Deep Learning-based Automated Rare Sperm Identification from Testes Biopsies
Bibliographic record
Abstract
ABSTRACT Non-obstructive azoospermia (NOA), the most severe form of male infertility, is currently treated using microsurgical sperm extraction (microTESE) to retrieve sperm cells for in vitro fertilization via intracytoplasmic sperm injection (IVF-ICSI). The success rate of this procedure for NOA patients is currently limited by the ability of andrologists to identify a few rare sperm cells among millions of background testis cells. To improve this success rate, we developed a convolution neural network (CNN) to detect rare sperm from low-resolution microscopy images of microTESE samples. Our CNN uses the U-Net architecture to perform pixel-based classification on image patches from brightfield microscopy, which is followed by morphological analysis to detect individual sperm instances. This CNN is trained using microscopy images of fluorescently labeled sperm, which is fixed to eliminate their motility, and doped into testis biopsies obtained from NOA patients. We initially tested this algorithm using purified sperm samples at different imaging magnifications in order to determine the upper bounds of performance. We then tested this algorithm by doping rare sperm cells into testis biopsy samples from NOA patients and found a sperm detection F1 score of 85.2%. These results demonstrate the potential to use automated microscopy to dramatically increase the amount of testis biopsy tissue that could be comprehensively examined, which greatly increases the chance of finding rare viable sperm, and thereby increases the success rates of IVF-ICSI for couples with NOA.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 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 source (direct Gemma or distilled Codex), 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".