Sperm motility analysis: a novel algorithm for tracking image sequences

GUSTAVO MONTE; JUAN IGNACIO PASTORE; LUCIA, ZALAZAR; ANDREINA CESARI; BALLARIN, VIRGINIA

World Congress on Medical Physics and Biomedical Engineering

Congreso; World Congress on Medical Physics and Biomedical Engineering; 2018

Czech Society for Biomedical Engineering and Medical Informatics and for the Czech Association of Medical Physicists

Motion analysis on quality assessment of semen samples has a good correlation with male fertility and it is one of the most affected parameter after cryopreservation / thawing. However, sperm tracking is quite complex due to cell collision, occlusion and missed detection. making it difficult to calculate motility parameters for biological studies. We present a novel algorithm for tracking image sequences for the evaluation of sperm motility. The proposed tracking algorithm extracts essential information from a frame by minimizing redundant data and generating an output whose order of magnitude smaller with respect to the number of pixels in the image.The main idea of the proposed algorithm is to represent the signal as a concatenation of known functions instead of sequence of samples. In an oversampling condition, the value of the sample loses relative importance while the comparison of its value with the past samples allows inferring trajectory behaviors. Under this condition, the trajectories are smooth and the signal is proposed as originated by a generating function which contracts, dilates and adapts to follow the trajectory.Uniform sampling is maintained but the signal is oversampled with respect to the Nyquist frequency. A redundant vector is obtained from which the algorithm is fed. First a segmentation based on adaptive linear interpolation is performed, then the segments are labeled reflecting the behavior within it. With this information, sequences of segments that identify global behaviors are detected.The system was compared with the Microptic´s Sperm Class Analyzer® according to the monitoring performance, evaluating the accuracy of the system and the percentage of fragmented paths to measure the reliability of the monitoring. The results show that the proposed method presented a better performance according to these indicators. Finally, motility percentages were found to corroborate the largest number of detected mobile trajectories of our method.