Sperm Icon

Classification and Clustering of Human Sperm Swimming Patterns

Authors: ∗J. Choi, ∗C. Wang, L.F. Urbano, ‡P. Masson, §M. VerMilyea, ∗M. Kam. *New Jersey Institute of Technology, Newark; ‡Penn Fertility Care, Philadelphia; §Ovation Fertility, Austin.

Summary abstract: The principal observed that the progressive swim types of sperm cells are linear mean and circular swim. Using motility characteristic parameters produced by CASA systems, we performed a parameter subset search to produce distinct clusters of the different swim types. For this task, the artificial bee colony algorithm (an iterative search algorithm modeled after the collective behavior of bees) and the well-studied k-means clustering algorithm were used on simulated and human sperm swim data. The result was distinct clusters with features of each type of swim. The clustering approach displays potential as a tool for automated sperm swim subpopulation analysis.

What is known already: Sperm cell movements can be divided into three major categories: progressive, non-progressive and immotile. Immotile movements define cells that show no movement at all, and non-progressive movements are defined by motility that lacks progression, where the sperm head shows little displacement over time. Progressive movements characterize actively moving sperm cells. The two progressive movements, linear-mean and circular, differ by the presence and absence of head rolling. Sperm cells of different movement types have different values of motility parameters, therefore clustering by motility parameters may provide a means of identifying the subpopulation of swimming patterns. Such clustering would allow better understanding of the cause and importance of these two swimming patterns.

Participants/materials, setting and methods: Video clips of human semen specimens prepared and collected by the IVF laboratories at Penn Fertility Care were used in this study. The video clips are 200 magnified images of 640×480 pixel resolution (0.857 μm/pixel). The sperm swim track data were obtained from the video clips by segmentation, localization and track data association process. In an attempt to find the subset of motility parameters for k-means clustering that would provide distinct clusters of each swimming type, we used the artificial bee colony (ABC) algorithm.

Main results: K-means clustering guided by the ABC algorithm (for parameter selection) provided distinct clusters of the different sperm swimming types. The algorithm produced clusters of moving sperm swim types characterized by the presence or the absence of sperm head roll.

Limitations, reasons for caution: For both sperm samples, using more parameters than the selected parameters chosen by the ABC algorithm did not provide better clusters. Also, the parameters used for clustering were not the same for the two samples. The use of the ABC algorithm allows sample-specific clustering, which leads to clusters whose membership can be described by swimming characteristics. We were able to separate cells based on the types of the two progressive movements from each other. In all the examples, three clusters were specified for the k-means clustering. Attempts to increase the number of clusters to four resulted in poor performance.

Wider implications of the findings: The ability to categorize sperm swim movement could provide a tool that will lead to a better understanding of sperm swimming patterns.