Standardisation of Dog Sperm Morphology Classification

Sperm morphology is an important component of dog fertility analysis. Visual examination is the gold standard technique for evaluation, but large variations within and across evaluators have been documented. Databases of sperm morphology images are valuable practical aids for training and performance testing and can improve sperm classification standardisation; however, such databases are only useful when the material is expertly classified to serve as a reference. The objective of this study was to build an expertly annotated database of sperm images and to generate an illustrated guide with examples of the different sperm morphology presentations to promote standardisation in dog sperm morphology classification. Semen samples (n = 42) were obtained either from dogs presented for fertility evaluation or from semen doses used for artificial insemination. For image acquisition, wet preparations were prepared and evaluated using phase-contrast optics under 1000× magnification. Selected images were cropped so that only one sperm was included in each file, and a database of 8817 images was created. The images were classified by three veterinarians, specialists in animal reproduction (Diplomates American College of Theriogenologists) according to different criteria. Agreement among evaluators was the criterion used to create a database with reconciled classification. Complete agreement on defect category and morphological classification was observed on 70.2% of the images, and at least two veterinarians agreed on the classification of another 27.4% of the images; therefore, the reconciled database classification was based on the consensus of at least two veterinarians on 97.5% of the images. The percentage of morphologically normal sperm in the reconciled database was 57%, with the primary defect involving the sperm head, midpiece, or tail in 18%, 20%, and 5% of the cases, respectively. Among defective sperm, 39% had a single defect and 4% had multiple defects. When the total prevalence of specific defects was computed, the reconciled database contained 18% sperm head abnormalities, 22% midpiece abnormalities, and 7% tail abnormalities. Comprehensive illustrations of the variation in normal and abnormal dog sperm morphology from the database built based on reconciled expert classification are provided for reference.