Abstract
Training to improve the standardization of subjective assessments in biological science is crucial to improve and maintain accuracy. However, in reproductive science there is no standardized training tool available to assess sperm morphology. Sperm morphology is routinely assessed subjectively across several species and is often used as grounds to reject or retain samples for sale or insemination. As with all subjective tests, sperm morphology assessment is liable to human bias and without appropriate standardization these assessments are unreliable. This proof-of-concept study aimed to develop a standardized sperm morphology assessment training tool that can train and test students on a sperm-by-sperm basis. The following manuscript outlines the methods used to develop a training tool with the capability to account for different microscope optics, morphological classification systems, and species of spermatozoa assessed. The generation of images, their classification, organization, and integration into a web interface, along with its design and outputs, are described. Briefly, images of spermatozoa were generated by taking field of view (FOV) images at 40× magnification on DIC optics, amounting to a total of 3,600 FOV images from 72 rams (50 FOV/ram). These FOV images were cropped to only show one sperm per image using a novel machine-learning algorithm. The resulting 9,365 images were labelled by three experienced assessors, and those with 100% consensus on all labels (4821/9365) were integrated into a web interface able to provide both (i) instant feedback to users on correct/incorrect labels for training purposes, and (ii) an assessment of user proficiency. Future studies will test the effectiveness of the training tool to educate students on the application of a variety of morphology classification systems. If proven effective, it will be the first standardized method to train individuals in sperm morphology assessment and help to improve understanding of how training should be conducted.