline C-reactive protein (CRP; ≥ 70 mg/L) and/or IgM (≤ 0.8 g/L). Results: Overall, there was no statistically significant difference in VFDs between trimodulin (mean 11.0, median 11 [n = 81]) and placebo (mean 9.6; median 8 [n = 79]; p = 0.173). Twenty-eight-day all-cause mortality was 22.2% vs. 27.8%, respectively (p = 0.465). Time to discharge from intensive care unit and mean duration of hospitalization were comparable between groups. Adverse-event incidences were comparable. Post hoc subset analyses, which included the majority of patients (58–78%), showed significant reductions in all-cause mortality (trimodulin vs. placebo) in patients with high CRP, low IgM, and high CRP/low IgM at baseline. Conclusions: No significant differences were found in VFDs and mortality between trimodulin and placebo groups. Post hoc analyses supported improved outcome regarding mortality with trimodulin in subsets of patients with elevated CRP, reduced IgM, or both. These findings warrant further investigation. Trial registration: NCT01420744., 0, SCOPUS: ar.j, info:eu-repo/semantics/published

This paper proposes the application of nonlinear measures based on nonlinear dynamics for emotional speech characterization. Measures such as mutual information, dimension correlation, entropy correlation, Shannon entropy, Lempel-Ziv complexity and Hurst exponent are extracted from the samples of a database of emotional speech. Then, statistics such as mean, standard deviation, skewness and kurtosis are applied on the extracted measures. Experiments were conducted on the Berlin emotional speech database for a three-class problem (neutral, fear and anger emotional states). Feature selection is accomplished to select a reduced number of features. In order to evaluate the discrimination ability of the selected features a neural network classifier is used. A global success rate of 93.78% is obtained.

In this work a novel system for sea surface images pre-processing and processing has been developed in order to detect bubbles on the sea surface. This application is fundamental to verify radiometer satellite systems which are used to the study of the floor humidity and the sea salinity. 160 images of 8 kinds of salinity have been processed, 20 per class. Two main steps have been implemented; the first step is the image pre-processing and enhancing, in order to improve the bubbles detection. The second step is the segmentation and the bubbles detection. A combination system has been used in order to improve the final result, getting a recognition rate of 95.43%.

Transcranial magnetic stimulation induces weak currents within the cranium to activate neuronal firing and its response is recorded using electroencephalography in order to study the brain directly However, different artifacts contaminate the results The goal of this study is to process these artifacts and reduce them digitally Electromagnetic, blink and auditory artifacts are considered, and Signal-Space Projection, Independent Component Analysis and Wiener Filtering methods are used to reduce them These last two produce a successful solution for electromagnetic artifacts Regarding the other artifacts, processed with Signal-Space Projection, the method reduces the artifact but modifies the signal as well Nonetheless, they are modified in an exactly known way and the vector used for the projection is conserved to be taken into account when analyzing the resulting signals A system which combines the proposed methods would improve the quality of the information presented to physicians.

This present work presents different feature reduction methods, applied to Deoxyribonucleic Acid (DNA) marker, and in order to identify a success of 100% based on Discriminate Common Vectors (DCV), Principal Component Analysis (PCA), and Independent Component Analysis (ICA) using as classifiers Support Vector Machines (SVM) and Artificial Neural Networks. In particular, the biochemical parameterization has 89 Random Amplified polymorphic DNA (RADPS) markers of Pejibaye palm landraces, and it has been reduced from 89 to a 3 characteristics, for the best method using ICA. The interest of this application is due to feature reduction and therefore, the reduction of computational load time versus the use of all features. This method allows having a faster supervised classification system for the process of the plant certification with origin denomination. Therefore, this system can be transferred to voice applications in order to reduce load time, keeping or improving the success rates.

This present work presents a parameterization system based on angles from signature edge (2D-shape) for off-line signature identification. We have used three different classifiers, the Nearest Neighbor classifier (K-NN), Neural Networks (NN) and Hidden Markov Models (HMM). Our off-line database has 800 writers with 24 samples per each writer; in total, 19200 images have been used in our experiments. We have got a success rate of 84.64%, applying as classifier Hidden Markov Model, and only used the information from this edge detection method.

The final publication is available at Springer via http://dx.doi.org/10.1007/978-3-642-04391-8_31, Proceedings of Joint COST 2101 and 2102 International Conference, BioID_MultiComm 2009, Madrid, Spain., In this paper we present an autonomous biometric device developed in the framework of a national project. This system is able to capture speech, hand-geometry, online signature and face, and can open a door when the user is positively verified. Nevertheless the main purpose is to acquire a database without supervision (normal databases are collected in the presence of a supervisor that tells you what to do in front of the device, which is an unrealistic situation). This system will permit us to explain the main differences between what we call "real conditions" as opposed to "laboratory conditions"., This work has been supported by FEDER and MEC, TEC2006-13141-C03/TCM, and COST-2102.

A neural network with characteristic parameters to recognize abnormalities in ultrasound images acquired from echographic tissue-mimicking materials is proposed. The neural network has been implemented in MATLAB and it can be used in real time to assist the clinical diagnoses in the early phases. The parameters are extracted from a database of B-mode ultrasound images. After training and testing the network, using a statistically significative set of experimental data and a non-commercial phantom, results show that the proposal can be successfully applied to efficiently deal with this problem.

In this paper, we describe two different hand image databases. One has been acquired in laboratory condition with a document scanner, and the other one in operational conditions using a webcam and an infrared filter. The experimental part describes some verification experiments and extracts relevant conclusions about image acquisition and biometric classification.

This paper proposes to use the Fisher kernel for planar shape recognition. A synthetic experiment with artificial shapes has been built. The difference among shapes is the number of vertexes, links between vertexes, size and rotation. The 2D-shapes are parameterized with sweeping angles in order to obtain scale and rotation invariance. A Hidden Markov Model is used to obtain the Fisher score which feeds the Support Vector Machine based classifier. Noise has been added to the shapes in order to check the robustness of the system against noise. Hit ratio score over 99%, has been obtained, which shows the ability of the Fisher kernel tool for planar shape recognition.

In this paper we propose the use of nonlinear speech features to improve the voice quality measurement. We have tested a couple of features from the Dynamical System Theory, namely: the Correlation Dimension and the largest Lyapunov Exponent. In particular, we have studied the optimal size of time window for this type of analysis in the field of the characterization of the voice quality. Two systems of automatic detection of laryngeal pathologies, one of them including these features, have been implemented with the purpose of validating the usefulness of the suggested nonlinear features. We obtain slight improvements with respect to a classical system.

In this paper we present the development of BrailLector, a system able to speak from Braille writing. By means of dynamic thresholding, adaptive Braille grid, recovery dots techniques and TTS software (Text-To-Speech), BrailLector translates Braille scanned images into normal text, and not only that, it speaks the translated text. BrailLector is a robust application with innovative thersholding and Braille grid creation algorithms which detects and read Braille characters with 99.9% of correct symbols and an error variance below 0.012. The conversion time is only 26 secs for double-sided documents by MATLAB programming language.