Geraldo Braz

Breast Tissues Classification Based on the Application of Geostatistical Features and Wavelet Transform

Female breast cancer is the major cause of death in occidental countries. Efforts in computer vision have been made in order to help improving the diagnostic accuracy by radiologists. We propose a methodology to distinguish Mass and Non-Mass tissues on mammograms. It is based on the computation of geostatistical measures (Morans Index and Gearys Coefficient) over a multiresolution image representation trough wavelet transform. The computed measures are classified through a Support Vector Machine (SVM). The methodology reaches 98.36% of Specificity, 98.13% of Sensitivity and a rate of 98.24% to discriminate Mass from Non-Mass elements, using the Gearys Coefficient application.

Classification of breast tissues using Getis-Ord statistics and support vector machine

Female breast cancer is the major cause of cancer-related deaths in western countries. Efforts in computer vision have been made in order to help improving the diagnostic accuracy by radiologists. In this paper, we present a methodology that intends to use Getis Index spatial texture measures in order to distinguish mass and non-mass tissues extracted from mammograms. The computed measures are classified through a One-Class and a Two-Class Support Vector Machine (SVM). The proposed method reaches 99.33% of accuracy using One-Class SVM and 94.21% of accuracy using Two-Class SVM.

Semi-automatic methodology for augmented panorama development in industrial outdoor environments

Abstract The use of virtual reality (VR) environments based on augmented panoramas enables the creation of visualization scenarios in different application areas. These panorama-based environments significantly reduce the cost and the timeframe of developing VR applications. The use of this visualization modality in industrial environments is appropriate because in these scenarios there is little or no change of the viewable physical environment during long periods of operation. The combined use of VR concepts and augmented reality (AR) enables the creation of applications that use object detection based on image characteristics and the inclusion of information through an augmented environment. With this in mind, this paper presents a methodology for creating VR environments based on augmented panoramas that uses semi-automatic object detection using Haar-like features and real images of the operating environment. The methodology was used to create an augmented environment based on an electric power substation as a case study. This environment is used for visualizing power substation equipment variables, and it shows the possibilities of using several panorama formats for the visualization.

Augmented visualization using homomorphic filtering and Haar-based natural markers for power systems substations

Abstract This paper presents an approach for annotating real data of power system equipment and has a main goal of improving visualization in outdoor scenarios where lighting presents itself as a problem for the detection of objects. It proposes the creation of object detectors as natural markers using Haar-like features and homomorphic filtering to include real information in an augmented visualization of a substation. The proposed system provides a real-time solution for displaying the data that are acquired from the SCADA/EMS automation system over the real scenario of the substation by providing an augmented visualization. The proposed system achieves an acceptable response time and the object detection step receives updates on each frame from the camera. Thus, it allows the use of augmented reality within operation and maintenance activities in the substation equipment, thereby providing data visualizations at the location where the demand exists instead requiring one to move to the control room to visualize the actual systems status. Equipment detection is performed on the video camera of a mobile device, frame by frame, by using a cascade classifier that is based on Haar-like features for the training and detection processes and by applying homomorphic filtering to reduce illumination problems. The proposed system can be used for training several detectors for substation equipment with the same technique. As a proof of concept, this work presents the results that are obtained using the power transformer. Thus, an augmented reality system prototype was developed that achieved good detection rates, thereby showing that the use of theses features is promising for augmented reality implementation in the daily routines of an electrical company.

Semivariogram and Semimadogram functions as descriptors for AMD diagnosis on SD-OCT topographic maps using Support Vector Machine

BackgroundAge-related macular degeneration (AMD) is a degenerative ocular disease that develops by the formation of drusen in the macula region leading to blindness. This condition can be detected automatically by automated image processing techniques applied in spectral domain optical coherence tomography (SD-OCT) volumes. The most common approach is the individualized analysis of each slice (B-Scan) of the SD-OCT volumes. However, it ends up losing the correlation between pixels of neighboring slices. The retina representation by topographic maps reveals the similarity of these structures with geographic relief maps, which can be represented by geostatistical descriptors. In this paper, we present a methodology based on geostatistical functions for the automatic diagnosis of AMD in SD-OCT.MethodsThe proposed methodology is based on the construction of a topographic map of the macular region. Over the topographic map, we compute geostatistical features using semivariogram and semimadogram functions as texture descriptors. The extracted descriptors are then used as input for a Support Vector Machine classifier.ResultsFor training of the classifier and tests, a database composed of 384 OCT exams (269 volumes of eyes exhibiting AMD and 115 control volumes) with layers segmented and validated by specialists were used. The best classification model, validated with cross-validation k-fold, achieved an accuracy of 95.2% and an AUROC of 0.989.ConclusionThe presented methodology exclusively uses geostatistical descriptors for the diagnosis of AMD in SD-OCT images of the macular region. The results are promising and the methodology is competitive considering previous results published in literature.

Lung nodule classification based on shape distributions

This paper presents the application of shape distributions to diagnose lung nodules in computerized tomography images. Our study uses the nodule surface and also interior surfaces generated by their voxels joint attributes and distribution. The surfaces are captured using 3D Alpha Shapes algorithm. All surfaces are characterized using shape distributions D1, D2, D3, D4 and A3 in order to represent shape behavior based on statistical distribution of their points in a 3D contour. We evaluate experiments with these features using Support Vector Machines and the combination of all descriptors proving above of 90% of accuracy. This indicates a promising feature to distinguish malignant from benign nodules for lung cancer diagnosis.

An Immersive Virtual Reality Application for Collaborative Training of Power Systems Operators

The use of immersive Virtual Reality applications for training in industrial areas has been increasing due to the benefits related to that technology. This paper presents an application to perform training of power system operators in a collaborative and immersive environment. This application aims to enhance the user immersion and increase collaborative training in a Virtual Reality using Collaborative Virtual Environment and a Problem Based Learning approach. It was build in Unity engine and presents a fully integrated scenario of power system visualization with a supervisor module that improves training through the simulation of real events.

DIAGNOSIS OF BREAST REGIONS THROUGH THE USE OF RIPLEY'S K FUNCTION AND SVM

Breast cancer has become increasingly common among the female population over 40 years and is the type of cancer that affects more women worldwide. One way to early detect non-palpable tumors that cause breast cancer is to perform an X-ray (mammogram) of the breasts. It is known that the chances of curing breast cancer is high if detected in early stages. However, the sensitivity of this test may vary widely due to factors such as examination quality or specialist experience. Thus, the use of diagnose systems in order to assist the specialist, have increased the chances of correct diagnoses. This paper presents a methodology for image spatial texture analysis and recognition of patterns present in mass extracted from images of mammograms, according to their malignant or benign behavior. Therefore, this paper uses the Ripley’s K function to extract texture and SVM for pattern recognition. The best result achieved was 85% accuracy, 88.23% sensitivity and 80.76% specificity with Az = 0.84.