Multi-Target Tracking of Human Spermatozoa in Phase-Contrast Microscopy Image Sequences using a Hybrid Dynamic Bayesian Network.

Male infertility is mostly related to semen and spermatozoa, and any diagnosis or treatment requires the investigation of the motility patterns of spermatozoa. The movements of spermatozoa are fast and involve collision and occlusion with each other. In order to extract the motility patterns of spermatozoa, multi-target tracking (MTT) of spermatozoa is necessary. One of the most important steps of MTT is data association, in which the newly arrived observations are used to update the previous tracks. Dynamic Bayesian network (DBN) is a powerful tool for modeling and solving various types of problems such as tracking and classification. There can also be a hybrid-DBN (HDBN), in which both continuous and discrete nodes are present. HDBN has a suitable structure for modeling problems that have both discrete and continuous parameters like MTT. In this research, the data association for MTT of human spermatozoa has been studied. The proposed algorithm was tested over hundreds of manually extracted spermatozoa tracks and evaluated using several standard measures. The superior results of the proposed algorithm in comparison to the other well-known algorithms, show that it could be considered as an improved alternative to traditional computer assisted sperm analysis (CASA) algorithms.