A new region-based adaptive thresholding for sperm motility segmentation

Infertility cases have shown increasing growth in recent years where approximately 40% of root causes of infertility cases are related to men. Researchers have shown that the sperm motility has significantly contributed towards infertility as compared to its concentration and morphology. However, existing technique has faced difficulties in segmenting motile sperm in the low contrast regions. The movements of the motile sperms that are normally fast further complicate the automated segmentation. In this paper, we present a new region-based adaptive thresholding technique that consists of four main stages. Pixels of the images are classified based on the intensity distribution and those pixels are grouped and processed separately. Multiple thresholds are generated based on the classified group to ensure objects in low contrast regions are segmented. In addition, the proposed method does not require external pre-processing tool or phase contrast accessories prior to the sperm segmentation. Our experimental evaluations show that the proposed method produces significant improvement from the existing technique with the average accuracy of 95.74%. The qualitative results also indicate that the proposed method is able to segment the motile sperms in the low contrast region. These results of sperm segmentation are in agreement with the quantitative measurement of non-uniformity where the proposed method attains lower non-uniformity with respect to the results achieved by the other method