Deep Learning-enabled Sperm Morphology Analysis of Bovine Sperm for label-free Imaging Flow Cytometry

Data analysis of sperm morphology is critical for evaluating bull fertility, yet it is often performed using light microscopy and staining techniques in a highly subjective and manual manner. In this study, we introduce a scalable, high-resolution approach combining label-free Imaging Flow Cytometry (IFC) with deep learning for automated classification of bovine sperm morphology. We analyzed 436,374 single-cell images obtained at 40× magnification from three prominent bull breeds in Kazakhstan—Kazakh Whitehead, Auliekol, and Simmental from fresh and cryopreserved sperm—providing a uniquely large and diverse dataset. The dataset was used for training and evaluation of deep learning models, among which the convolutional neural network (CNN) MobileNetV4 yielded superior results, achieving 92.3% accuracy and a 0.91 F1-score after training with a Layer-wise Pretraining and Fine-Tuning (LP-FT) strategy. The model classified spermatozoa into eight distinct morphological categories. The CNN-based pipeline ensured consistent, observer-independent classification across all samples. Testing across different conditions and breeds resulted in a 5–10% drop in generalization performance, highlighting the impact of domain-specific biases and underscoring the need for larger, standardized datasets. The proportion of morphologically abnormal spermatozoa varied between seasons and after cryopreservation. This study highlights the advantages of integrating IFC and artificial intelligence (AI) algorithms for robust, high-throughput, and objective label-free spermatozoa morphology assessment in fresh and cryopreserved sperm, offering a promising tool for improving fertility diagnostics and breeding strategies in veterinary practice.