Influência da qualidade de imagens e técnicas de rastreamento nos resultados de análise espermática assistida por computador (CASA-Mot) em peixes

Abstract

This doctoral dissertation is based on the elaboration of two scientific articles and a process that normalize specific execution steps in a CASA-Mot software: 1) the first article discusses and evaluates different image enhancement techniques available in ImageJ software wich has the potential to treat spermatic images when using the CASA-Mot plugin. For the tests, were recorded videos of the spermatic movement of gray jundiá (Rhamdia quelen) under different contrast conditions (high, medium and low) in the spermatic images, through the change in the condenser of the microscope. The effects of seven images enhancement techniques were checked under the results of the number of spermatozoa tracked, motility rate and curvilinear velocity. The images enhancement techniques Median, Bandpass and Rolling-Ball significantly improved the contrast between sperm and background at all contrast levels, producing reliable results; 2) the second article treats and evaluates the effects of spermatic tracking techniques of the Nearest Neighbor (NN) native from CASA-Mot plugin, and a robust Multiple Hypothesis Tracking (MHT) technique in simulated motion videos of the spermatic movement of fishes, containing different numbers of spermatozoa in the visual field (50, 100, 200 and 300 sperm) and different motility rates (30, 60 and 90%). The results presented by the different trackers were evaluated under motility rates, number of spermatozoa traces, number of tracking errors and time of analysis. All results were influenced (p < 0.05) by the tracking technique used; 3) The analytical process of sperm movement is described which comprises seven steps: importing video scenes, applying image enhancement technique according to the results of the first article, segmenting sperm images, tracking sperm using NN or MHT techniques, calculating motility parameters and provide results organized in specific directories. Additionally, the software screens developed to facilitate and automate the analysis are presented.