Elastic shape models for sperm morphology

Male fertility relies on a standardized analysis of the semen parameters, which include the mobility, concentration and the morphology of the sperm cell in a collected human semen sample. Mobility can be captured by microscope video camera, concentration can be obtained by counting the number of sperm cell in the a unit semen sample, whereas the morphology of the sperm can be detected by the microscope imaging. Among these parameters, sperm morphology worth further analysis using the imaging technique due to the optimized and advancement of computer techniques.

Usually, sperm test is done by human specialists in hospitals and research institutes. And complying with regular criteria from World Health Organization, this test is used to determine whether human sperm can function successful or not in fertilization, and finally give birth to a live birth. There are multiple tests available for evaluating different aspects of these functions. In order to use these functional tests accurately, clinicians must understand the content of the test, the indications for the test, and how to interpret the output to guide other tests or patient management.

One interesting subject in sperm morphology analysis (SMA) involves the application of automated algorithm, which can accelerate the detection of typical and atypical forms in recent years. There are different types of sperm abnormalities in general. For example, morphological head defect includes conical, pear-shaped, round, pin shape; an amorphous head, two heads, with or without acrosome, and vacuolar head are also counted as defect sperm. Neck and mid-segment defects consist of curved or asymmetric neck, and thin or thick midpiece. Tail defects are those such as short, folded and rolled tails and two tails. Cytoplasm exceeding one third of the head size can also cause defects. However, for a normal sperm that is able to fertilize, the criteria to make judgement is certain. In this research, we will present a novel technique that combine the elastic shape into the morphological analysis of human sperm selection.

All kinds of efforts have already been introduced to assist the assessment of sperm morphology detection. However, due to the limitations of imaging techniques in the past years, lots of manual methods are still subjective, sub-precise, non-repeatable and have large variability. Even methods of Computer-assisted sperm morphology assessment (CASA) reflect variation in most situations.

For the purpose of developing a new method to further assist the specialists to free them from tedious manual detection, an automated system based on Bayesian elastic active contour model is introduced to help detecting sperm cells in sperm images that come from the microscope. With each sperm cell detected, shape registration is applied to locate different parts on sperm contours. Morphological parameters of sperm cells are computed to compare with the World Health Organization (WHO) laboratory manual. We have evaluated our method on more than 100 real sperm images collected from the patients data in the hospital, where each images contain around 10 to 40 sperm cells randomly. The experimental results have shown great efficiency and accuracy. The proposed method also has great potential in other applications.