Garde A.,Polytechnic University of Catalonia |
Garde A.,Institute for Bioengineering of Catalonia IBEC |
Garde A.,CIBER ISCIII |
Voss A.,Jena University of Applied Sciences |
And 7 more authors.
Computers in Biology and Medicine | Year: 2013
Classification algorithms with unbalanced datasets tend to produce high predictive accuracy over the majority class, but poor predictive accuracy over the minority class. This problem is very common in biomedical data mining. This paper introduces a Support Vector Machine (SVM)-based optimized feature selection method, to select the most relevant features and maintain an accurate and well-balanced sensitivity-specificity result between unbalanced groups. A new metric called the balance index (B) is defined to implement this optimization. The balance index measures the difference between the misclassified data within each class. The proposed optimized feature selection is applied to the classification of patients' weaning trials from mechanical ventilation: patients with successful trials who were able to maintain spontaneous breathing after 48h and patients who failed to maintain spontaneous breathing and were reconnected to mechanical ventilation after 30min. Patients are characterized through cardiac and respiratory signals, applying joint symbolic dynamic (JSD) analysis to cardiac interbeat and breath durations. First, the most suitable parameters (C+, C-, σ) are selected to define the appropriate SVM. Then, the feature selection process is carried out with this SVM, to maintain B lower than 40%. The best result is obtained using 6 features with an accuracy of 80%, a B of 18.64%, a sensitivity of 74.36% and a specificity of 82.42%. © 2013 Elsevier Ltd.
Kim S.,Biomedical Engineering Research Center |
Choi J.,Biomedical Engineering Research Center
ICCAS 2015 - 2015 15th International Conference on Control, Automation and Systems, Proceedings | Year: 2015
For the design of an electrothermal bipolar vessel sealing device, the consideration of thermal damage to the vessel is important. To design a radio frequency (RF) electrode that imparts minimal thermal damage to the tissue, we proposed four different types of RF electrode arrangements. We also simulated the temperature distribution in each type of arrangement using an analysis program. We assumed the material of the RF electrodes to be heat-resistant stainless steel alloy 321 and defined the heat transfer parameters of the blood vessel. A volume-type heat source was set to constantly generate 5 W of heat in a period of 10 s. We did not consider the conditions of the blood vessel, such as humidity at the contact surface, effect of the surrounding tissues, and blood flow, for the purpose of simplifying the model under simulation. The maximum temperature of the blood vessel in each case was similar but the temperature distribution was different depending on the shapes of the RF electrodes. We identified the best operation time for maintaining the temperature under 50 °C, the threshold for minimal thermal damage to the tissue . Using the results of this study, we will be able to obtain the appropriate operation time and a more suitable design for the RF electrode. © 2015 Institute of Control, Robotics and Systems - ICROS.
Rojas-Martinez M.,CIBER ISCIII |
Rojas-Martinez M.,Biomedical Engineering Research Center |
Rojas-Martinez M.,Polytechnic University of Catalonia |
Mananas M.A.,CIBER ISCIII |
And 5 more authors.
Journal of NeuroEngineering and Rehabilitation | Year: 2012
Background: sEMG signal has been widely used in different applications in kinesiology and rehabilitation as well as in the control of human-machine interfaces. In general, the signals are recorded with bipolar electrodes located in different muscles. However, such configuration may disregard some aspects of the spatial distribution of the potentials like location of innervation zones and the manifestation of inhomogineties in the control of the muscular fibers. On the other hand, the spatial distribution of motor unit action potentials has recently been assessed with activation maps obtained from High Density EMG signals (HD-EMG), these lasts recorded with arrays of closely spaced electrodes. The main objective of this work is to analyze patterns in the activation maps, associating them with four movement directions at the elbow joint and with different strengths of those tasks. Although the activation pattern can be assessed with bipolar electrodes, HD-EMG maps could enable the extraction of features that depend on the spatial distribution of the potentials and on the load-sharing between muscles, in order to have a better differentiation between tasks and effort levels. Methods. An experimental protocol consisting of isometric contractions at three levels of effort during flexion, extension, supination and pronation at the elbow joint was designed and HD-EMG signals were recorded with 2D electrode arrays on different upper-limb muscles. Techniques for the identification and interpolation of artifacts are explained, as well as a method for the segmentation of the activation areas. In addition, variables related to the intensity and spatial distribution of the maps were obtained, as well as variables associated to signal power of traditional single bipolar recordings. Finally, statistical tests were applied in order to assess differences between information extracted from single bipolar signals or from HD-EMG maps and to analyze differences due to type of task and effort level. Results: Significant differences were observed between EMG signal power obtained from single bipolar configuration and HD-EMG and better results regarding the identification of tasks and effort levels were obtained with the latter. Additionally, average maps for a population of 12 subjects were obtained and differences in the co-activation pattern of muscles were found not only from variables related to the intensity of the maps but also to their spatial distribution. Conclusions: Intensity and spatial distribution of HD-EMG maps could be useful in applications where the identification of movement intention and its strength is needed, for example in robotic-aided therapies or for devices like powered- prostheses or orthoses. Finally, additional data transformations or other features are necessary in order to improve the performance of tasks identification. © 2012 Rojas-Martínez et al.; licensee BioMed Central Ltd.
Massanet-Vila R.,University of Barcelona |
Massanet-Vila R.,Biomedical Engineering Research Center |
Massanet-Vila R.,CIBER ISCIII |
Caminal P.,University of Barcelona |
And 5 more authors.
2010 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10 | Year: 2010
Previous studies have suggested that some graph properties of protein interaction networks might be related with gene morbidity. In particular, it has been suggested that when a polymorphism affects a gene, it is more likely to produce a disease if the node degree in the interaction network is higher than for other genes. However, these results do not take into account the possible bias introduced by the variance in the amount of information available for different genes. This work models the relationship between the morbidity associated with a gene and the degrees of the nodes in the protein interaction network controlling the amount of information available in the literature. A set of 7461 genes and 3665 disease identifiers reported in the Online Mendelian Inheritance in Man (OMIM) was mined jointly with 9630 nodes and 38756 interactions of the Human Proteome Resource Database (HPRD). The information available from a gene was measured through PubMed mining. Results suggest that the correlation between the degree of a node in the protein interaction network and its morbidity is largely contributed by the information available from the gene. Even though the results suggest a positive correlation between the degree of a node and its morbidity while controlling the information factor, we believe this correlation has to be taken with caution for it can be affected by other factors not taken into account in this study. © 2010 IEEE.
Hong S.,Korea University |
Hong S.,Biomedical Engineering Research Center |
Lee J.Y.,Korea University |
Hwang C.,Biomedical Engineering Research Center |
And 2 more authors.
Tissue Engineering - Part A | Year: 2016
The aggregation of multiple cells, such as mesenchymal condensation, is an important biological process in skeletal muscle development, osteogenesis, and adipogenesis. Due to limited in vivo study model systems, a simple and effective in vitro three-dimensional (3D) aggregation system is required to study the mechanisms of multicellular aggregation and its applications. We first generated controlled mesenchymal stem cell (MSC) aggregates using a bioprinting technique to monitor their aggregation and sprouting. We induced the angiogenic potential of the MSCs through chemical inhibition of the Rho/Rho-associated protein kinase (ROCK) pathway, which led to hairy sprouting in the aggregates. The angiogenic potential of this 3D construct was then tested by subcutaneously implanting the Matrigel with 3D MSC aggregates in a rat. Treatment of 3D MSCs with the ROCK inhibitor, Y27632, increased their angiogenic activity in vivo. The gene expressions and histological staining indicated that angiogenesis and neovascularization were mainly regulated by the paracrine factors secreted from human 3D MSC constructs. Our results demonstrate the enhancement of the angiogenic potential of the MSC constructs through the secretion of vascular endothelial growth factor (VEGF) and epidermal growth factor (EGF) by the inhibition of the Rho/ROCK pathway. © Copyright 2016, Mary Ann Liebert, Inc. 2016.