Detection of Human Sperm Tracks Using Video Processing Techniques

Abstract

Sperm motility analysis is very important for human fertility assessment. It is often carried on manually, but this could be susceptible to mistakes due to the nature of procedure. In addition to being time-consuming, results are merely subjective and non-repeatable. In order to overcome this, we present a semi-automated algorithm that tracks accurately the sperm movements. Adaptive Gaussian models are implemented for detecting moving spermatozoa and segment them throughout video frames. Morphological operators and connected-components labeling are applied to reduce noise and calculate centroids, respectively. Then, the Munkres algorithm along with the Kalman filter are used for the purpose of assigning centroids to tracks. Finally, tracks are displayed on screen. Outcomes show a 90.91 % of accuracy regarding to manual analysis. This algorithm aims only to detect spermatozoa movement and trace its displacement in video for human sperm samples. Moreover, it allows andrology experts to perform a more exact analysis of the individual characteristics of spermatozoa, having so a low cost, accurate and repetitive technological support that will allow them to emit more precise diagnosis. Thus, this method will help specialists to reduce time periods and make more objective analysis of sperm motility. In this way, fertility diagnosis will be more reliable.