Automatic Detection of Regions of Interest in Makler Images by Combinational Approach and Sperms Analysis by Fuzzy C-means

Abstract

Background: The male-based infertility test known as a spermiogram involves a manual count using a Makler counting chamber. There is a need to develop an automated sperm-counting system to provide more precise diagnoses. To that end, the automatic detection of Regions of Interest (ROI) in Makler images constitutes the first phase to using the advantages of the Makler chamber in a computerized counting system. Methods: ROI are defined between grids, hence, another challenging issue, that of exact grid detection, is examined. In this study, initially we reviewed several line detection algorithms with their applications and possible usage on the grid-detection problem of Makler images. Next, a combinational grid-detection technique, particularly for Makler images, was improved upon. Results: In summary, the Hough transform method has been enhanced by a combined approach of using Line Segment Detector, the clustering of slope angles, and post processing. The K-means method is deployed to refine the grids and to find the direction of grid lines to use in Hough transform. In the grid-detection step, the presented technique is evaluated with a template-matching technique following the Sorensen-Dice index. It gives 95.3% accuracy and 88.5% F-measure scores. Discussion: ROI extraction is performed based on grid detection output by multiple logical queries. Each extracted region, clarified from the grid lines, was identically examined for sperm count. Fuzzy c-means clustering was first performed to segment the objects in ROI, then blob analysis was utilized to eliminate non-sperm objects. Coclusion: The proposed sperm analysis approach was then compared to the visual assessment technique. Results indicate that the proposed system might be useful in laboratories, but still needs to be improved in the feature extraction process.