This work proposes an algorithm oriented to the detection of tuberculosis bacilli in digital images of sputum samples, inked with the Ziehl Neelsen method and prepared with the direct, pellet and diluted pellet methods. The algorithm aims at automating the optical analysis of bacilli count and the calculation of the concentration level. Several algorithms have been proposed in the literature with the same objective, however, in no case is the performance in sensitivity and specificity evaluated for the 3 preparation methods. The proposed algorithm improves the contrast of the colors of interest, then thresholds the image and segments by labeling the objects of interest (bacilli). Each object then has its geometrical descriptors and photometric descriptors. With all this, a characteristic vector is formed, which are used in the training and classification process of an SVM. For the training 225 images obtained by the 3 preparation methods were used. The proposed algorithm reached, for the direct method, a sensitivity level of 93.67% and a specificity level of 89.23%. In the case of the Pellet method, a sensitivity of 92.13% and a specificity of 82.58% was obtained, while for diluted Pellet the sensitivity was 92.81% and the specificity 83.61%.

Background: Direct sputum smear is still the first-choice tool for screening of tuberculosis worldwide. Variants of this technique, to improve the sensitivity are desired. Methods: Two microbiological variants of the standard sputum smear (”pellet” and “diluted-pellet”) for both Ziehl-Neelsen (ZN) and auramine fluorescence (AF) staining were evaluated. In addition, two methods for concentration of mycobacteria in sputum, using positive and negative pressure filtration, were tested and compared. The evaluation of the microbiological variants was performed on 98 culture positive sputum samples from different TB patients. The diagnostics sensitivity and the level of detritus in the processed sputum were determined. Bacilli load in the smear variants was determined by microscopic observation and by manual inspection of microscopic digital images. The comparison of the mycobacteria filtration methods was performed on 76 smear positive sputum samples. Filters retaining the concentrated mycobacteria were stained with AF and compared with the direct smear. Bacilli load, detritus level, filtered volume, filtration time and background noise level, were determined. Results: The sensitivity of microscopy with the microbiological variants was 7.1% and 2% higher in ZN and AF respectively, compared to direct smear. The sensitivity of AF in diluted pellet was significantly higher than all ZN variants (P < 0.05). Detritus level observed in slides was significantly lower in the diluted pellet than the pellet and direct smear in ZN and AF (P < 0.001). A significant increase in the bacilli load in microscopic observation and digital images analysis was observed in pellet and diluted pellet than the direct method (P <0.0001). The concentration of mycobacteria using positive-pressure filtration showed a trend to produce a higher bacilli load compared to the negative-pressure filtration and direct smear, although it was not significant. Detritus levels were significantly higher in both variants of filtration (P < 0.0001). Filtered volumes were higher in positive-pressure compared to negative-pressure filtration. Filtration times were significantly higher in negative-pressure compared to positive-pressure filtration (P < 0.0001). Conclusion: The proposed variants improved the performance of the standard sputum smear, making it an important test for settings with high rates of smear-negative TB cases.

In this article is developed a technique that allows to calculate the presence of vegetation, glaciers and water bodies through multispectral image processing employing a Multi-layer Perceptron Neural Netwok, giving the option to discriminate the presence of lakes to generate the cadastral registration of these. The supervised classification that was implemented has a high level of robustness and reliability, since the validation of the data obtained at a geolocation level have a 0% of error and the parameters of the area and perimeter an approximate error of 10%.

This work proposes a method to estimate landscape units contained in a region (using satellite WorldView-2 imagery as input) for urban planning. Number of landscape units contained in a region and their extension are estimated by a graph-based segmentation algorithm while the composition of each landscape unit is estimated by a modular neural network. The proposed method, despite of the subjectivity of what represents a landscape unit, achieves the following results: vegetation estimation accuracy: 99.58%, water estimation accuracy: 98.08%, urban area estimation accuracy: 90.38% and soil estimation accuracy: 90.25%, over 2400 testing pixels (600 pixels/image - 4 satellite images).