Search

The use of the electrical activity from the muscles may provide a natural way to control exoskeletons or other robotic devices seamlessly. The major challenges to achieve this goal are human motor redundancy and surface electromyography (sEMG) variability. The goal of this work is to find a feature extraction and classification procedures to estimate accurately elbow angular trajectory by means of a NARX Neural Network. The processing time-step should be small enough to make it feasible its further use for online control of an exoskeleton. In order to do so we analysed the Biceps and Triceps Brachii data from an elbow flexo-extension Coincident Timing task performed in the horizontal plane. The sEMG data was pre-processed and its energy was divided in five frequency intervals that were fed to a Nonlinear Auto Regressive with Exogenous inputs (NARX) Neural Network. The estimated angular trajectory was compared with the measured one showing a high correlation between them and a RMSE error maximum of 7 degrees. The procedure presented here shows a reasonably good estimation that, after training, allows real-time implementation. In addition, the results are encouraging to include more complex tasks including the shoulder joint.

Simultaneous Localization and Mapping (SLAM) is an area of enormous growth in the robotics field. With the current evolution of microprocessors technology and the development of high-speed and precise SLAM solutions, running real-time SLAM algorithms on embedded platforms starts to become a feasible task. For this challenge to be achieved, evaluations on state-of-the-art SLAM methods running on the most up-to-date microcomputers at the moment must be done. This work presents results of ORB-SLAM2 running on both a Raspberry Pi 3B+ board and a NVIDIA Jetson Nano and suggests optimizations to be performed on future works.

Several computational activities, such as image segmentation and classification, use texture information. This information, also called descriptors, is typically calculated through the Gray Level Occurrence Matrix (GLCM), which has a quadratic cost. In this work, we discuss an alternative to this method using Sum and Difference Histograms (SDH), which has linear cost. A set of nine equations that already exist in the literature, but presenting high values of difference with GLCM, were investigated and adapted in this study. The differences (which in some cases were very high) were eliminated in most of the descriptors. The few remaining differences present values very close to zero, with a ratio of 10−1. In addition, six other equations are proposed and discussed in this study. To validate these equations, a series of experiments are performed, using real and synthetic textures. Each texture contains different combinations that represent the same motif with different perspectives and scales. These served to validate the equations proposed in this study. Moreover, in applications always present the same pattern of behavior for the equivalent equations of the original ones described by using GLCM. All the results were favorable to the use of SDH that presented a reduction of the complexity wich speeds up to 98.6% the computational time.

Protein-ligand interaction (PLI) networks show how proteins interact with small non-protein ligands through noncovalent bonding. Understanding such interactions is a crucial step towards ligand prediction, target identification and drug design. We propose CALI (Complex network-based Analysis of protein-Ligand Interactions), a graph-based, visual strategy coupled with complex network topological properties to summarize and detect frequent patterns in PLIs. Patterns obtained with CALI were compared to experimentally determined protein-ligand interactions from the CDK2 and Ricin dataset. For the CDK2, CALI found 90% of interacting residues, and all residues of the Ricin that interact with ligands. We devised a powerful visual and interactive strategy to analyze the data, providing a general view of the interaction dataset, showing the most common PLIs from a global perspective.

Socio-technical systems play an important role in public health screening programs to prevent cancer. Cervical cancer incidence has significantly decreased in countries that developed systems for organized screening engaging medical practitioners, laboratories and patients. The system automatically identifies individuals at risk of developing the disease and invites them for a screening exam or a follow-up exam conducted by medical professionals. A triage algorithm in the system aims to reduce unnecessary screening exams for individuals at low-risk while detecting and treating individuals at high-risk. Despite the general success of screening, the triage algorithm is a one-size-fits all approach that is not personalized to a patient. This can easily be observed in historical data from screening exams. Often patients rely on personal factors to determine that they are either at high risk or not at risk at all and take action at their own discretion. Can exploring patient trajectories help hypothesize personal factors leading to their decisions? We present Portinari, a data exploration tool to query and visualize future trajectories of patients who have undergone a specific sequence of screening exams. The web-based tool contains (a) a visual query interface (b) a backend graph database of events in patients' lives (c) trajectory visualization using sankey diagrams. We use Portinari to explore diverse trajectories of patients following the Norwegian triage algorithm. The trajectories demonstrated variable degrees of adherence to the triage algorithm and allowed epidemiologists to hypothesize about the possible causes., Conference paper published at ICSE 2017 Buenos Aires, at the Software Engineering in Society Track. 10 pages, 5 figures

Interactions between proteins and ligands are relevant in many biological processes. In the last years, such interactions have gained even more attention as the comprehension of protein-ligand molecular recognition is an important step to ligand prediction, target identificantion, and drug design, among others. This article presents GReMLIN (Graph Mining strategy to infer protein-Ligand INteraction patterns), a strategy to search for conserved protein-ligand interactions in a set of related proteins, based on frequent subgraph mining, that is able to perceive structural arrangements relevant for protein-ligand interaction. When compared to experimentally determined interactions, our in silico strategy was able to find many of relevant binding site residues/atoms for CDK2 and active site residues/atoms for Ricin.

Mammographic Computer-Aided Diagnosis systems are applications designed to assist radiologists in diagnosis of malignancy in mammographic findings. Most methods described in the literature do not perform a proper preprocessing step in mammographic images prior to classification, which can generate inconsistent results due to the potentially large amount of noise in medical images. This paper proposes a new method based on Information Theory and Data Compression for detection of random noise in image bit planes. In order to validate the efficiency of the proposed noise removal method, we used Machine Learning algorithms to classify mammographic findings from the Digital Database for Screening Mammography. Results using texture features indicate that a reduction in the radiometric resolution of 4 or 5 bit planes in digitized screen film mammographic images result in a better classification performance.

In this paper, we propose an interactive visualization called ADVISe (Annotation Dynamics Visualization), which tackles the problem of visualizing evolutions in enzyme annotations across several releases of the UniProt/SwissProt database. More specifically, we visualize the dynamics of Enzyme Commission numbers (EC numbers), which are a numerical and hierarchical classification scheme for enzymes based on the chemical reactions they catalyze. An EC number consists of four numbers separated by periods and represents a progressively finer classification of the catalyzed reaction. The proposed interactive visualization gives a macro view of the changes and presents further details on demand, such as frequencies of change types segmented by levels of generalization and specialization as well as by enzyme families. Users can also explore entry metadata. With this tool, we were able to identify trends of specialization, database growth and exceptions in which EC numbers were deleted, divided or created and revisions of past annotation errors.

A experimental scheme for identification of breast diseases based on thermal images is presented. A set of infrared images from a database that is being developed were used for an experimental investigation considering the image Lacunarity measures by the gliding box algorithm. This approach generates a parameter to distinguish from normal to abnormal breast diagnostics. We propose two interpretations based in similarity of the breast anatomic aspects. The original contributions of this work are the use of thermal image on diagnosis of breast disease and an approach that classifies images based on lacunarity indexes.

Three aspects of texture are distinguished by fractal geometry: fractal dimension (FD), lacunarity and succolarity. Although, FD has been well studied and lacunarity has been more and more used, succolarity, until now, has not been considered. This work presents a method to compute succolarity. It considers its applicability and relations (differences and similarities) with of each other fractal measure, as well. The proposed method for this computation is based on the FD box counting approach adapted to the notions of succolarity. A simple example is shown step by step to easily explain how to compute the succolarity for binary images. Furthermore, it is shown that the proposed method is able to correctly distinguish social aspects of cities on satellite images. The main idea is the use of succolarity as a new feature in the pattern recognition process mainly for identification of natural textures. Succolarity works very well when the input images have direction and/or flow information associated with it.