Date of Award
Daniel Felix Ritchie School of Engineering and Computer Science, Computer Science
Anneliese Andrews, Ph.D.
Catherine Durso, Ph.D.
Male fertility, Spermatoza, Assisted reproductive technology, Computer-assisted sperm analysis
Effective Assisted Reproductive Technology (ART) relies in part upon accurate but easily conducted measurements of sperm motion parameters. Several established methods are widely used to assess possible reasons for male infertility, in human and veterinary Andrology clinics. Computer-assisted sperm analysis (CASA) devices quantitatively assess sperm motion parameters, which have been defined by the World Health Organization, and include the percentage of motile cells in a sample and the motion characteristics of individual cells, such as curvilinear velocity (VCL), average path velocity (VAP) and straight line velocity (VSL). However, CASA analyses fail to define hyperactive sperm motility or determine the prevalence of hyperactively motile sperm in the sample. Hyperactively motile sperm swim in an erratic pattern, and this occurs only at the very end of sperm capacitation, a series of biochemical changes occurring in a sperm which enables it to fertilize an oocyte. The computational challenge for detecting hyperactivated sperm motility lies in precisely modeling sperm movement changes that accurately reflect the sperm's biomedical function, by developing an algorithm that detects and classifies these unique motility patterns. Currently, no such algorithms reliably classify hyperactivated spermatozoa. Therefore, several methods to automatically identify and classify hyperactivated spermatozoa trajectories are described and their performance compared to 'the gold standard' of visual classification, by experts.
The methods considered were: two existing methods, a mathematical modification to one of these, and three new methods, each examined independently and then two were combined to produce an integrated approach.
Evaluation of each method was performed by using each to analyze an initial data set containing tracks of hyperactivated and progressive sperm, which had been classified by experts in the field, and then to analyze data sets obtained from actual laboratory samples. Classifications as well as misclassifications were recorded in diffusion matrices. Two methods, the Minimum Bounding Square Ratio (MBSR) and the Rotated Rectangular Linearity (RRL) were more effective in accurately detecting hyperactivated sperm and were similar in correctly classifying hyperactivated sperm. However, RRL misclassified twice as many sperm as MBSR. MBSR also outperformed the other methods in correctly classifying progressively motile sperm and sperm exhibiting transitional motility.
After developing this algorithm, it was applied to evaluate sperm from a large experiment to determine if sperm treated with different phosphodiesterase inhibitors, used in erectile dysfunction drugs, exhibit sperm motility. The experiment would not have been possible without these new computer algorithms. Taken together, this research demonstrates that newly developed algorithms can be used to identify critically important features of sperm, such as hyperactivity. One algorithm, MBSR may become an important tool improving Assisted Reproductive Technology's success
Copyright is held by the author. User is responsible for all copyright compliance.
Received from ProQuest
Kaula, Norbert, "A Computer Science Approach to Identify and Classify Hyperactivated Spermatozoa" (2009). Electronic Theses and Dissertations. 841.
Computer science, Physiology