A new fully automated system for the morphometric analysis of human sperm heads

To develop a fully automated image analysis system to provide objective, quantitative, and reproducible assessment of the head morphology of human spermatozoa.

Shorr-stained smears of washed sperm were imaged with a microscope (x100) and black and white video camera and digitized for computer analysis. Automatically selected sperm head images are analyzed in terms of 32 parameters that reflect size, shape, and staining heterogeneity. Unique features of the analysis include extraction of a 40% density contour in addition to the usual boundary contour and parameterization of optical density profiles. The morphometry of a semen sample is summarized by evaluation of the sample mean and standard deviation for each morphometric parameter. In addition, each analyzed sperm in a sample is assessed for conformity to a reference morphometric data set.

The system filters seminal debris, superimposed cells, and lysed sperm, with an average sperm recognition error of 2.3% and contour interpretation error of 1.7%. The average proportion of variance attributable to slide preparation and analysis methods (2.6%) was low relative to that between infertility patients (12.4%). The relatively slow assessment times would be improved substantially with upgraded hardware.

The precision and sensitivity offered by the detailed image analysis of this system provides a powerful tool for morphometric semen analysis.