A digital image based method to quantify intracellular polyphosphates in microbial aggregates
Year: 2018
Type: dissertation
Abstract: Eutrophication is a serious problem of aquatic ecosystems that can threaten biodiversity and human health. It is caused by a high concentration of nutrients, such as phosphorus, leading to excessive growth of aquatic plants and, consequently, to stagnation and decreased levels of oxygen in water systems. To prevent this problem, wastewater from rural, industrial and urban backgrounds are treated using enhanced biological phosphorous removal. This technique uses microorganisms already present in wastewater that incorporate phosphorous into their cells in the form of polyphosphate, removing the nutrient from the water. A very important part of this technique consists of evaluating the polyphosphate levels present in the microorganisms, which is usually done through expensive or time consuming analytical methods. To improve the polyphosphate measurement step, the objective of this thesis was to create a quantitative image analysis program that can correlate image obtained data with analytical data, creating linear regression models useful for polyphosphate quantification. This program used color segmentation between polyphosphate and biomass regions, based on RGB, HSV and LAB color spaces, and was applied on images of samples dyed with different stains (methylene blue and toluidine blue) and with different dilution factors (no dilution and 10x dilution). These images experienced gamma values and contrast adjustments, to evaluate different image settings on color segmentation. Models for methylene blue stained and diluted samples presented the highest regression coefficients between the image data results and analytical data (0.920 and 0.992, respectively), also existing some differences between models of images with different contrast and gamma values. The color space chosen for image analysis seems to be very important with RGB and LAB presenting more satisfying results than HSV. More work is needed before this method can be used, however the results seen on this thesis are promising.
Author: André Paço da Fonseca
Institution: University of Minho
Language: English
Subfield: Pollution
Field: Environmental Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W3081488037
