Estimation of Boar Sperm Status Using Intracellular Density Distribution in Grey Level Images

In this work we review three methods proposed to estimate the fraction of alive sperm cells in boar semen samples. Images of semen samples are acquired, preprocessed and segmented in order to obtain images of single sperm heads. A model of intracellular density distribution characteristic of alive cells is computed by averaging a set of images of cells assumed to be alive by veterinarian experts. We quantify the deviation of the distribution of a cell from this model and use it for classification deploying three different approaches. One is based on a decision criterion used for single cell classification and gives misclassification error of 20.40%. The other two methods are focused on estimating the fraction of alive sperm in a sample, instead of single cell classification. One of them applies the least squares method, achieving an absolute error below 25% for 89% of the considered sample images. The other uses an iterative procedure to find an optimal decision criterion that equalizes the number of misclassifications of alive and dead cells. It provides an estimation of the fraction of alive cells that is within 8% of its actual value for 95% of the samples.