P-093 Single sperm morphokinetic variables during ICSI at the time of sperm aspiration into the microneedle
Bibliographic record
Abstract
Abstract Study question Is it possible to classify sperm morphology as normal or abnormal based on their kinetic characteristics? Summary answer Using different classifiers, it was found that balanced dataset the classification scores of individual sperm morphology can be optimized during their motility just before ICSI What is known already Sperm selection plays a crucial role during ICSI. This selection by the embryologists is accompanied by two factors: motility parameters of the sperm to be injected, as well as the morphology. Motility affects morphological decision-making which may be subjective due to spermatozoa not being in a narrow vertical space in a PVP droplet, i.e. counting chamber. Machine/Deep Learning models -as classifiers- are gaining popularity in IVF as it could minimize subjectivity in gamete selection. SiD software is an algorithm that from a group of spermatozoa characteristics can support the selection of a single sperm during real time ICSI. Study design, size, duration In this prospective study 1699 individual spermatozoa were video-recorded (resolution of 200 X 200 pix.) during sperm selection in a 7%PVP solution. Motility variables (VSL, VCL, LIN, VAP, ALH, WOB, STR, MAD) were obtained from each video using the software SiD. Each sperm was classified as normal or abnormal. Based on motility variables, different classification models were used to make a morphokinetic association of the variables with the classification of each sperm. Participants/materials, setting, methods 1699 individual spermatozoa were classified and labeled by three senior embryologists into two categories: normal or anormal. To belong to any category, at least two embryologists had to agree. A normalization was applied to the motility data obtained with SiD. Machine learning classification models were applied. The three classification models with the best results were selected to optimize the hyper-parameters and improve their performance. Main results and the role of chance A set of motility variables were obtained using the software SiD1 (IVF2.0 Ltd., UK) these were used as features for each of the sperm samples and were tagged either as normal or abnormal having a total of 257 normal samples and 1442 abnormal samples. Different classification algorithms were used to perform sperm classification as normal and abnormal. Then hyperparameter tuning algorithms such as GridSearch were used to compute the optimum values, finding KNN algorithm to achieve the best results to classify normal and abnormal spermatozoa with 80% accuracy. Given the nature of the unbalanced dataset, the F1 score was calculated, achieving 0.77, other algorithms such as Decision-Tree and Random-Forest were used achieving similar results in the F1 score. We will continue growing the dataset until it is balanced and run the same algorithms expecting the performance to improve. Limitations, reasons for caution Each clinic's laboratory setup, including the camera used to record the ICSI procedure, is a clinic-specific configuration. Sperm abnormality detection is sensitive to camera resolution, showing the classifiers are resolution-dependent observers. To replicate these results, it is important to consider the quality of the camera. Wider implications of the findings Using the selected kinematics features, it is possible to classify individual spermatozoa during motility as having normal or abnormal morphology. Results may improve using standard morphological examination of each population and having more normal sperm samples. Trial registration number Not applicable
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".