Search

Targeted environmental and ecosystem management remain crucial in control of dengue. However, providing detailed environmental information on a large scale to effectively target dengue control efforts remains a challenge. An important piece of such information is the extent of the presence of potential dengue vector breeding sites, which consist primarily of open containers such as ceramic jars, buckets, old tires, and flowerpots. In this paper we present the design and implementation of a pipeline to detect outdoor open containers which constitute potential dengue vector breeding sites from geotagged images and to create highly detailed container density maps at unprecedented scale. We implement the approach using Google Street View images which have the advantage of broad coverage and of often being two to three years old which allows correlation analyses of container counts against historical data from manual surveys. Containers comprising eight of the most common breeding sites are detected in the images using convolutional neural network transfer learning. Over a test set of images the object recognition algorithm has an accuracy of 0.91 in terms of F-score. Container density counts are generated and displayed on a decision support dashboard. Analyses of the approach are carried out over three provinces in Thailand. The container counts obtained agree well with container counts from available manual surveys. Multi-variate linear regression relating densities of the eight container types to larval survey data shows good prediction of larval index values with an R-squared of 0.674. To delineate conditions under which the container density counts are indicative of larval counts, a number of factors affecting correlation with larval survey data are analyzed. We conclude that creation of container density maps from geotagged images is a promising approach to providing detailed risk maps at large scale., Author summary Providing detailed environmental information on a large scale to effectively target dengue control efforts remains a challenge. In this paper we present the design and implementation of a pipeline to detect outdoor open containers which constitute potential dengue vector breeding sites from geotagged images and to create highly detailed container density maps at unprecedented scale. Specifically, we use convolutional neural networks to detect a variety of types of breeding site container types in Google street view images and use the container counts to create container density maps. Evaluation of the approach is carried out over three provinces in Thailand: Bangkok, Krabi, and Nakhon Si Thammarat. Our evaluation shows that the object recognition network can accurately recognize several of the most important types of containers in Thailand. The container counts obtained from the street view images agree well with container counts from available manual surveys. We further show that simple multi-linear models using container density values provide good predictions of Breteau index (number of positive containers per 100 houses inspected) values. This is the first study to present results validating container counts from image analysis against such data.

Introduction This paper discusses the presence of Aedes aegypti pupae in different types of containers considering: volume, pH of the container, among other variables. Methods A nonlinear method for selection was applied, based on Mutual Information, by placing in order of importance the most appropriate variables for identifying containers with and without Aedes aegypti pupae. Such variables were used for input into a Neural Network in layers for classification. Results Among the experiments carried out, the best result obtained used the first eight variables selected by order of importance. The percentage of hits for containers which had no Aedes aegypti pupae was 73.3%, and 80.9% for those which did have Aedes aegypti pupae in the containers. This Neural Network method, a model with the capacity to emulate non-linear data, got better results in comparison with the discriminant power of the Logistic Regression model. Thus, the outcomes of using the Neural Networks method achieved better separability in classifying the containers with pupae and those with no pupae. Conclusion This type of analysis will aid in the efforts to design an efficient program to control Aedes aegypti that can concentrate principally on containers which present the greatest productivity., Author summary The authors discuss a nonlinear method, based on Mutual Information, for selection of the presence of Aedes aegypti pupae in different breeding sites. The authors compare this method with the logistic regression model. In this study, using the Neural Network model, the variables that better discriminate the containers with Aedes aegypti pupae from the containers without pupae were size—the larger the container volume the greater the proportion of containers positive for pupae–, the presence of another type of pupa other than Aedes aegypti and the location of the container outside the home. The Neural Network method, a model with the capacity to emulate non-linear data, got better results in comparison with the discriminant power of the Logistic Regression model. Thus, the outcomes of the Neural Networks method achieved better separability in classifying the containers with pupae and those with no pupae. This type of analysis will aid in the efforts to design an efficient program to control Aedes aegypti that can concentrate principally on containers which present the greatest productivity.

In this paper we report a quantitative laser Biospeckle method using VDRL plates to monitor the activity of Trypanosoma cruzi and the calibration conditions including three image processing algorithms and three programs (ImageJ and two programs designed in this work). Benznidazole was used as a test drug. Variable volume (constant density) and variable density (constant volume) were used for the quantitative evaluation of parasite activity in calibrated wells of the VDRL plate. The desiccation process within the well was monitored as a function of volume and of the activity of the Biospeckle pattern of the parasites as well as the quantitative effect of the surface parasite quantity (proportion of the object’s plane). A statistical analysis was performed with ANOVA, Tukey post hoc and Descriptive Statistics using R and R Commander. Conditions of volume (100μl) and parasite density (2-4x104 parasites/well, in exponential growth phase), assay time (up to 204min), frame number (11 frames), algorithm and program (RCommander/SAGA) for image processing were selected to test the effect of variable concentrations of benznidazole (0.0195 to 20μg/mL / 0.075 to 76.8μM) at various times (1, 61, 128 and 204min) on the activity of the Biospeckle pattern. The flat wells of the VDRL plate were found to be suitable for the quantitative calibration of the activity of Trypanosoma cruzi using the appropriate algorithm and program. Under these conditions, benznidazole produces at 1min an instantaneous effect on the activity of the Biospeckle pattern of T. cruzi, which remains with a similar profile up to 1 hour. A second effect which is dependent on concentrations above 1.25μg/mL and is statistically different from the effect at lower concentrations causes a decrease in the activity of the Biospeckle pattern. This effect is better detected after 1 hour of drug action. This behavior may be explained by an instantaneous effect on a membrane protein of Trypanosoma cruzi that could mediate the translocation of benznidazole. At longer times the effect may possibly be explained by the required transformation of the pro-drug into the active drug., Author Summary Biospeckle refers to a pattern that occurs when a laser beam illuminates a dynamic surface, such as a liquid that contains microorganisms. The movement or the roughness of its surface causes the wave fronts to interfere and produce a pattern of moving dots that resemble boiling water. This research describes the application of Biospeckle to Trypanosoma cruzi, the parasite that causes Chagas disease. The purpose was to observe the movement of the Biospeckle dots and to detect differences depending on the presence of the parasite, the quantity of the parasite and the conditions of the parasites when they are affected by a drug. We designed a method using VDRL plates where the sample has a relatively small volume and is flat shaped, a laser, a camera and a lens. The Biospeckle pattern is recorded in a video in a computer and shows the Biospeckle dots which move rapidly as the concentration of parasites increases and less rapidly as the concentration decreases or as the parasites are affected by a drug such as benznidazole. We designed algorithms which take the difference between successive frames and expressed them in a program in Java, in a script in R Commander and SAGA and in ImageJ. Thus we obtained a quantitative description of the movement of T. cruzi.