Research On Sperm Morphology Evaluation And Morphological Classification

Human Sperm Morphology research means through the assessment of spermmorphology to provide the reference for clinical diagnosis of andrology. Spermmorphology is one of the most important parameters of human semen analysis. It can beapplied to the diagnosis of male infertility, evaluation of contraceptive effectiveness,reproductive physiology research, assisted reproductive technology el.al. Because ofhuman sperm morphology diversity, difference of processing approach, as well as thevariation in criteria, sperm morphology assessment is the predicament of semen analysis.With the development of computer and information technology, sperm morphologyresearch based on microscope images, in somehow, make up for the deficiencies of thetraditional method based on slides, and made an exciting achievement. The thesis is alsobased on sperm morphology microscope images, using the method of image processing,pattern recognition, statistical analysis for sperm morphology assessment andclassification.The variation in sperm morphology assessment caused by the subjectivity ofevaluators is always the focus of andrologists. The research of the variation is helpful forthe phycians to reduce the variation and gain more consistent result of andrology diagnosis.Quality Control (QC) and evaluation training are important to reduce or avoid thevariations. The traditional method used for QA or training is cockamamie and timeconsuming, and the effect is limited. The method based on sperm microscope imagescollected the results of andrologists and put them into the software, making the QA andtraining more convenient and efficient. In addition, this method made the quality controland training routine possible. In computer assisted sperm morphology assessmenttechniques, the accurate classification of sperms is the precondition to get precisediagnosis. To obtain better classification results, extract sperm morphology featurecharacteristic parameters is the key issue. On the basis of traditional method, we increasedsome sperm morphology feature characteristic parameters, SVM, BP Neural Networkclassifiers were used to verify the effectiveness of these parameters, and a good result wasobtained. The main content of this thesis can be summarized in the following sections:1）Introduction of sperm morphology:background, parameters of sperm morphology,method, and the predicament;2）Through the analysis the variation in the results of andrologists, the characteristicand reasons of the variation were discussed;3）The software which can be used for QA and training was developed with theapplication of VB programing language and Access database; 4） The feature characteristic parameters were increased and the effectivenessverification were completed.