Evaluation of Multi-Target Human Sperm Tracking Algorithms in Synthesized Dataset

Evaluation of Multi-Target Human Sperm Tracking Algorithms in Synthesized Dataset

Abdollah Arasteh (School of Electrical Engineering, Sharif University of Technology, Tehran, Iran) and Bijan Vosoughi Vahdat (School of Electrical Engineering, Sharif University of Technology, Tehran, Iran)
DOI: 10.4018/IJMSTR.2016040102


Infertility is an important issue for many couples and male infertility is highly related to semen and spermatozoa which can be surveyed by means of semen/sperm analysis. Many fertility assessment laboratories are now equipped with automatic systems called Computer Aided Semen/Sperm Analysis (CASA) for doing this task. Evaluation of such systems is very important. In this research a web-based simulator is developed which facilitates evaluation of CASA systems. The developed software has many useful parameters such as blurring images or adding noise and it also gives full control of sperm counts and types. To illustrate performance of the developed simulator, many parameters such as spermatozoa population, standard deviation of Gaussian blur filter and noise intensity have been swept and the results of two well-known multi-target tracking systems (Linear Kalman Filter and Particle Filter) were compared and discussed.
Article Preview


One in every four couples in developing countries had been found to be affected by infertility (Mascarenhas, Flaxman, Boerma, Vanderpoel, & Stevens, 2012; WHO, 2012), so infertility is one of the most important issues for many couples. Male infertility is related to semen and sperm cells in majority of cases and can be measured by semen and sperm cell analysis for more advanced diagnosis and treatment (D. Mortimer, Pandya, & Sawers, 1986; Nieschlag & Lenzi, 2013; Staff, 2010). Many fertility assessment laboratories are now equipped with automatic systems called Computer Aided Semen/Sperm Analysis (CASA). A CASA is a setup of hardware and software together which can observe sperm cells throughout a microscope lens and measure many parameters of them including their speed, average path, total movement, and finally classify them into one of three standard classes defined in (WHO, 2010a).

After extraction of each sperm cell trajectory, all measured parameters can be calculated, so they are all functions of sperm cell trajectory. It means that precision in extraction of sperm cell trajectory causes precision in measured parameters and imprecise sperm cell trajectory results in imprecise measured parameters. Hence, the major problem to solve is precise extraction of trajectories of sperm cells, which means a good performance in multi-target tracking. There are many multiple object tracking algorithms developed for wide variety of problems such as tracking cars, humans, sperm cells, etc. (Chen, Li, Qin, & Hao, 2015; Fu & Han, 2012; Sørensen, Østergaard, Johansen, & de Bruijne, 2008), but cell and especially sperm cell tracking is a particular problem, which has its own considerations.

Evaluation of multi-target tracking algorithms needs very wide range of data to be valid, but Video sequences recorded and reported are very limited in number of cells and subjects for sperm tracking problem. The lack of an appropriate dataset should be considered for developing new algorithms or evaluating currently available ones.

Sperm cells movement recording mostly depends on the sample and there is no control on the parameters such as cell concentration, movements speed (or speed distribution), and other cells (like debris or white blood cells), so for recording a suitable sample, the only available procedure is recording more and more data from new subjects. Among these drawbacks, there is a big obstacle ahead for evaluation of algorithms and that is the real sperm cell positions are not available. It means if there should be an evaluation for an algorithm, the real trajectory of each sperm cell must be available for error calculation, however, because the recorded samples are just an image sequence, there is no information about the real position of each sperm cell unless the manual annotation is done, which is a very tedious and cumbersome task.

Considering all aforementioned reasons, developing a software simulator with controllable parameters at hand may be a helpful tool for evaluation and comparison of different multi-target tracking algorithms.

Developing a portable software which can be run properly in many platforms e.g. Windows, Linux, Mac OS, Android, iOS, etc. is an important issue. Web application is a very good choice for maximizing portability in life cycle of a software (Abran, Al-Sarayreh, & Cuadrado-Gallego, 2013; Kantee & Vuolteenaho, 2006; Wasserman, 2010). There is now a very good opportunity for developing graphic web application with the help of many new methods and technologies such as HTML5 and Canvas, SVG and D3JS libraries, etc. D3JS is a library developed in the Stanford University (Data Driven Documents website, n. d.) for better visualization and presentation of web graphics. It has many useful tools and functions available for developing graphic-based web applications.

In this research, the previous study (Arasteh & Vahdat, 2015) is extended. In the previous study a web application for simulation of sperm cells movement and generating related image sequence was developed using D3JS library. The output of this simulation was then used as input to multi-target tracking algorithms which were written by MATLAB.

Complete Article List

Search this Journal:
Open Access Articles: Forthcoming
Volume 5: 4 Issues (2017)
Volume 4: 4 Issues (2016)
Volume 3: 4 Issues (2015)
Volume 2: 4 Issues (2014)
Volume 1: 4 Issues (2013)
View Complete Journal Contents Listing