Bull Sperm Morphology Assessment: Agreement Between AI Station and Trumorph

This study investigated the workflow-level agreement between AI Station and Trumorph systems under practical conditions in bull sperm morphology assessment in 41 ejaculates from 13 Brahman bulls. Agreement was evaluated using mixed-effects models, Bland-Altman plots, and Passing-Bablok regression, applying a ±20% acceptance threshold. Although both systems yielded similar mean values for normal morphology, Bland-Altman analysis demonstrated wide limits of agreement, indicating limited interchangeability at the individual-sample level and substantial disagreement for specific morphoanomalies. AI Station yielded higher estimates of proximal and distal cytoplasmic droplets, whereas Trumorph produced higher estimates of bent-tail and coiled-tail defects. Several traits showed evidence of constant or proportional bias, pointing to systematic differences between the two methods. Overall, these findings indicate that the two systems are not interchangeable under the analytical conditions evaluated here and require platform-specific validation before routine diagnostic or classification use.