P-093 Single sperm morphokinetic variables during ICSI at the time of sperm aspiration into the microneedle

Abstract Study question Is it possible to classify sperm morphology as normal or abnormal based on their kinetic characteristics? Summary answer Using different classifiers, was found that balanced dataset the classification scores of individual can be optimized during motility just before ICSI What is known already Sperm selection plays a crucial role ICSI. This by embryologists accompanied two factors: parameters injected, well morphology. Motility affects morphological decision-making which may subjective due spermatozoa not being in narrow vertical space PVP droplet, i.e. counting chamber. Machine/Deep Learning models -as classifiers- are gaining popularity IVF could minimize subjectivity gamete selection. SiD software an algorithm from group characteristics support single real time design, size, duration In this prospective study 1699 were video-recorded (resolution 200 X pix.) 7%PVP solution. variables (VSL, VCL, LIN, VAP, ALH, WOB, STR, MAD) obtained each video using SiD. Each classified abnormal. Based variables, used make morphokinetic association with sperm. Participants/materials, setting, methods and labeled three senior into categories: anormal. To belong any category, at least had agree. A normalization applied data Machine learning applied. The best results selected optimize hyper-parameters improve performance. Main chance set SiD1 (IVF2.0 Ltd., UK) these features for samples tagged either having total 257 1442 samples. Different algorithms perform Then hyperparameter tuning such GridSearch compute optimum values, finding KNN achieve 80% accuracy. Given nature unbalanced dataset, F1 score calculated, achieving 0.77, other Decision-Tree Random-Forest similar score. We will continue growing until run same expecting performance improve. Limitations, reasons caution clinic's laboratory setup, including camera record procedure, clinic-specific configuration. abnormality detection sensitive resolution, showing classifiers resolution-dependent observers. replicate results, important consider quality camera. Wider implications findings kinematics features, Results standard examination population more Trial registration number Not applicable