My thesis is linked to the topic of routine medical urine analysis. I presented urine sediment analysis techniques, and uncertainty factors of manual microscopic methods in detail. I compared UriSed automatic urine sediment analysis method of 77 Elektronika Kft. – which is based on pattern recognition – to manual microscopy, as the gold standard method. An essential element of my work was to determine the effectiveness of UriSed technology, and to further develop the particle recognition module.
I compared these two diagnostic techniques, using 500 urine samples and biostatistical methods. First, it was necessary to make the calculation of statistical indicators, such as specificity, sensitivity, positive and negative predictive values, agreement of the two methods, and McNemar factor to determine the abnormal particle concentrations in UriSed method. An important element of my work was to develop new critical concentration levels (cut-off values) which may prove to be better in practice, than the cut-off levels that were used before.
A crucial part of my thesis was to develop a computer program, which aims to improve particle recognition quality. My program is capable of the optimization of decision-making modules by selecting the the modules that can perform together the best. The program executes particle recognition in all modul combinations and calculates an optimization factor for all these combinations. During investigations on testfiles I proved that significant improvements were achieved by appropriate selection of decision-making modules.
In conclusion, I can say that the UriSed method performs well in terms of diagnostic accuracy and is also advantageous in many other ways, compared to manual microscopy. Thanks to my thesis the further development of this world-famous automated urine diagnostic method was succefully implemented.