Automated follicle segmentation in ultrasound scans of bovine ovaries

Abstract

Important reproductive information can be determined by analyzing the follicles in an ultrasound scan of the ovary. The currently used process requires an expert to manually mark and measure the size of the follicles. Manual analysis is time-consuming and often conflicting. A robust computer assisted approach to automatically detect and quantify follicles would improve both the efficiency and accuracy of the process. The multi-step algorithm presented involves coarse ovarian segmentation, follicle detection, and post processing using level sets to automatically determine follicle boundaries. In comparison with expert evaluations, experiments on bovine ovarian images show encouraging results. The algorithm correctly detected approximately 65% of the ovarian follicles.